Detecting Cross-Site Scripting Attack using Machine Learning Algorithms

Cross-Site scripting attacks can have a significant impact on a user's computer due to their ability to infect it with malware. This is because the code is embedded in a vulnerable program, allowing attackers to gain access through session cookies and carry out any action that is allowed by the user's account. This can range from credential theft to session hijacking, as well as a variety of other security vulnerabilities. Therefore the detecting and preventing the XSS attacks are critical to assure the safeguard of web-based applications. This research work concentrates on the automation of Cross-Site scripting attack detection by employing machine learning algorithms like Naive Bayes, K-Nearest Neighbors, Decision Trees, Random Forest, and Support Vector Machines. XSS attacks frequently exhibit patterns and characteristics that can be learnt from historical data. These patterns are well recognized by supervised machine learning algorithms, which can also generalize from training data to detect similar harmful patterns in real-time web traffic. XSSDataSet1, a large dataset of both genuine and malicious On-the-Web (OTW) requests and attack methods is utilized for training and testing the Machine learning models. While extracting the data, the request's payload, request parameters, and contextual information are taken into account. The evaluation of each algorithm includes an analysis of various performance metrics such as accuracy, precision, recall, and the F1 Score. An accuracy level of 99.91% has been attained with decision tree algorithm. The results of the extensive analysis showed that decision tree classifier, random forest, support vector machine, and K-Nearest Neighbors are more efficient than Naive Bayes in detecting the XSS attack.