Spaces:
Running
Running
File size: 20,046 Bytes
3ad73e3 | 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 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 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 | # ============================================================
# PhishGuard v5 β HuggingFace Spaces Deployment
# File : app.py
# Port : 7860 (required by HuggingFace)
#
# Endpoints:
# GET / β health check
# POST /predict β predict if URL is phishing or legitimate
# POST /predict/batch β predict multiple URLs at once
#
# Features: 48 hybrid features
# 42 structural + 2 knowledge-based + 4 pure structural
#
# Author : Uzman Zahid
# Dublin Business School β 2026
# ============================================================
from fastapi import FastAPI, HTTPException
from fastapi.middleware.cors import CORSMiddleware
from pydantic import BaseModel
import joblib
import json
import re
import math
import pandas as pd
import os
# ββ 1. INITIALISE APP ββββββββββββββββββββββββββββββββββββββββ
app = FastAPI(
title = "PhishGuard v5 API",
description = "Real-time phishing URL detection using Machine Learning β 48 hybrid features β Uzman Zahid, Dublin Business School 2026",
version = "5.0.0"
)
# ββ 2. CORS βββββββββββββββββββββββββββββββββββββββββββββββββββ
app.add_middleware(
CORSMiddleware,
allow_origins = ["*"],
allow_credentials = True,
allow_methods = ["*"],
allow_headers = ["*"],
)
# ββ 3. LOAD MODEL ββββββββββββββββββββββββββββββββββββββββββββ
print("Loading PhishGuard v5 model...")
model_path = "model_compressed.pkl" if os.path.exists("model_compressed.pkl") else "model.pkl"
try:
model = joblib.load(model_path)
print(f"β
Model loaded from {model_path}")
except Exception as e:
print(f"β Error loading model: {e}")
raise
try:
with open("features.json") as f:
features = json.load(f)
print(f"β
Features loaded: {len(features)} features")
except Exception as e:
print(f"β Error loading features: {e}")
raise
# ββ 4. KNOWLEDGE LOOKUP TABLES βββββββββββββββββββββββββββββββ
# Based on APWG Phishing Activity Trends Reports (2016-2024)
SUSPICIOUS_TLDS = {
'tk', 'ml', 'ga', 'cf', 'gq',
'xyz', 'top', 'club', 'online', 'site',
'fun', 'icu', 'vip', 'cyou', 'lat',
'space', 'live', 'pw', 'cc', 'su',
'ws', 'bz', 'name', 'mobi', 'link',
'click', 'download', 'loan', 'win',
'racing', 'stream', 'trade', 'review',
'accountant', 'science', 'work', 'party',
'faith', 'date', 'cricket', 'ninja',
'bid', 'webcam', 'rocks', 'country',
}
# Well-established legitimate second-level domains
KNOWN_SAFE_SLDS = {
'google', 'microsoft', 'apple', 'amazon',
'facebook', 'youtube', 'netflix', 'github',
'stackoverflow', 'wikipedia', 'twitter',
'linkedin', 'instagram', 'spotify', 'stripe',
'paypal', 'dropbox', 'slack', 'zoom',
'cloudflare', 'netlify', 'vercel', 'heroku',
'reddit', 'pinterest', 'tiktok', 'discord',
'notion', 'figma', 'canva', 'shopify',
'wordpress', 'adobe', 'salesforce', 'hubspot',
'openai', 'anthropic', 'huggingface', 'kaggle',
'coursera', 'udemy', 'edx', 'khanacademy',
'duolingo', 'bbc', 'cnn', 'reuters', 'bloomberg',
'techcrunch', 'mit', 'stanford', 'harvard',
'oxford', 'cambridge', 'dbs', 'ucd', 'tcd', 'dcu',
'npmjs', 'pypi', 'kubernetes', 'mozilla', 'docker',
'spinbot', 'ilovepdf', 'smallpdf', 'mp3cut',
'convertio', 'tinypng', 'grammarly',
'wise', 'revolut', 'coinbase', 'binance',
'vimeo', 'twitch', 'dailymotion', 'soundcloud',
'gov', 'nasa', 'digitalocean',
}
# ββ 5. REQUEST / RESPONSE MODELS βββββββββββββββββββββββββββββ
class URLRequest(BaseModel):
url: str
class PredictionResponse(BaseModel):
url : str
prediction : int
label : str
confidence : float
message : str
# ββ 6. FEATURE EXTRACTION (48 HYBRID FEATURES) βββββββββββββββ
def extract_features(raw_url):
"""
Extracts 48 hybrid features from any URL format.
