URL Feature Analysis for Effective Phishing Detection using Machine Learning

URL Feature Analysis for Effective Phishing Detection using Machine Learning

Dr. S. Subashini¹, A. MohammedNavas², T. S. P. Ksheerabdinath³, M. H. Ksheerabdhinath⁴,

T. S. Muneeshwara Venkatesh5

¹²³45Department of Computer Science and Engineering, KLN College of Engineering, Tamil Nadu, India.

---------------------------------------------------------------------***---------------------------------------------------------------------

Abstract - In the digital age, phishing websites pose a significant threat to online security by mimicking trusted platforms to steal sensitive user information. The Phishing Website Detection system leverages Machine Learning (ML) to automate the classification of URLs as either phishing or benign. By analyzing a dataset of 10,000 URLs, consisting of 5,000 phishing and 5,000 legitimate links, the system extracts 17 key features related to address bar, domain, and HTML/JavaScript indicators. Among the algorithms tested, including Decision Tree, Random Forest, and SVM, the XGBoost classifier achieved the highest accuracy of 86.4%. Developed with Flask as the backend, the web-based application ensures real-time detection, providing users with phishing alerts while maintaining memory efficiency. The tool contributes to enhancing web security, offering reliable protection in various digital environments.

Keywords: Machine Learning, Phishing Detection, URL Feature Extraction, XGBoost Classifier.