import os import re import math import pickle import joblib from urllib.parse import urlparse from collections import defaultdict, Counter from difflib import SequenceMatcher # Socket fallback for local DNS checking import socket import datetime import requests from bs4 import BeautifulSoup from flask import Flask, request, jsonify, render_template from flask_cors import CORS # --- SAFE WHOIS IMPORT --- try: import whois except ImportError: whois = None # ========================================== # 🛠️ DYNAMIC PATHS FOR CLOUD DEPLOYMENT # ========================================== BASE_DIR = os.path.dirname(os.path.abspath(__file__)) TEMPLATE_DIR = os.path.join(BASE_DIR, 'Web') if os.path.exists(os.path.join(BASE_DIR, 'Web')) else os.path.join(BASE_DIR, 'templates') MODEL_PATH = os.path.join(BASE_DIR, 'Web', 'Models') if os.path.exists(os.path.join(BASE_DIR, 'Web', 'Models')) else os.path.join(BASE_DIR, 'Models') app = Flask(__name__, template_folder=TEMPLATE_DIR) CORS(app) # ========================================== # 🧠 MODEL CLASSES & DEFINITIONS # ========================================== class MarkovPasswordModel: def __init__(self): self.n = 3 self.transitions = defaultdict(Counter) def score_password(password: str): if not password: return 0, 'Too Short' # 🛠️ HIGH-FIDELITY HYBRID FALLBACK ALGORITHM # If file loader fails due to structural environment guards, calculate real complexity dynamically if not markov_model or not hasattr(markov_model, 'transitions') or len(markov_model.transitions) == 0: length = len(password) pool = 0 if re.search(r'[a-z]', password): pool += 26 if re.search(r'[A-Z]', password): pool += 26 if re.search(r'[0-9]', password): pool += 10 if re.search(r'[^a-zA-Z0-9]', password): pool += 32 if pool == 0: pool = 1 entropy = length * math.log2(pool) score = (entropy / 80) * 100 score = max(5, min(score, 100)) if length < 6: score = min(score, 25) elif length < 8: score = min(score, 45) labels = [(20, 'Very Weak'), (40, 'Weak'), (60, 'Moderate'), (80, 'Strong')] label = next((l for s, l in labels if score < s), 'Very Strong') return round(score, 2), label # Standard Markov Calculation if loaded successfully order, text = 2, '<^^' + password + '<' log_prob, count, vocab = 0.0, 0, 150 for i in range(order, len(text)): ctx = text[i - order:i] nx = text[i] tot = sum(markov_model.transitions[ctx].values()) freq = markov_model.transitions[ctx].get(nx, 0) prob = (freq + 1) / (tot + vocab) if tot > 0 else 1 / vocab log_prob += math.log2(prob) count += 1 if count == 0: return 0, 'Too Short' avg_ll = log_prob / count score = (avg_ll - (-3.5)) / ((-8.0) - (-3.5)) * 100 score = max(0, min(score, 100)) labels = [(20, 'Very Weak'), (40, 'Weak'), (60, 'Moderate'), (80, 'Strong')] label = next((l for s, l in labels if score < s), 'Very Strong') return round(score, 2), label def load_model_file(filename): path = os.path.join(MODEL_PATH, filename) if not os.path.exists(path): print(f"[WARNING] Cloud model file not found at: {path}") return None try: return joblib.load(path) except Exception as e: print(f"[ERROR/JOBLIB] Direct load failed: {e}") try: with open(path, "rb") as f: return pickle.load(f) except Exception as e2: print(f"[ERROR/PICKLE] Fallback read failed: {e2}") return None # --- MODELS LOADING --- password_model = load_model_file("markov_password_model.joblib") markov_model = password_model phishing_svm_model = load_model_file("phishing_svm_model.pkl") url_data = load_model_file("phishing_model_url.pkl") url_legitimacy_model = url_data["model"] if isinstance(url_data, dict) else url_data print("[SYSTEM] Cloud-ready configuration loaded.") # ========================================== # 🔍 TYPO DETECTION # ========================================== TOP_TARGET_DOMAINS = [ "google.com", "facebook.com", "amazon.com", "paypal.com", "microsoft.com", "apple.com", "netflix.com", "instagram.com", "twitter.com", "linkedin.com", "github.com", "yahoo.com" ] def get_domain(url): parsed = urlparse(url if url.startswith("http") else "http://" + url) return parsed.netloc.lower().replace("www.", "") def similarity(a, b): return SequenceMatcher(None, a, b).ratio() def detect_typo(domain): best_match = None best_score = 0 for legit in TOP_TARGET_DOMAINS: sim = similarity(domain, legit) if sim > best_score: best_score = sim best_match = legit return best_match, best_score def extract_url_features(url: str): parsed = urlparse(url if url.startswith("http") else "http://" + url) domain = parsed.netloc.lower().replace("www.", "") feats = [len(url), len(re.findall(r"\d", url)), url.count('.'), domain.count('-'), 1 if parsed.scheme == "https" else 0] while len(feats) < 18: feats.append(0) return feats # ========================================== # 🛠️ ROBUST FEATURE EXTRACTION # ========================================== def extract_real_features(url): try: parsed = urlparse(url if url.startswith("http") else "http://" + url) domain = parsed.netloc url_length = len(url) num_subdomains = domain.count(".") has_ip = 1 if re.match(r"\d+\.