Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -3,11 +3,26 @@ import torch
|
|
3 |
from transformers import AutoTokenizer, AutoModelForSequenceClassification
|
4 |
import logging
|
5 |
import re
|
|
|
|
|
|
|
|
|
|
|
|
|
6 |
|
7 |
-
# Configure logging
|
8 |
logging.basicConfig(level=logging.INFO)
|
9 |
logger = logging.getLogger(__name__)
|
10 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
11 |
# Model configuration
|
12 |
MODEL_NAME = "cybersectony/phishing-email-detection-distilbert_v2.4.1"
|
13 |
|
@@ -26,6 +41,50 @@ Higher percentages indicate greater certainty.
|
|
26 |
tokenizer = None
|
27 |
model = None
|
28 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
29 |
def load_model():
|
30 |
"""Load the model and tokenizer with basic error handling"""
|
31 |
global tokenizer, model
|
@@ -49,9 +108,9 @@ def is_valid_email_text(text):
|
|
49 |
return True, ""
|
50 |
|
51 |
def get_colored_bar(percentage):
|
52 |
-
"""Create a
|
53 |
if percentage >= 85:
|
54 |
-
color = "🟢"
|
55 |
elif percentage >= 50:
|
56 |
color = "🟡"
|
57 |
else:
|
@@ -59,8 +118,11 @@ def get_colored_bar(percentage):
|
|
59 |
bar_length = max(1, int(percentage / 5)) # Scale to 20 characters
|
60 |
return color * bar_length + "⚪" * (20 - bar_length)
|
61 |
|
62 |
-
def predict_email(email_text):
|
63 |
-
"""
|
|
|
|
|
|
|
64 |
# Input validation
|
65 |
valid, message = is_valid_email_text(email_text)
|
66 |
if not valid:
|
@@ -97,7 +159,251 @@ def predict_email(email_text):
|
|
97 |
# Get top prediction
|
98 |
max_label, max_prob = max(results.items(), key=lambda x: x[1])
|
99 |
|
100 |
-
# Risk levels with
|
101 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
102 |
|
|
|
|
|
|
|
103 |
|
|
|
|
|
|
3 |
from transformers import AutoTokenizer, AutoModelForSequenceClassification
|
4 |
import logging
|
5 |
import re
|
6 |
+
from flask import Flask, request, jsonify, render_template_string, redirect, url_for
|
7 |
+
from flask_login import LoginManager, UserMixin, login_required, login_user, logout_user, current_user
|
8 |
+
import sqlite3
|
9 |
+
from functools import wraps
|
10 |
+
from datetime import datetime
|
11 |
+
import bleach
|
12 |
|
13 |
+
# Configure logging
|
14 |
logging.basicConfig(level=logging.INFO)
|
15 |
logger = logging.getLogger(__name__)
|
16 |
|
17 |
+
# Flask app setup
|
18 |
+
app = Flask(__name__)
|
19 |
+
app.secret_key = 'secure_gov_key_2025' # Replace with a secure key in production
|
20 |
+
|
21 |
+
# Initialize Flask-Login
|
22 |
+
login_manager = LoginManager()
|
23 |
+
login_manager.init_app(app)
|
24 |
+
login_manager.login_view = "login"
|
25 |
+
|
26 |
# Model configuration
|
27 |
MODEL_NAME = "cybersectony/phishing-email-detection-distilbert_v2.4.1"
|
28 |
|
|
|
41 |
tokenizer = None
|
42 |
model = None
|
43 |
|
44 |
+
# User class for Flask-Login
|
45 |
+
class User(UserMixin):
|
46 |
+
def __init__(self, user_id, role):
|
47 |
+
self.id = user_id
|
48 |
+
self.role = role
|
49 |
+
|
50 |
+
# Database setup
|
51 |
+
def init_db():
|
52 |
+
with sqlite3.connect('phishguardian.db') as conn:
|
53 |
+
c = conn.cursor()
|
54 |
+
c.execute('''CREATE TABLE IF NOT EXISTS users
|
55 |
+
(id TEXT PRIMARY KEY, username TEXT, password TEXT, role TEXT)''')
