Design And Implementation Of Intrusion Detection System Based-on Support Vector Machine And Agent

With the rapid growth of internet dimensions and complexity, people nowadays rely more and more on network. While at the same time, there are more network system attacks, technique of which tends to become complex. As an auto-detecting system for computer system or network intrusion behavior, Intrusion Detection System(IDS) is certainly a significant part of network security.This paper firstly introduces network security condition and concept, character and division of IDS. The development tendency of existing intrusion system is discussed, pointing out the disadvantage of traditional IDS. To develop a new IDS, the problem of detecting accuracy and efficiency need to be solved, and distributing intellectualized method should be considered. Therefore, we bring forward IDS based on SVM and mobile Agent technology. It makes use of SVM as a detecting tool for its good classification performance when dealing with problems of small sample set, high dimension non-linear data. Intellectuality and transferability mobile Agent is utilized to move and detect intrusion among network nodes.In chapter 3, we present basic theories of SVM and formula of soft margin, and indicate that SVM learning performance severely rely on selection of kernel function and kernel parameters. Using an example, we show visually how SVM kernel parameters affect detection performance. Through large amount of experiments, optimal kernel function and kernel parameters for intrusion detection are selected. Chapter 4 gives the details of Aglet, the Agent tool we used, and its system framework, object model and communicating model. Chapter 5 and chapter 6 present configuration of system model based on SVM and Agent, and introduce some functional model design and system environment.