Evaluating Learning Models for Phishing Detection: A Comparative Literature Review

Abstract—Phishing email identification is one of the major issues in cybersecurity as attackers have advanced techniques to overcome conventional methods. In this paper, we carried out a comparative literature review on advanced techniques such as machine learning (ML), deep learning (DL), hybrid ensemble models, and explainable artificial intelligence (XAI). Key studies are analyzed with respect to accuracy, adaptability, transparency, and scalability. Their results show that hybrid ensembles of models with XAI provide a good trade-off between robustness and interpretability compared to single-model solutions. Nevertheless, there are some open issues including dataset diversity, computational expense, and industrial applicability to enterprise. This review not only combines the contemporary trends of phishing studies but also indicates gaps and future work directions for building effective, scalable, interpretable, adversarial-resistant phishing detection systems. Keywords— phishing detection, machine learning (ML), deep learning (DL), explainable artificial intelligence (XAI), cybersecurity, ensemble approach, email security, term frequency– inverse document frequency (TF-IDF), local interpretable modelagnostic explanations (LIME), SHapley additive exPlanations (SHAP)