Paper Title
AN EXPLAINABLE ENSEMBLE LEARNING FRAMEWORK FOR PHISHING WEBSITE DETECTION USING RANDOM FOREST AND XGBOOST WITH SHAP EXPLAINER

Abstract
Phishing contributes to the most common cyberattacks, because it manipulates user behavior rather than system weak points. Even with improvements in security mechanisms, many existing solutions still depend on constant methods such as blacklisting, which are not useful for handling newly created threats like Zero Day phishing websites. These websites normally exist only for a short duration, making them difficult to detect using traditional approaches. In this work, an ensemble based phishing detection model is proposed by combining Random Forest and XGBoost. The model uses around lexical and domain based features extracted from URLs to identify whether a website is legitimate or malicious. Instead of depending on a single model, the approach helps improves the consistency of prediction and reduces the chances of overfitting. Many machine learning models function like black boxes, making it difficult to understand how the concept arrive to the prediction. To overcome this issue SHAP(SHapley Additive exPlanations) is used to understand how each feature contributes to the prediction. This improves the transparency of the system and easier to trust, especially in realtime security applications. The model was evaluated on the UCI phishing dataset and achieved the Accuracy of 97. 1% and a recall of 98. 5%, which is performing better than individual models like Decision Tree and SVM. Besides the performance, the paper focus on the transparency gap in the cybersecurity field by adding the SHAP explainability and implementing a diagnostic dashboard, which is live and Streamlit-based. This way, security analysts can understand the results by viewing Reason Codes generated on the spot and making the system a useful tool for professional security audits. Keywords - Phishing Detection, Ensemble Learning, Random Forest, XGBoost, Explainable AI (XAI), SHAP.