Genetic Algorithm-Based Feature Selection for Optimizing Website Intrusion Detection Model

Abstract

The widespread use of the internet and digital platforms has made websites essential for communication, business, and accessing information. However, this reliance has also heightened cybersecurity concerns, as cyber-attacks on web applications pose risks such as data theft, malware installation, and redirection to malicious sites.  Cybersecurity is increasingly leveraging machine learning and predictive analysis to proactively identify and mitigate potential web application attacks. Nonetheless, the key challenge in website attack prediction is identifying the most relevant features that balance accurate predictions with manageable computational overhead. While Genetic Algorithm-based feature selection techniques hold promise, their effectiveness should be evaluated on datasets containing common web attacks such as SQL injection, cross-site scripting (XSS), and cross-site request forgery (CSRF). Additionally, metrics beyond accuracy should be considered to assess their generalizability and overall performance. This research investigates the application of Genetic Algorithms (GA) for feature selection in website attack prediction using Decision Tree and Logistic Regression models. The algorithms were evaluated using accuracy and F1-score metrics. The findings revealed that feature selection using GA significantly reduces the number of features to up to 86.8% while maintaining or improving detection accuracy and reduced overfitting. This 86.8% reduction in features simplifies the model while saving resources. The findings highlight the potential of GA as a valuable method for feature selection in website attack prediction.