Optimized Feature Selection and Enhanced Phishing Detection Using Multi-Objective Techniques and XGBoost

Abstract

Web spoofing attacks pose a serious threat to the confidentiality and integrity of online interactions, often tricking users into disclosing sensitive information by mimicking legitimate websites. While existing server-side defenses—such as SSL/TLS protocols and domain validation—offer some degree of protection, they are inherently reactive and frequently fall short in real-time threat mitigation. These approaches often fail to detect spoofed sites promptly and accurately, leading to both false positives and undetected threats. Given the evolving sophistication of spoofing techniques, a proactive client-side detection system is essential for timely and effective protection. In response to this gap, the proposed pishcatcher system introduces a machine learning-driven client-side defense mechanism designed to detect and block web spoofing attempts in real time. By extracting and analyzing features from web pages—such as HTML structure, CSS layout, and JavaScript behaviors—the system utilizes advanced classification algorithms to differentiate between genuine and malicious websites. Furthermore, its adaptive learning capability allows continuous improvement in detecting emerging spoofing patterns, ensuring resilience against evolving attack vectors. This approach not only enhances the accuracy and responsiveness of spoofing detection but also provides users with a reliable tool to safeguard online interactions before damage occurs