Web Phishing Detection Using a Modified Spider Monkey Optimization Algorithm

Abstract

In today's digital age, technology is both a powerful tool and a potential threat, with its widespread adoption bringing significant advancements as well as new vulnerabilities. As we increasingly rely on digital solutions for everyday tasks, the importance of safeguarding against cyber threats has become paramount. This is particularly true in the context of web security, where cybercriminals are continuously developing more sophisticated methods to exploit online systems. Given the increasing sophistication and frequency of phishing attacks, traditional phishing detection methods remain necessary. However, more advanced methods are needed to detect new attacks. This study presents a web phishing detection approach using machine learning (ML) algorithms and an enhanced feature selection (FS) through the modified Spider Monkey Optimization (SMO) algorithm. The SMO algorithm was modified by adding an effective crossover function to improve the algorithm's performance. The study employs Mendeley phishing datasets containing features extracted from both phishing and legitimate web pages. In binary classification, the proposed detection model consistently demonstrated strong performance using XGBoost, LightGBM, and Random Forest.