Web Phishing Detection and Awareness Utilizing Hybrid Machine Learning Algorithms

The growing risks of web phishing attacks necessitate advanced detection mechanisms to safeguard users from phishing sites. This paper proposes a Layered Web Phishing Detection (LWPD) system that combines hybrid machine learning algorithms with real-time user awareness mechanisms. The system involves two stages: (1) a hybrid ML model integrating unsupervised clustering (K-means) and supervised classification (Random Forest, XGBoost) to detect phishing URLs with 99.7% accuracy after feature engineering; and (2) a browser extension that alerts users in real time regarding suspicious links. Experiments with a balanced dataset (11,430 URLs, 87 features) demonstrate that feature reduction to 34 significant attributes enhances model performance without compromising high precision. The method put forward in this paper outperforms conventional ML methods by reducing feature dimensionality by 60% at the expense of just a 2% loss in accuracy. This study bridges the gap between theoretical detection and actual user awareness in the real world by presenting a viable solution for cybersecurity application.