Enhancing Bank Customer Protection Against Phishing Attacks Through Xgboost-Based Feature Analysis

Abstract

Internet usage in Indonesia has significantly increased, with approximately 175.4 million people or 64% of the population actively using the internet. While the internet provides numerous benefits, such as easy access to information and faster communication, this rise in usage also opens opportunities for cybercriminals to exploit user vulnerabilities. One of the most common forms of cybercrime is phishing, which attempts to steal users' personal information by impersonating a trusted entity. Current methods for detecting phishing are ineffective against zero-day phishing attacks. Therefore, this study employs the XGBoost algorithm to detect phishing websites. The results show that the XGBoost model, using feature selection techniques, can enhance phishing detection accuracy to 95.5%, with a precision of 95.5%, recall of 95.1%, and F1-score of 95.3%. With these capabilities, XGBoost can be used to protect internet users from evolving phishing threats and assist banks in anticipating customer losses.