Enhancing Web Security: A Multi-Layer Approach for Malicious URL Detection

Phishing attacks often impersonate trusted websites, such as those of financial institutions, to steal sensitive data like login credentials and payment information. This project introduces a three-layered detection framework—Static Detection, Heuristic Detection, and Machine Learning Detection—to enhance the accuracy and reliability of phishing website identification. Features are extracted from legitimate and phishing websites, and a comprehensive dataset in CSV format is used for training and testing purposes. The static detection layer analyzes fixed attributes such as domain age, URL structures, and SSL certificate usage. The heuristic detection layer focuses on behavioral patterns, including frequent redirections and anomalous link characteristics. Finally, the machine learning detection layer employs a Back Propagation Neural Network (BPNN) that integrates character embedding with data mining features to capture intricate details and overarching patterns within URLs. By combining these three detection layers, the system achieves high accuracy and provides a robust defense against phishing threats.