Phishing URL Detector

AI-powered system that detects phishing URLs using machine learning, built with Python.

Overview

This project uses a Random Forest classifier to identify malicious/phishing URLs, trained on a dataset of 58,000+ URLs.

Results

• Accuracy: 95.8%
• Precision & Recall: 0.96 (balanced across both phishing and legitimate URL classes)
• Only 490 misclassifications out of 11,729 test samples

Tech Stack

• Python
• Scikit-learn (Random Forest)
• Pandas for data processing
• Matplotlib for visualization

Files

• phishing_detector.py — Main model training and evaluation script
• explore_data.py — Data exploration and preprocessing
• confusion_matrix.png — Model performance visualization

Limitations

• Currently misses ~3.7% of phishing URLs (false negatives)
• Requires pre-extracted features; not yet tested on raw live URLs

Future Work

• Live URL feature extraction for real-time predictions
• Flask web interface for interactive URL checking
• Hyperparameter tuning to improve recall on phishing class

Author

Aroon Kumar Maheshwari

Built With

• machine-learning
• python
• random-forest
• scikit-learn