|
56 |
+
c.execute('''CREATE TABLE IF NOT EXISTS analysis_logs
|
57 |
+
(id INTEGER PRIMARY KEY AUTOINCREMENT, user_id TEXT, email_text TEXT,
|
58 |
+
result TEXT, timestamp TEXT)''')
|
59 |
+
# Add default admin user (password: 'admin123' for demo, use hashed passwords in production)
|
60 |
+
c.execute("INSERT OR IGNORE INTO users (id, username, password, role) VALUES (?, ?, ?, ?)",
|
61 |
+
('admin1', 'admin', 'admin123', 'Admin'))
|
62 |
+
conn.commit()
|
63 |
+
|
64 |
+
# Load user for Flask-Login
|
65 |
+
@login_manager.user_loader
|
66 |
+
def load_user(user_id):
|
67 |
+
with sqlite3.connect('phishguardian.db') as conn:
|
68 |
+
c = conn.cursor()
|
69 |
+
c.execute("SELECT id, role FROM users WHERE id = ?", (user_id,))
|
70 |
+
user = c.fetchone()
|
71 |
+
if user:
|
72 |
+
return User(user[0], user[1])
|
73 |
+
return None
|
74 |
+
|
75 |
+
# RBAC decorator
|
76 |
+
def role_required(*roles):
|
77 |
+
def decorator(f):
|
78 |
+
@wraps(f)
|
79 |
+
def decorated_function(*args, **kwargs):
|
80 |
+
if not current_user.is_authenticated:
|
81 |
+
return redirect(url_for('login'))
|
82 |
+
if current_user.role not in roles:
|
83 |
+
return render_template_string("<h1>403 Forbidden</h1><p>Unauthorized role.</p>")
|
84 |
+
return f(*args, **kwargs)
|
85 |
+
return decorated_function
|
86 |
+
return decorator
|
87 |
+
|
88 |
def load_model():
|
89 |
"""Load the model and tokenizer with basic error handling"""
|
90 |
global tokenizer, model
|
|
|
108 |
return True, ""
|
109 |
|
110 |
def get_colored_bar(percentage):
|
111 |
+
"""Create a colored bar based on percentage"""
|
112 |
if percentage >= 85:
|
113 |
+
color = "🟢"
|
114 |
elif percentage >= 50:
|
115 |
color = "🟡"
|
116 |
else:
|
|
|
118 |
bar_length = max(1, int(percentage / 5)) # Scale to 20 characters
|
119 |
return color * bar_length + "⚪" * (20 - bar_length)
|
120 |
|
121 |
+
def predict_email(email_text, user_id):
|
122 |
+
"""Prediction with actual labels and colored bars"""
|
123 |
+
# Sanitize input
|
124 |
+
email_text = bleach.clean(email_text, tags=[], strip=True)
|
125 |
+
|
126 |
# Input validation
|
127 |
valid, message = is_valid_email_text(email_text)
|
128 |
if not valid:
|
|
|
159 |
# Get top prediction
|
160 |
max_label, max_prob = max(results.items(), key=lambda x: x[1])
|
161 |
|
162 |
+
# Risk levels with 85% threshold
|
163 |
+
if "phishing" in max_label.lower() or "suspicious" in max_label.lower():
|
164 |
+
risk_level = "⚠️ Risky" if max_prob >= 85 else "⚡ Low Risk"
|
165 |
+
elif "spam" in max_label.lower():
|
166 |
+
risk_level = "🗑️ Spam" if max_prob >= 85 else "⚡ Low Risk"
|
167 |
+
else:
|
168 |
+
risk_level = "✅ Safe" if max_prob >= 85 else "❓ Uncertain"
|
169 |
+
|
170 |
+
# Format output
|
171 |
+
output = f"Result: {risk_level}\n"
|
172 |
+
output += f"Top Prediction: {max_label} ({max_prob:.1f}%)\n"
|
173 |
+
output += "Details:\n"
|
174 |
+
for label, prob in sorted(results.items(), key=lambda x: x[1], reverse=True):
|
175 |
+
output += f"{label}: {prob:.1f}% {get_colored_bar(prob)}\n"
|
176 |
+
|
177 |
+
# Simple recommendation
|
178 |
+
if "phishing" in max_label.lower() or "suspicious" in max_label.lower():
|
179 |
+
output += "Advice: Avoid clicking links or sharing info."
