Explainable Deep Learning Model for Phishing Email Detection Using XAI Techniques

ABSTRACT Phishing email attacks remain one of the most prevalent cybersecurity threats, causing significant financial and information losses worldwide. Although deep learning techniques have demonstrated high detection accuracy, their black-box nature limits trust and interpretability in real-world applications. This study proposes an explainable deep learning framework for phishing email detection based on a Bidirectional Long Short-Term Memory (BiLSTM) model integrated with SHAP (SHapley Additive exPlanations) for interpretability. A curated dataset of 3,332 email messages was constructed and preprocessed for experimental evaluation. The proposed model achieved an accuracy of 0.96, an F1-score of 0.95, and a ROC-AUC of 0.97, outperforming traditional machine learning (SVM) and CNN-based models. SHAP-based global and local explanations provide insights into feature contributions, enhancing transparency and trust in the model’s decisions. The results demonstrate that integrating explainable artificial intelligence techniques with deep learning significantly improves the reliability and applicability of phishing detection systems. Keywords: Phishing Detection, Deep Learning, LSTM, Explainable AI, SHAP, Cybersecurity, Email Security, Email Analysis, Deep Neural Networks