Media
4 items
Description
Browser-based phishing detection using Hybrid Machine Learning and Adaptive Learning Phishora is an AI-powered browser extension that detects phishing websites using a hybrid machine learning approach. Simply click "Analyze Website" to instantly check if the site you're visiting is safe, suspicious, or a phishing threat. How it works: Phishora extracts 20 structural features from the URL (such as URL length, HTTPS status, subdomain count, obfuscation patterns, and redirect behaviour) and analyses them using three machine learning models working together — a Random Forest classifier, a Deep Neural Network, and an Adaptive Random Forest that continuously learns from your feedback. Key features: — One-click phishing analysis for any website — Three-tier classification: Legitimate, Suspicious, or Phishing — Individual confidence scores from each model (RF, DNN, and Combined) — Risk level indicators: Low, Medium, High, or Critical — Adaptive learning that improves detection accuracy over time using your feedback — Concept drift detection using ADWIN to handle evolving phishing tactics — Fast analysis with response times under 200 milliseconds — Complete privacy: all processing happens locally on your machine, no data is sent to external servers How to use: 1. Visit any website in Chrome 2. Click the Phishora extension icon in the toolbar 3. Click "Analyze Website" 4. View the verdict, confidence scores, and risk level 5. Optionally provide feedback (Yes/No) to help the model learn Important note: Phishora requires a local backend server running on your machine for ML inference. Please refer to the project documentation for setup instructions. Built as a research project at the Informatics Institute of Technology (IIT) in collaboration with the University of Westminster. Developed by Pooja Malagala | Supervised by Mr Imesh Pathirana