A Client-Side AI Tool for Real-Time Phishing URL Detection

Abstract

Phishing attacks have increasingly exploited user trust by mimicking legitimate websites to steal personal information. This project introduces PhishCatcher, a machine learning-based client-side solution that detects spoofed web pages in real time. By extracting URL-level features and applying a Random Forest classifier, the tool effectively differentiates between legitimate and malicious web pages. Implemented as a Google Chrome extension, PhishCatcher achieved a 98.5% accuracy on a dataset containing both phishing and genuine URLs. The system demonstrated minimal latency with an average response time of 62.5 milliseconds, making it suitable for real-time applications. This approach enhances user security by reducing reliance on outdated blacklist-based detection, offering a robust defense mechanism against evolving phishing threats.