An Ids Based On Agent And Optimizations RBF Neural Network

As computer networks and Internet applications developing, a wide range of information-sharing more depth into the people's working and living in all areas. It depends on the information sharing is gradually increased, and security as the based of information-sharing technology becomes more important. Intrusion Detection System (IDS) is a model of information security; it is an important component of dynamic security technology. Intrusion detection is essentially an electronic data processing, which is analyzed and processed in accordance with a predetermined strategy on the security audit collected data; it can make invaded conclusions which are based on the analysis made by the system.Firstly, the thesis summarizes Intrusion Detection technology and analyzes the deficiencies of traditional IDS. Secondly, the thesis studies the application of neural network in IDS in depth. BP network is one of the most widely used neural networks. In comparison with BP network, RBF neural network is not an error back propagation method but an entire front propagation method. Its training time is short and it is not easy to converge to the local minimum. RBF network has the best approaching capability, simple network structure, and high leaning speed. SO the RBF network is very suitable for intrusion detection test which has the high requirements of efficiency and speed. But the parameters of RBF is based on the partial information in the whole parameter space, it may not be able to reach the global optimal value. So the thesis uses PSO algorithm to optimize RBF network simply. The PSO algorithm is a tool which is based on an iterative work to search for the global optimal solution. It has the function of global optimization, and it can make up the deficiencies of parameters which RBF neural network sets up. Thirdly, the thesis introduces Agent technology and the application model of it in IDS.On these bases, the thesis designs an Intrusion Detection System which is combination of the Agent and optimizations RBF neural network. The system can automatically adapt to the complex and volatile environment, through self-study, self-evolution, it can improve the system's intrusion detection capabilities; it can make full use of network resources to cooperatively complete the detection task. Using this prototype system, the function of each module is described detailed, while design structure chart and implementation techniques of the system are discussed. Finally, the main realization process of the core module (that is, neural network module) is presented and using KDD CUP 99 data which can simulate intrusion of network to set an intrusion simulation experiments.