Research On Web Theme Content Security Analysis And Monitor

The opening and the increasing scale have made the network a convenient means by which people exchange information freely. However, at the same time, conveniency and openness also bring many negative effects for the network, such as the spreading of various kinds of superstition, pornography, violence, reactionary and some other illegal information, or leaking of secret information, etc. And the traditional filtering techniques, such as keywords-based filtering, IP address filtering etc. cannot effectively solve these problems by now.With such demand in mind, this thesis, based on support vector machines, analyzes and studies the text filtering technology, in order to analyze network information content and filtrate the network information. Based on the preliminary study on the SVM (Support Vector Machines) , we proposed a text classification filtering algorithm which is based on the KKT feedback to improve conditions for learning mechanism of SVM. Firstly, we extract features from Web theme webpages that user collected from Web and then train the SVM according to the user's need. Secondly, we employed the SVM to analyze and filter the theme page text. Finally, we improve the performance of the SVM via feedback mechanism based on KKT conditions. Preliminary experimental results on a mid-size scale of web theme show that the proposed algorithm can do the filtering and classification on the malicious webpages and get the better results on specific information security.