42 structural + 2 knowledge-based + 4 pure structural.
Identical to feature extraction used during training.
"""
try:
url = str(raw_url).strip()
# ββ Scheme ββββββββββββββββββββββββββββββββββββββββββββ
has_https = 1 if url.lower().startswith('https://') else 0
has_http = 1 if url.lower().startswith('http://') else 0
# ββ Parse components ββββββββββββββββββββββββββββββββββ
url_no_scheme = url
if '://' in url:
url_no_scheme = url.split('://', 1)[1]
domain_with_port = url_no_scheme.split('/')[0].split('?')[0].split('#')[0]
domain_clean = domain_with_port.split(':')[0]
parts = url_no_scheme.split('/', 1)
path = '/' + parts[1] if len(parts) > 1 else ''
query = url.split('?', 1)[1].split('#')[0] if '?' in url else ''
fragment = url.split('#', 1)[1] if '#' in url else ''
domain_no_www = domain_clean
if domain_no_www.lower().startswith('www.'):
domain_no_www = domain_no_www[4:]
if domain_clean.lower().startswith('www.'):
domain_normalised = domain_clean
else:
domain_normalised = 'www.' + domain_clean
domain_parts_full = domain_normalised.split('.')
# ββ TLD and SLD βββββββββββββββββββββββββββββββββββββββ
domain_parts = domain_no_www.split('.')
tld = domain_parts[-1].lower() if domain_parts else ''
sld = domain_parts[-2].lower() if len(domain_parts) >= 2 else ''
if len(domain_parts) >= 3 and domain_parts[-1] in [
'uk','au','in','jp','nz','za','br','sg','ie','pk'
]:
tld = f"{domain_parts[-2]}.{domain_parts[-1]}"
sld = domain_parts[-3].lower() if len(domain_parts) >= 3 else ''
# ββ URL features ββββββββββββββββββββββββββββββββββββββ
url_length = len(url)
number_of_dots_in_url = url.count('.')
digits_in_url = re.findall(r'\d', url)
having_repeated_digits_in_url = 1 if len(digits_in_url) != len(
set(digits_in_url)) and len(digits_in_url) > 0 else 0
number_of_digits_in_url = sum(c.isdigit() for c in url)
number_of_special_char_in_url = sum(not c.isalnum() for c in url)
number_of_hyphens_in_url = url.count('-')
number_of_underline_in_url = url.count('_')
number_of_slash_in_url = url.count('/')
number_of_questionmark_in_url = url.count('?')
number_of_equal_in_url = url.count('=')
number_of_at_in_url = url.count('@')
number_of_dollar_in_url = url.count('$')
number_of_exclamation_in_url = url.count('!')
number_of_hashtag_in_url = url.count('#')
number_of_percent_in_url = url.count('%')
# ββ Domain features βββββββββββββββββββββββββββββββββββ
domain_length = len(domain_no_www)
number_of_dots_in_domain = domain_no_www.count('.')
