Explainable Ensemble Machine Learning Framework for Finding Phishing Websites in Real Time

Abstract: One of the most significant cyber risks in today's digital communication is phishing websites that prey on users through fake web pages that look like real websites to steal important information such as login passwords, banking details, and personal data. The current blacklist based phishing detection systems are less successful against the emerging and zero-day phishing assaults, as they cannot identify the unseen malicious URLs in real-time. In order to overcome these limitations, this study presents an Explainable Ensemble Machine Learning Framework for Real-Time Phishing Website Detection that utilizes advanced feature engineering, ensemble learning algorithms, and explainable artificial intelligence (XAI) techniques to enhance detection performance and interpretability.

The proposed methodology uses the entire range of URL-based, domain-based and webpage content-based attributes derived from the phishing and legal websites collected from publicly available cyber security datasets PhishTank, UCI Repository and Kaggle. Ensemble method based on voting combines and implements multiple machine learning algorithms such as Random Forest, Gradient Boosting and XGBoost classifiers. Moreover, there is explainable AI based on SHAP for transparent decision-making and feature importance analysis. The experimental results show that the proposed ensemble framework provides better performance than standard standalone machine learning models in terms of accuracy, precision, recall, F1-score, and ROC-AUC score. The platform also offers real-time phishing detection for browser extensions and cybersecurity apps. The suggested research contributes to the development of an intelligent, scalable, accurate and interpretable phishing detection system for modern cyber security infrastructures.

Keywords: Phishing Website Detection, Machine Learning, Ensemble Learning, Explainable AI, SHAP, Cybersecurity, XGBoost, Random Forest, Real-Time Detection