|
180 |
+
elif "spam" in max_label.lower():
|
181 |
+
output += "Advice: Mark as spam or delete."
|
182 |
+
else:
|
183 |
+
output += "Advice: Appears safe, but stay cautious."
|
184 |
+
|
185 |
+
# Log analysis
|
186 |
+
with sqlite3.connect('phishguardian.db') as conn:
|
187 |
+
c = conn.cursor()
|
188 |
+
c.execute("INSERT INTO analysis_logs (user_id, email_text, result, timestamp) VALUES (?, ?, ?, ?)",
|
189 |
+
(user_id, email_text[:1000], output, datetime.utcnow().isoformat()))
|
190 |
+
conn.commit()
|
191 |
+
|
192 |
+
return output
|
193 |
+
|
194 |
+
except Exception as e:
|
195 |
+
logger.error(f"Error during prediction: {e}")
|
196 |
+
return f"❌ Error: Analysis failed - {str(e)}"
|
197 |
+
|
198 |
+
# Flask routes
|
199 |
+
@app.route('/')
|
200 |
+
def index():
|
201 |
+
if not current_user.is_authenticated:
|
202 |
+
return redirect(url_for('login'))
|
203 |
+
return render_template_string("""
|
204 |
+
<!DOCTYPE html>
|
205 |
+
<html>
|
206 |
+
<head>
|
207 |
+
<title>PhishGuardian - MDA Email System</title>
|
208 |
+
<style>
|
209 |
+
body { font-family: Arial, sans-serif; margin: 20px; background-color: #f4f4f9; }
|
210 |
+
h1 { color: #333; }
|
211 |
+
.container { max-width: 800px; margin: auto; }
|
212 |
+
textarea { width: 100%; height: 200px; margin-bottom: 10px; }
|
213 |
+
button { padding: 10px 20px; margin-right: 10px; }
|
214 |
+
pre { background-color: #fff; padding: 15px; border: 1px solid #ddd; }
|
215 |
+
.error { color: red; }
|
216 |
+
</style>
|
217 |
+
</head>
|
218 |
+
<body>
|
219 |
+
<div class="container">
|
220 |
+
<h1>🛡️ PhishGuardian - MDA Email System</h1>
|
221 |
+
<p>Analyze emails for safety. Paste email text below.</p>
|
222 |
+
<p><b>Labels</b>: Legitimate (safe), Phishing (scam), Suspicious (questionable), Spam (junk). Percentages show confidence (0-100%).</p>
|
223 |
+
{% if current_user.is_authenticated %}
|
224 |
+
<p>Logged in as: {{ current_user.id }} ({{ current_user.role }}) | <a href="{{ url_for('logout') }}">Logout</a></p>
|
225 |
+
{% if current_user.role in ['Admin', 'Analyst'] %}
|
226 |
+
<form method="POST" action="{{ url_for('analyze') }}">
|
227 |
+
<textarea name="email_text" placeholder="Paste email here..."></textarea>
|
228 |
+
<button type="submit">🔍 Check</button>
|
229 |
+
<button type="button" onclick="document.querySelector('textarea').value=''">🗑️ Clear</button>
|
230 |
+
</form>
|
231 |
+
{% if result %}
|
232 |
+
<h3>Results</h3>
|
233 |
+
<pre>{{ result }}</pre>
|
234 |
+
{% endif %}
|
235 |
+
{% endif %}
|
236 |
+
{% if current_user.role in ['Admin', 'Auditor'] %}
|
237 |
+
<p><a href="{{ url_for('view_logs') }}">View Analysis Logs</a></p>
|
238 |
+
{% endif %}
|
239 |
+
{% if current_user.role == 'Admin' %}
|
240 |
+
<p><a href="{{ url_for('manage_users') }}">Manage Users</a></p>
|
241 |
+
{% endif %}
|
242 |
+
{% endif %}
|
243 |
+
</div>
|
244 |
+
</body>
|
245 |
+
</html>
|
246 |
+
""", result=request.args.get('result', ''))
|
247 |
+
|
248 |
+
@app.route('/login', methods=['GET', 'POST'])
|
249 |
+
def login():
|
250 |
+
if request.method == 'POST':
|
251 |
+
username = bleach.clean(request.form['username'])
|
252 |
+
