XSS Attack Detection Using Machine Learning

Abstract

Cross-site scripting (XSS) is one of the severe problems in Web Applications. With more connected devices which use different Web Applications, the risk of XSS attacks is increasing. In a cross-site scripting attack, the attacker injects malicious client-side scripts into the source code of a reliable program or website. If the website does implement sane input validation and filtration methods, the script will execute later on the client browser who visits such websites. Cross-site scripting attacks have been happening since the 1990s. Even today, XSS is considered a significant threat to many web applications as E-commerce websites with user reviews and Social Networking Services (SNSs) such as Facebook and Twitter. Such platforms rely on Javascript and Ajax for seamless user experience with high communication frequency among its users, making them very prone to XSS attacks. In this paper, we focus on detecting XSS attacks using six different machine learning classifiers that classify web pages into two categories: XSS or non-XSS based on several extracted features. With the highest classifier achieving accuracy of 0.993, precision and recall of 0.99, the evaluation results show that our approach achieves high performance in detecting XSS attacks.