Real-Time Explainable AI for Phishing Detection

This paper presents an AI-powered phishing detection system designed to identify malicious URLs using machine learning techniques and feature engineering. The system leverages a custom-built feature extractor that processes URL characteristics such as length, special character presence, IP usage, and redirection count to generate meaningful numerical attributes. Utilizing a labelled dataset of phishing and legitimate URLs, multiple machine learning models, including Random Forest and Support Vector Machines, were trained and optimized through comprehensive hyperparameter tuning. The trained models are deployed within an interactive web application built using Python’s Flask framework, facilitating real-time phishing detection. Experimental evaluation demonstrates high classification accuracy exceeding 97%, with precision and recall metrics validating its robustness. The system’s architecture, supported by thorough dataset preprocessing and scalable deployment configurations, addresses modern phishing challenges with adaptive and automated defence mechanisms. This work contributes a practical and effective AI-driven tool to enhance cybersecurity measures against phishing attacks