password = bleach.clean(request.form['password'])
|
253 |
+
with sqlite3.connect('phishguardian.db') as conn:
|
254 |
+
c = conn.cursor()
|
255 |
+
c.execute("SELECT id, role FROM users WHERE username = ? AND password = ?", (username, password))
|
256 |
+
user = c.fetchone()
|
257 |
+
if user:
|
258 |
+
login_user(User(user[0], user[1]))
|
259 |
+
return redirect(url_for('index'))
|
260 |
+
return render_template_string("<h1>Login Failed</h1><p>Invalid credentials.</p>")
|
261 |
+
return render_template_string("""
|
262 |
+
<!DOCTYPE html>
|
263 |
+
<html>
|
264 |
+
<head>
|
265 |
+
<title>Login - PhishGuardian</title>
|
266 |
+
<style>
|
267 |
+
body { font-family: Arial, sans-serif; margin: 20px; background-color: #f4f4f9; }
|
268 |
+
.container { max-width: 400px; margin: auto; }
|
269 |
+
input { width: 100%; padding: 10px; margin-bottom: 10px; }
|
270 |
+
button { padding: 10px 20px; }
|
271 |
+
</style>
|
272 |
+
</head>
|
273 |
+
<body>
|
274 |
+
<div class="container">
|
275 |
+
<h1>Login</h1>
|
276 |
+
<form method="POST">
|
277 |
+
<input type="text" name="username" placeholder="Username" required>
|
278 |
+
<input type="password" name="password" placeholder="Password" required>
|
279 |
+
<button type="submit">Login</button>
|
280 |
+
</form>
|
281 |
+
</div>
|
282 |
+
</body>
|
283 |
+
</html>
|
284 |
+
""")
|
285 |
+
|
286 |
+
@app.route('/logout')
|
287 |
+
@login_required
|
288 |
+
def logout():
|
289 |
+
logout_user()
|
290 |
+
return redirect(url_for('login'))
|
291 |
+
|
292 |
+
@app.route('/analyze', methods=['POST'])
|
293 |
+
@role_required('Admin', 'Analyst')
|
294 |
+
def analyze():
|
295 |
+
email_text = request.form['email_text']
|
296 |
+
result = predict_email(email_text, current_user.id)
|
297 |
+
return redirect(url_for('index', result=result))
|
298 |
+
|
299 |
+
@app.route('/logs')
|
300 |
+
@role_required('Admin', 'Auditor')
|
301 |
+
def view_logs():
|
302 |
+
with sqlite3.connect('phishguardian.db') as conn:
|
303 |
+
c = conn.cursor()
|
304 |
+
c.execute("SELECT user_id, email_text, result, timestamp FROM analysis_logs ORDER BY timestamp DESC")
|
305 |
+
logs = c.fetchall()
|
306 |
+
logs_html = "<h3>Analysis Logs</h3><ul>" + "".join(
|
307 |
+
f"<li><b>{log[3]}</b> | User: {log[0]} | Email: {log[1][:50]}... | Result: {log[2][:100]}...</li>"
|
308 |
+
for log in logs
|
309 |
+
) + "</ul>"
|
310 |
+
return render_template_string("""
|
311 |
+
<!DOCTYPE html>
|
312 |
+
<html>
|
313 |
+
<head>
|
314 |
+
<title>Logs - PhishGuardian</title>
|
315 |
+
<style>
|
316 |
+
body { font-family: Arial, sans-serif; margin: 20px; background-color: #f4f4f9; }
|
317 |
+
.container { max-width: 800px; margin: auto; }
|
318 |
+
ul { list-style-type: none; padding: 0; }
|
319 |
+
li { margin-bottom: 10px; }
|
320 |
+
</style>
|
321 |
+
</head>
|
322 |
+
<body>
|
323 |
+
<div class="container">
|
324 |
+
<h1>Analysis Logs</h1>
|
325 |
+
<p><a href="{{ url_for('index') }}">Back to Home</a></p>
|
326 |
+
{{ logs_html | safe }}
|
327 |
+
</div>
|
328 |
+
</body>
|
329 |
+
</html>
|
330 |
+
""", logs_html=logs_html)
|
331 |
+
|
332 |
+
@app.route('/users', methods=['GET', 'POST'])
|
333 |
+
