The Research On Security Detection Of Web Client Script Based On Decision Tree

With the rapid development of Internet and web applications, the interaction between users and applications is getting more frequent, which leads to a large number of user data spreading on the Internet. It has become a research object of web security to ensure the security of these data. Originally, studies on web security mainly focus on the server defense. But on account of the rapid development of Web application, the interactive demands between web applications and users become more and more complex. In order to reduce the amount of information transmission between client and server, the client undertakes many functional requirements of logic control. Scripting technology represented by JavaScript is a powerful tool to help web client perform the logic control function. However, the strong control ability of script procedure makes it easier for hackers to attack the web applications by using scripts. Existing approaches of script security protection are implemented based on the input and output filtering and coding, constrained operations of script, the penetration test of static codes, and so on. There is still a lack of effective schemes that can defend and eliminate the client script attack.This paper presents a scripting attack detection method based on decision tree algorithm. By using the category advantage of CART decision tree, the paper analyzes the feature attributes that web applications may be attacked by script, and determines whether the value contains malicious script based on the value of the feature attributes. It involves the principles and harms of scripting attack, and studies measures of existing defense and decision tree. As for the selection of attributes, this paper fully considers the key points that web applications are easy to be under scripting attacks. It also comprehensively analyzes the object and attribute value of these key points. By using Java, a detection system based on this way has been made and passed the experimental test.