DETEKSI SERANGAN REMOTE CODE EXECUTION DAN CROSS SITE SCRIPTING MENGGUNAKAN MACHINE LEARNING

Abstract

With the increasing connectivity and rapid growth of information technology, Indonesia is confronted with formidable challenges in the domain of cybersecurity. The incidence of cyber attacks has reached an alarming threshold, approaching nearly one billion occurrences throughout the year 2022. This not only serves as a disconcerting indicator of the intensification of cyber threats but also underscores the imperative for efficacious solutions to detect and mitigate these attacks. Such measures are deemed necessary for the comprehensive safeguarding of data security and privacy. This research endeavors to develop a machine learning-based system for the detection of cyberattacks. The swift evolution of technology has rendered cyberattacks increasingly intricate to identify, with methods and vectors becoming more sophisticated and diverse. Consequently, this study employs machine learning methods for detection, with a specific focus on two types of cyberattacks: Remote Code Execution and Cross-Site Scripting. To attain a precise detection model, three algorithms were scrutinized in this research: a) Support Vector Machine; b) Gradient Boosting; and c) Logistic Regression. According to the conducted research, the Support Vector Machine algorithm achieved the highest accuracy rates, specifically 0.9876 for Remote Code Execution and 0.9961 for Cross-Site Scripting. Meanwhile, Logistic Regression yielded accuracy rates of 0.9537 (Remote Code Execution) and 0.9939 (Cross-Site Scripting), while Gradient Boosting demonstrated accuracy rates of 0.9475 (Remote Code Execution) and 0.9939 (Cross-Site Scripting). After conducting penetration tests using the Arachni and ZAP applications, it can be concluded that the Remote Code Execution detection model successfully detected 100% of the attacks. Conversely, the Cross Site Scripting detection model managed to identify up to 70% of the tested attacks.