Design and Experimental Evaluation of an AI-Based Multi-Source Phishing Detection System Using Explainable AI

– Phishing is a major cybersecurity threat in which attackers use fake emails and websites to steal sensitive user information. Many existing detection systems rely on a single source of data and do not clearly explain their decisions, which limits their effectiveness and reduces user trust. This paper proposes an AI-based multi-source phishing detection system that analyzes both emails and websites. The system extracts features from text content, URL structures, and metadata, and processes them using machine learning and deep learning models. A Random Forest classifier and a Convolutional Neural Network (CNN) were used to identify phishing attacks. The system was tested using publicly available datasets, including the PhishTank dataset and the Enron Email Dataset. The CNN model achieved 95% accuracy, outperforming the Random Forest model, which achieved 93% accuracy. To improve transparency, an Explainable AI (XAI) module using SHAP and LIME was included to highlight important features influencing the predictions. The results show that the proposed approach improves detection accuracy, reduces false positives, and provides clear explanations, making it suitable for real-time phishing detection systems.