Enhancing Phishing URL Detection Through Optimized XGBoost Classifier

Modern phishing attacks continue to thrive for the exploitation of users and with it, financial and data losses. Traditional approaches to detection are not able to keep pace with the growing sophistication of criminals. This research proposes a machine learning-based phishing detection framework that uses XGBoost to classify URLs as either phishing or legitimate. A publicly available dataset with URLs that have a multitude of features yields excellent performance. The analysis of feature importance reflects key URL features that led to the classification to phishing or legitimate URLs. XGBoost is a reliable and scalable technique to detect phishing in real-time. Future work will investigate a hybrid approach, integrating deep learning techniques, to allow for even greater adaptability against emerging phishing techniques.