This project implements a Machine Learning and Deep Learning hybrid approach to detect phishing websites. By analyzing URLs and their associated features, the system predicts whether a given website is legitimate or phishing, leveraging multiple ML algorithms and neural networks for improved accuracy.

• 📊 Collects phishing URLs from PhishTank and legitimate URLs from University of New Brunswick datasets
• 🔍 Extracts 17 handcrafted features (address bar, domain, HTML & JavaScript based)
• 🧠 Trains multiple ML & DL models including Decision Trees, Random Forest, XGBoost, SVMs, Autoencoders, and MLPs
• 📈 Evaluates models and compares performance
• 🏆 Best performing model: XGBoost (86.4% accuracy)
• 💾 Saves trained model as .pickle file for future predictions

1. Clone the Repo

   git clone https://github.com/jasoncobra3/Website_Phishing_Detection.git
   cd Website_Phishing_Detection
   
   

2. Create Virtual Environment

   python -m venv phishing_env
    # Windows:
    phishing_env\Scripts\activate
    # macOS/Linux:
    source phishing_env/bin/activatee
   
   

3. Install Dependencies

   pip install -r requirements.txt
   
   

4. Run Jupyter Notebooks

   jupyter notebook
   

• Open URL Feature Extraction.ipynb - extract features
• Open Phishing Website Detection_Models & Training.ipynb - train & evaluate models

• ✅ XGBoost achieved 86.4% accuracy, outperforming other models
• ✅ Random Forest and SVM performed moderately well
• ✅ Autoencoder and MLP showed promising results for DL integration

• 🌐 Develop a browser extension to detect phishing in real-time
• 🎨 Build a GUI/web app for user-friendly phishing detection
• 🔄 Improve hybrid ML-DL pipeline with larger datasets

• PhishTank Dataset : https://www.phishtank.com/developer_info.php
• UNB Dataset : https://www.unb.ca/cic/datasets/url-2016.html
• UCI Phishing Features : https://archive.ics.uci.edu/ml/datasets/Phishing+Websites

Feel free to fork, star, or submit a pull request to contribute improvements!