\d+\.\d+\.\d+", domain) else 0 has_https = 1 if parsed.scheme == "https" else 0 num_special = len(re.findall(r"[!@#$%^&*(),?\":{}|<>]", url)) url_depth = parsed.path.count("/") try: socket.gethostbyname(domain) dns_exists = 1 except: return None age_days = 0 if whois: try: domain_info = whois.whois(domain) creation_date = domain_info.creation_date if isinstance(creation_date, list): creation_date = creation_date[0] if creation_date: age_days = (datetime.datetime.now() - creation_date).days except: age_days = 0 try: response = requests.get(url, timeout=4) html = response.text soup = BeautifulSoup(html, "html.parser") except: return None forms = soup.find_all("form") iframes = soup.find_all("iframe") scripts = soup.find_all("script", src=True) has_login_form = 0 for form in forms: if form.find("input", {"type": "password"}): has_login_form = 1 break iframe_count = len(iframes) external_scripts = 0 for script in scripts: src = script.get("src") if src and domain not in src: external_scripts += 1 ssl_valid = 1 if parsed.scheme == "https" else 0 features = [ url_length, num_subdomains, has_ip, 0, num_special, has_https, url_depth, 0, 0, len(domain), age_days, ssl_valid, dns_exists, 1, 1, external_scripts, has_login_form, 0, iframe_count, 0, 0 ] return features except: return None def check_blacklist(url): try: test_url = f"https://transparencyreport.google.com/transparencyreport/api/v3/safebrowsing/status?site={url}" r = requests.get(test_url, timeout=4) if "unsafe" in r.text.lower(): return 1 return 0 except: return 0 # ========================================== # 🛣️ ROUTES # ========================================== @app.route('/') def index(): return render_template('Index.html') @app.route('/url-checker') def url_page(): return render_template('url-checker.html') @app.route('/phishing') def phishing_page(): return render_template('phishing.html') @app.route('/password-checker') def password_checker_page(): return render_template('password-checker.html') @app.route('/hash') def hash_page(): return render_template('hash.html') @app.route('/analysis') def analysis_page(): return render_template('Analysis.html') # ========================================== # 🔌 API ENDPOINTS # ========================================== @app.route('/check_url', methods=['POST']) def check_url(): if not url_legitimacy_model: return jsonify({"error": "URL legitimacy model missing"}), 500 url = request.get_json().get("url", "") domain = get_domain(url) prob = url_legitimacy_model.predict_proba([extract_url_features(url)])[0][1] ml_score = (1 - prob) * 100 matched_domain, sim_score = detect_typo(domain) typo_penalty = 0 is_typo = False if sim_score > 0.75 and domain != matched_domain: is_typo = True typo_penalty = sim_score * 50 final_score = min(ml_score + typo_penalty, 100) status = "Not Safe" if is_typo or final_score < 75 else "Safe" return jsonify({ "score": round(final_score, 2), "status": status, "typo_detected": is_typo, "matched_domain": matched_domain if is_typo else None, "similarity": round(sim_score, 2) }) @app.route('/check_phishing', methods=['POST']) def check_phishing(): if not phishing_svm_model: return jsonify({"error": "Phishing SVM model missing"}), 500 url = request.get_json().get("url", "") model = phishing_svm_model['pipeline'] features = extract_real_features(url) if features is None: return jsonify({"score": 0, "status": "Website Not Found"}) blacklist_flag = check_blacklist(url) probs = model.predict_proba([features])[0] phishing_prob = probs[1] * 100 final_score = 100 if blacklist_flag else phishing_prob domain = get_domain(url) matched_domain, sim_score = detect_typo(domain) if sim_score > 0.75 and domain != matched_domain: final_score += 30 final_score = min(final_score, 100) return jsonify({ "score": round(final_score, 2), "status": "Not Safe" if final_score > 50 else "Safe" }) @app.route('/check_password', methods=['POST']) def route_score(): data = request.get_json() s, lbl = score_password(data.get('password', '')) return jsonify({'score': s, 'label': lbl}) if __name__ == '__main__': port = int(os.environ.get("PORT", 5000)) app.run(host='0.0.0.0', port=port)