E2Phish: Explainable Ensemble Machine Learning Model for Enhanced Phishing URL Detection

This paper presents a framework E2Phish which is integrated with explainable ensemble models for detection of phishing URLs. This includes Hard Voting, Soft Voting and Stacking to increase classification accuracy. The Mutual Information (MI) Score was used for feature selection to identify features that can enhance the performance of the detection models from the dataset. For explainability of these models, Explainable Artificial Intelligence (XAI) techniques such as SHAP and LIME and PDP were employed. These techniques focus on the most important features that have impact on the predictions of the model. The framework E2Phish is effective in detecting phishing URLs and at the same time focuses on the explainability of the models which is important in cybersecurity. Code of this project is available at https://github.com/tejas-kalal/E2Phish