PHISHING DETECTION SYSTEM USING MULTI-SOURCE URL AND CONTENT-BASED FEATURES

Abstract

Phishing continues to be a major online security threat, deceiving users into revealing sensitive data. Attackers design fake websites that mimic legitimate ones to steal credentials and financial details. Conventional blacklist-based systems fail to detect new or evolving phishing sites. This study introduces a machine learning-based phishing detection model. It analyses both URL structure and webpage content for suspicious patterns. URL features include length, domain type, and special symbols. Content features focus on HTML tags, scripts, and embedded links. Decision Tree, MLP, and XGBoost algorithms are trained on labelled datasets. Among them, XGBoost achieves the best accuracy with minimal false detections. The system offers a fast, scalable, and effective defence against phishing attacks