Design and Comparative Evaluation of a Hybrid Rule-Based and Machine Learning Phishing URL Detection System

Abstract

This dissertation presents the design and evaluation of a hybrid phishing URL detection system that combines rule-based techniques with machine learning. The system was evaluated using a dataset of 11,054 website samples with 31 features. Three detection methods were compared: a rule-based model, a Random Forest model, and a hybrid model. Performance was measured using accuracy, precision, recall, F1-score, and confusion matrix analysis. The Random Forest model achieved the highest accuracy (96.92%), while the hybrid model achieved 96.31% with improved phishing detection sensitivity. The results demonstrate that machine learning significantly outperforms traditional rule-based methods and that hybrid approaches provide an effective solution for phishing website detection and enhanced cybersecurity.