ML-Based Phishing Website Detection System

This project includes code based on publicly available content and might not have been entirely written by me.

ML-Based Phishing Website Detection System (Content-Based)

This project is part of a research-based effort focused on detecting phishing websites using content-based features like HTML tags. The repository includes code for feature extraction, data collection, preparation, and building machine learning models to classify websites as phishing or legitimate.

File Inputs

• CSV files containing phishing and legitimate URLs:
  • verified_online.csv - Phishing URLs from Phishtank.org
  • tranco_list.csv - Legitimate URLs from Tranco-list.eu

Process Overview

1. Load URLs from CSV files.
2. Fetch content for each URL using Python’s requests library.
3. Parse the content with BeautifulSoup to extract numerical features.
4. Create a structured dataframe, add labels (1 for phishing, 0 for legitimate), and save the data as CSV files.
   • See: structured_data_legitimate.csv and structured_data_phishing.csv
5. Split the data for training and testing or use K-fold cross-validation (K=5) as shown in the machine_learning.py script.

Implemented Machine Learning Models

• Support Vector Machine (SVM)
• Naive Bayes
• Decision Tree
• Random Forest
• AdaBoost

Evaluate models using accuracy, precision, recall, and visualize performance.

Dataset

You can create your own dataset using the data_collector.py script with a custom URL list. I used “phishtank.org” and “tranco-list.eu” as data sources.

• Total websites: 24,228
  • 12,001 legitimate websites
  • 12,227 phishing websites
• Dataset was created in January 2024.
• I may update the dataset every year.

Website

Visit the live demo at, feel free to reach me out, if site is frozen due not active users: ML Based Phishing Detection Visit my Github, if you need the source code.