Intelligent Browser-Extension For Real-Time Phishing Detection Using Hybrid Machine Learning Models

Abstract

Phishing attacks remain a significant cybersecurity threat due to their evolving nature and reliance on social engineering techniques. While machine learning-based phishing detection models have demonstrated high accuracy in offline evaluations, their real-time deployment in client-side environments remains challenging. This paper presents a hybrid phishing detection framework that integrates offline machine learning analysis with real-time browser-based deployment. Multiple machine learning classifiers are evaluated using URL based features, and the Random Forest model is identified as the most effective classifier. Feature importance analysis is employed to extract the most influential phishing indicators, which are subsequently translated into a lightweight rule-weighted detection mechanism. This mechanism is implemented as a browser extension to enable real-time phishing detection without relying on external servers. Experimental results demonstrate that the proposed approach achieves high detection accuracy while maintaining low computational overhead. The system provides explainable detection decisions, preserves user privacy, and effectively bridges the gap between machine learning research and practical phishing defense systems suitable for real-world deployment.