Phishing is one of the most dangerous attacks in cybersecurity, which has increased since the introduction of IoT devices, involving attempts to trick users into handing over their passwords and sensitive data. Since most existing detection mechanisms are either nonintuitive or untrusted from the user's perspective, this project attempted to create a phishing detection system that relies on machine learning with explainable AI (XAI). Considering the results of previous studies that stress the importance of accurate and understandable phishing detection models, a five-phase framework was adopted: data collection, data cleaning, data modeling, XAI, and design of an interactive mechanism. The PhishTank dataset was preprocessed to improve model performance by optimizing the feature set and eliminating noise. Random Forest (RF) was selected, which was the best in terms of accuracy, precision, recall, and F1 score compared to Logistic Regression (LR) and Decision Trees (DT) models. LIME and SHAP were used to offer interpretability and present feature importance at the instance and global levels, respectively. Through an engaging mechanism, users can input URLs, obtain predictions regarding possible phishing attempts, and even read explanations, promoting comprehension and trust. This research shows that including XAI can improve not only the efficacy of the phishing detection systems, but also the level of trust that users have in such systems and be the basis for even more robust and more explainable cybersecurity mechanisms.
Fatima et al. (2025) studied this question.