Secure Browse: AI-Powered Phishing Defense for Browsers

Abstract

Purpose: With the rising threat of phishing attacks exploiting user naivety, this report introduces a novel approach to bolster web security. Traditional rule-based systems and existing solutions fall short in addressing sophisticated phishing attempts. The proposed solution entails a Chromium-based browser extension that leverages machine learning classification techniques.

Methodology:  A Python web server, utilizing decision trees, k- nearest neighbors, and random forests, assesses the legitimacy of a given URL. The extension communicates with the server, providing real-time notifications to users when visiting potential phishing sites.

Findings: Experimental results demonstrate the effectiveness of the ensemble model with an accuracy of 90.68%, marking a significant improvement over rule-based alternatives.

Unique contribution to theory, policy and practice: Future work includes refining models, incorporating user feedback, and expanding the application to diverse platforms and contexts.