number_of_hyphens_in_domain = domain_no_www.count('-')
having_special_characters_in_domain = 1 if re.search(
r'[^a-zA-Z0-9\.\-]', domain_no_www) else 0
number_of_special_characters_in_domain = sum(
not c.isalnum() and c not in '.-' for c in domain_no_www)
having_digits_in_domain = 1 if any(
c.isdigit() for c in domain_no_www) else 0
number_of_digits_in_domain = sum(
c.isdigit() for c in domain_no_www)
digits_in_domain = re.findall(r'\d', domain_no_www)
having_repeated_digits_in_domain = 1 if len(
digits_in_domain) != len(set(digits_in_domain)) and \
len(digits_in_domain) > 0 else 0
# ββ Subdomain features ββββββββββββββββββββββββββββββββ
number_of_subdomains = max(0, len(domain_parts_full) - 2)
subdomains = domain_parts_full[:-2] if len(
domain_parts_full) > 2 else []
subdomain_depth = len(subdomains)
having_hyphen_in_subdomain = 1 if any(
'-' in s for s in subdomains) else 0
average_subdomain_length = sum(
len(s) for s in subdomains) / len(subdomains) \
if subdomains else 0.0
average_number_of_hyphens_in_subdomain = sum(
s.count('-') for s in subdomains) / len(subdomains) \
if subdomains else 0.0
having_special_characters_in_subdomain = 1 if any(
re.search(r'[^a-zA-Z0-9\-]', s)
for s in subdomains) else 0
number_of_special_characters_in_subdomain = sum(
sum(not c.isalnum() and c != '-' for c in s)
for s in subdomains)
having_digits_in_subdomain = 1 if any(
any(c.isdigit() for c in s) for s in subdomains) else 0
number_of_digits_in_subdomain = sum(
sum(c.isdigit() for c in s) for s in subdomains)
all_sub_digits = re.findall(
r'\d', ''.join(subdomains))
having_repeated_digits_in_subdomain = 1 if len(
all_sub_digits) != len(set(all_sub_digits)) and \
len(all_sub_digits) > 0 else 0
# ββ Path/Query features βββββββββββββββββββββββββββββββ
having_path = 1 if len(path) > 1 else 0
path_segments = [p for p in path.split('/') if p]
path_length = len(path_segments)
having_query = 1 if len(query) > 0 else 0
having_fragment = 1 if len(fragment) > 0 else 0
having_anchor = 1 if '#' in url else 0
# ββ Entropy features ββββββββββββββββββββββββββββββββββ
if len(url) > 0:
prob_url = [url.count(c)/len(url) for c in set(url)]
entropy_of_url = -sum(p*math.log2(p) for p in prob_url if p > 0)
else:
entropy_of_url = 0.0
if len(domain_no_www) > 0:
prob_dom = [domain_no_www.count(c)/len(domain_no_www)
for c in set(domain_no_www)]
entropy_of_domain = -sum(p*math.log2(p) for p in prob_dom if p > 0)
else:
entropy_of_domain = 0.0
# ββ NEW FEATURE 43: has_suspicious_tld ββββββββββββββββ
# Knowledge-based: APWG documented high-abuse TLDs
has_suspicious_tld = 1 if tld.lower() in SUSPICIOUS_TLDS else 0
# ββ NEW FEATURE 44: is_known_safe_sld βββββββββββββββββ
# Knowledge-based: established legitimate platforms
is_known_safe_sld = 1 if sld.lower() in KNOWN_SAFE_SLDS else 0
# ββ NEW FEATURE 45: consonant_vowel_ratio βββββββββββββ
# Pure structural: unnatural domains = phishing signal
vowels = set('aeiouAEIOU')
letters = [c for c in domain_no_www if c.isalpha()]
vowel_count = sum(1 for c in letters if c in vowels)
consonant_count = sum(1 for c in letters if c not in vowels)
consonant_vowel_ratio = round(
consonant_count / (vowel_count + 1), 4)
# ββ NEW FEATURE 46: longest_digit_sequence ββββββββββββ
# Pure structural: digit runs indicate random generation
digit_sequences = re.findall(r'\d+', domain_no_www)
longest_digit_seq = max(
(len(s) for s in digit_sequences), default=0)
# ββ NEW FEATURE 47: digit_letter_ratio ββββββββββββββββ
# Pure structural: digit-heavy domains = phishing
alpha_count = sum(c.isalpha() for c in domain_no_www)
digit_count = sum(c.isdigit() for c in domain_no_www)
digit_letter_ratio = round(
digit_count / (alpha_count + 1), 4)
# ββ NEW FEATURE 48: path_to_url_ratio βββββββββββββββββ
# Pure structural: bare phishing domains have ratio = 0
path_to_url_ratio = round(
len(path) / len(url), 4) if len(url) > 0 else 0.0
return {
'has_https' : has_https,
'has_http' : has_http,
'url_length' : url_length,
'number_of_dots_in_url' : number_of_dots_in_url,
'having_repeated_digits_in_url' : having_repeated_digits_in_url,
'number_of_digits_in_url' : number_of_digits_in_url,
'number_of_special_char_in_url' : number_of_special_char_in_url,
'number_of_hyphens_in_url' : number_of_hyphens_in_url,
