Based On Automatic Feature To Crawl And Behavior Associated With Network Intrusion Detection System

Network Intrusion has been a frequently discussed issuewithout a perfect solution. A variety of methods are proposed by numerous professionals for intrusion prevention and management, and these methods focus on different aspect in the technique level. The question is how to resist against network intrusion and where to start.In this paper, we base our research on text clustering and classification techniques and try to link all the elements in the network together by mining and association, in hope to discover logic relevance like "beer and diapers" . We examine and eliminate the network intrusion from a new perspective. We focus on the proposed model and the possible results via this model. Category clustering analysis is an important analytical tool in machine learning, pattern recognition and computer vision. The traditional category clustering is unsupervised, and it is widely used because of its low level of human interaction and little cost. However, the problem with unsupervised learning is that it always depends on some assumptions, e.g., the assumption that type distribution is uniform and Eigen-values are equal, which may not be able to resolve new problems we encountered. So this paper takes lot of consideration into the behavior characteristics of the original network intrusion.After all, it is a information world with globalization, and people are complaining the network intrusion while we are accumulating huge amount of information of intrusion - packets, coding, system logs and network flows. When such information is used for category clustering analysis, expected results could be obtained under human's supervision or interaction.Our system proposed a new perspective for network intrusion detection and prevention. A new criterion is suggested in a macro way to judge between the network behavior and the various network devices.