@role_required('Admin')
|
334 |
+
def manage_users():
|
335 |
+
if request.method == 'POST':
|
336 |
+
username = bleach.clean(request.form['username'])
|
337 |
+
password = bleach.clean(request.form['password'])
|
338 |
+
role = bleach.clean(request.form['role'])
|
339 |
+
user_id = f"user_{datetime.utcnow().timestamp()}"
|
340 |
+
with sqlite3.connect('phishguardian.db') as conn:
|
341 |
+
c = conn.cursor()
|
342 |
+
c.execute("INSERT INTO users (id, username, password, role) VALUES (?, ?, ?, ?)",
|
343 |
+
(user_id, username, password, role))
|
344 |
+
conn.commit()
|
345 |
+
return redirect(url_for('manage_users'))
|
346 |
+
with sqlite3.connect('phishguardian.db') as conn:
|
347 |
+
c = conn.cursor()
|
348 |
+
c.execute("SELECT id, username, role FROM users")
|
349 |
+
users = c.fetchall()
|
350 |
+
users_html = "<h3>Users</h3><ul>" + "".join(
|
351 |
+
f"<li>ID: {user[0]} | Username: {user[1]} | Role: {user[2]}</li>" for user in users
|
352 |
+
) + "</ul>"
|
353 |
+
return render_template_string("""
|
354 |
+
<!DOCTYPE html>
|
355 |
+
<html>
|
356 |
+
<head>
|
357 |
+
<title>Manage Users - PhishGuardian</title>
|
358 |
+
<style>
|
359 |
+
body { font-family: Arial, sans-serif; margin: 20px; background-color: #f4f4f9; }
|
360 |
+
.container { max-width: 800px; margin: auto; }
|
361 |
+
input, select { width: 100%; padding: 10px; margin-bottom: 10px; }
|
362 |
+
button { padding: 10px 20px; }
|
363 |
+
ul { list-style-type: none; padding: 0; }
|
364 |
+
</style>
|
365 |
+
</head>
|
366 |
+
<body>
|
367 |
+
<div class="container">
|
368 |
+
<h1>Manage Users</h1>
|
369 |
+
<p><a href="{{ url_for('index') }}">Back to Home</a></p>
|
370 |
+
<form method="POST">
|
371 |
+
<input type="text" name="username" placeholder="Username" required>
|
372 |
+
<input type="password" name="password" placeholder="Password" required>
|
373 |
+
<select name="role">
|
374 |
+
<option value="Admin">Admin</option>
|
375 |
+
<option value="Analyst">Analyst</option>
|
376 |
+
<option value="Auditor">Auditor</option>
|
377 |
+
</select>
|
378 |
+
<button type="submit">Add User</button>
|
379 |
+
</form>
|
380 |
+
{{ users_html | safe }}
|
381 |
+
</div>
|
382 |
+
</body>
|
383 |
+
</html>
|
384 |
+
""", users_html=users_html)
|
385 |
+
|
386 |
+
# Example emails
|
387 |
+
example_legitimate = """Dear Customer,
|
388 |
+
Thank you for your purchase from TechStore. Your order #ORD-2024-001234 is processed.
|
389 |
+
Order Details:
|
390 |
+
- Product: Wireless Headphones
|
391 |
+
- Amount: $79.99
|
392 |
+
- Delivery: 3-5 days
|
393 |
+
Best regards,
|
394 |
+
TechStore"""
|
395 |
+
example_phishing = """URGENT!!!
|
396 |
+
Your account is COMPROMISED! Click here to secure: http://fake-site.com/verify
|
397 |
+
Act NOW or your account will be suspended!
|
398 |
+
Security Team"""
|
399 |
+
example_neutral = """Hi team,
|
400 |
+
Reminder: meeting tomorrow at 2 PM. Bring project updates.
|
401 |
+
Thanks,
|
402 |
+
Sarah"""
|
403 |
|
404 |
+
# Initialize database and load model
|
405 |
+
init_db()
|
406 |
+
load_model()
|
407 |
|
408 |
+
if __name__ == "__main__":
|
409 |
+
app.run(ssl_context='adhoc', host='0.0.0.0', port=5000)
|