A Phishing Website Detection System Based on Hybrid Feature Engineering with SHAP Explainable Artificial Intelligence Technique

ورقة مؤتمر

2025

تم النشر فى: https://link.springer.com/chapter/10.1007/978-981-96-1483-7_1#citeas

Lecture Notes in Computer Science, vol 15463, Springer

Today, phishing website attacks have emerged as a prevalent threat affecting Internet users, governments, and businesses. The key challenge is that phishers continuously deploy new techniques to create zero-day phishing attacks. Recently, researchers have suggested anti-phishing techniques based on classifiers such as machine learning and deep learning methods. This research aims to develop a framework for predicting zero-day phishing websites through introducing new hybrid feature engineering methods that help adapt to evolving threats by extracting meaningful attributes or characteristics of phishing websites, including content-based and URL-based features. By combining diverse feature types, a detection system can capture a comprehensive set of characteristics associated with phishing behavior, improving the accuracy of the detection process. In addition, we also apply SHapley Additive exPlanations (SHAP) technique to interpret the behavior of the hybrid feature engineering by quantifying the impact of individual features on model predictions, providing a clear understanding of which features contribute most significantly to phishing detection. Moreover, the proposed approach aims to capture the most informative and discriminative features, reducing the likelihood of overlapping features.