DETECTION OF PHISHING WEBSITE USING SVM AND LIGHT GBM

Abstract

This paper presents a concise approach for detecting phishing websites using Support Vector Machines (SVM) and LightGBM. We extract a compact set of URL- and content-based features (domain age, URL length, SSL certificate presence, HTML/JavaScript anomalies, and page redirection patterns) and train both classifiers to distinguish phishing from legitimate pages. Experiments on benchmark datasets show that LightGBM achieves faster training and marginally higher detection performance while SVM offers robust, interpretable decision boundaries—making the two methods complementary for deployment. The proposed system is lightweight, suitable for real-time filtering, and reduces false positives compared to simple rulebased detectors.