jahan-zeb01/phishing-url-detector

🔍 Phishing URL Detection Model

Model Description

A Random Forest classifier that detects phishing URLs using 20 URL-based features extracted in real-time.

• Accuracy: 99.65%
• F1-Score: 0.9969
• Features: 20 URL-based features
• Training Data: 235,370 URLs

Features Used

#	Feature	Description
1	URLLength	Total characters in URL
2	DomainLength	Length of domain name
3	IsDomainIP	1 if URL contains IP address
4	CharContinuationRate	Character repetition pattern
5	TLDLegitimateProb	Probability TLD is legitimate
6	URLCharProb	Character probability distribution
7	TLDLength	Length of top-level domain
8	NoOfSubDomain	Number of subdomains
9	HasObfuscation	1 if obfuscated characters present
10	NoOfObfuscatedChar	Count of obfuscated characters
11	ObfuscationRatio	Ratio of obfuscated chars
12	NoOfLettersInURL	Count of letters
13	LetterRatioInURL	Ratio of letters
14	NoOfDegitsInURL	Count of digits
15	DegitRatioInURL	Ratio of digits
16	IsHTTPS	1 if HTTPS used
17	IsSuspiciousTLD	1 if TLD is suspicious
18	ContainsBrand	1 if URL contains a brand name
19	IsSuspiciousDomain	1 if domain = brand + extra text
20	SuspiciousKeywordCount	Count of suspicious keywords

Model Performance

Metric	Score
Accuracy	99.65%
Precision	99.57%
Recall	99.82%
F1-Score	99.69%

How to Use

import joblib
import pandas as pd

# Load the model
model = joblib.load('phishing_model_final.pkl')
scaler = joblib.load('scaler_final.pkl')
feature_names = joblib.load('feature_names_final.pkl')

# Extract features from URL
def predict_url(url):
    # ... (include your feature extraction code)
    pass

# Make prediction
result = predict_url("www.google.com")
print(result)  # {'url': 'www.google.com', 'is_phishing': False, 'confidence': 99.0%}