Testing Technology Malicious Web Based Active SVM Algorithm

In the network age, the applications which are based on the JavaScript and browser plug-in components spring up. With the convenient which the applications bring about, more and more problems emergence, such as the leakage of personal information information theft、 data tampering、 data updating、 computer virus, the exploits by the attackers are getting more and more daring in real time.As a result, the exploits of the web page, also mean the exploits of the browser are the main attacks happened. These attacks can exploit conveniently using JavaScript on a web site if we access the page and our browsers have the vulnerability.In addition, the JavaScript obfuscation technique is used to hide JavaScript containing the attack code. Obfuscation is a method that changes shape of data in order to avoid pattern-matching detection. Various methods are used to prevent attack made through Internet websites. Attackers circumvent the defense mechanism using obfuscation, with the effectiveness and promptness, spread like a net, threat the safety of information, which make the detection of the malicious codes much more difficult. And the obfuscation of JavaScript has become the most critical protection. How to detect these exploits and protect the users from attacks are serious problems in society.This paper introduced the Obfuscation of JavaScript and proposed the TF-IDF algorithm to feature extraction, combining the weight analysis of text categorization with the JavaScript character, which making the JavaScript feature extraction more scientific. And the experiments show that the extraction method performance has greatly improved. This article also supervised deficiencies of traditional SVM and proposed active learning machine learning strategy to simplify manual operations, improve efficiency and achieve highly intelligent systems. Experiments show that an Active SVM malicious page detection system has the smaller number of the samples marked but achieving better performance with fewer labor of human.