Performance Evaluation of Machine Learning Techniques for Detecting Cross-Site Scripting Attacks

The entire world shares the digital form of information through the Internet by accessing multiple websites. The link to these websites can be shared by authenticated or unauthenticated sources. Cross-site scripting attacks are the category of cyberattack in which malicious code can be inserted into the webpage code. These attacks can be so severe that they can steal the entire data of the machine or server and can harm the network infrastructure. Ransomware attackers also utilize cross-scripting attacks as the medium for breaking the security policies of the victim’s network. Machine learning techniques are widely employed to detect different types of cyberattacks in the industry. In this research paper, machine learning techniques, Naïve Bayes, Logistic Regression, AdaBoost, XGBoost, and decision tree are used to detect the cross-scripting attacks from the information about scripts. The results prove that the AdaBoost technique gains 97.29% accuracy for detecting cross-scripting attacks and the Naïve Bayes methodology shows minimum accuracy among the employed machine learning methodology. These achieved results form the future directions for building smart extensions for Internet browsers to detect such attacks in advance.