'number_of_underline_in_url' : number_of_underline_in_url,
'number_of_slash_in_url' : number_of_slash_in_url,
'number_of_questionmark_in_url' : number_of_questionmark_in_url,
'number_of_equal_in_url' : number_of_equal_in_url,
'number_of_at_in_url' : number_of_at_in_url,
'number_of_dollar_in_url' : number_of_dollar_in_url,
'number_of_exclamation_in_url' : number_of_exclamation_in_url,
'number_of_hashtag_in_url' : number_of_hashtag_in_url,
'number_of_percent_in_url' : number_of_percent_in_url,
'domain_length' : domain_length,
'number_of_dots_in_domain' : number_of_dots_in_domain,
'number_of_hyphens_in_domain' : number_of_hyphens_in_domain,
'having_special_characters_in_domain' : having_special_characters_in_domain,
'number_of_special_characters_in_domain' : number_of_special_characters_in_domain,
'having_digits_in_domain' : having_digits_in_domain,
'number_of_digits_in_domain' : number_of_digits_in_domain,
'having_repeated_digits_in_domain' : having_repeated_digits_in_domain,
'number_of_subdomains' : number_of_subdomains,
'subdomain_depth' : subdomain_depth,
'having_hyphen_in_subdomain' : having_hyphen_in_subdomain,
'average_subdomain_length' : average_subdomain_length,
'average_number_of_hyphens_in_subdomain' : average_number_of_hyphens_in_subdomain,
'having_special_characters_in_subdomain' : having_special_characters_in_subdomain,
'number_of_special_characters_in_subdomain': number_of_special_characters_in_subdomain,
'having_digits_in_subdomain' : having_digits_in_subdomain,
'number_of_digits_in_subdomain' : number_of_digits_in_subdomain,
'having_repeated_digits_in_subdomain' : having_repeated_digits_in_subdomain,
'having_path' : having_path,
'path_length' : path_length,
'having_query' : having_query,
'having_fragment' : having_fragment,
'having_anchor' : having_anchor,
'entropy_of_url' : entropy_of_url,
'entropy_of_domain' : entropy_of_domain,
# New 6 hybrid features
'has_suspicious_tld' : has_suspicious_tld,
'is_known_safe_sld' : is_known_safe_sld,
'consonant_vowel_ratio' : consonant_vowel_ratio,
'longest_digit_sequence' : longest_digit_seq,
'digit_letter_ratio' : digit_letter_ratio,
'path_to_url_ratio' : path_to_url_ratio,
}
except Exception as e:
raise HTTPException(
status_code=400,
detail=f"Feature extraction failed: {str(e)}"
)
# ββ 7. ENDPOINTS βββββββββββββββββββββββββββββββββββββββββββββ
@app.get("/")
def health_check():
return {
"status" : "running",
"model" : "PhishGuard v5",
"features" : len(features),
"version" : "5.0.0",
"message" : "PhishGuard v5 API is live β 48 hybrid features!",
"author" : "Uzman Zahid β Dublin Business School 2026",
"docs" : "/docs"
}
@app.post("/predict", response_model=PredictionResponse)
def predict(request: URLRequest):
"""
Predicts whether a URL is phishing or legitimate.
Returns prediction (0=legitimate, 1=phishing),
label, confidence score, and message.
"""
url = request.url.strip()
if not url:
raise HTTPException(
status_code=400,
detail="URL cannot be empty"
)
feat_dict = extract_features(url)
X = pd.DataFrame([feat_dict])[features]
prediction = int(model.predict(X)[0])
probability = model.predict_proba(X)[0]
confidence = float(round(max(probability), 4))
label = "phishing" if prediction == 1 else "legitimate"
message = (
f"β οΈ WARNING: This URL appears to be PHISHING! "
f"({confidence*100:.1f}% confidence)"
if prediction == 1 else
f"β
This URL appears to be LEGITIMATE. "
f"({confidence*100:.1f}% confidence)"
)
return PredictionResponse(
url = url,
prediction = prediction,
label = label,
confidence = confidence,
message = message
)
@app.post("/predict/batch")
def predict_batch(urls: list[str]):
"""
Predicts multiple URLs at once.
Accepts a list of URL strings.
Returns array of predictions with labels and confidence.
"""
results = []
for url in urls:
try:
feat_dict = extract_features(url.strip())
X = pd.DataFrame([feat_dict])[features]
prediction = int(model.predict(X)[0])
probability = model.predict_proba(X)[0]
confidence = float(round(max(probability), 4))
results.append({
"url" : url,
"prediction" : prediction,
"label" : "phishing" if prediction == 1 else "legitimate",
"confidence" : confidence
})
except Exception as e:
results.append({
"url" : url,
"error": str(e)
})
return {
"results": results,
"total" : len(results),
"model" : "PhishGuard v5"
} |