The Analysis And Research Of Intrusion Detection Model Based On Bayesian Network

With the development of computer technology and the popularity of internet , the network safety becomes more and more important.the management's primary target of intrusion detection is the safety which guarantees all computer's systems, network system and the whole information structure.Baysian network is not only a method of the indetermination deduction based on the probability, but also is the main tool which handles the indetermination information, This paper applies the characteristics of the Baysian network, investigates for intrusion detection system.Firstly,the paper analyzes the Navie Baysina Network(NB) in detail.the method regards all attributes in the training sample as sub- nodes of the root node, the sub- nodes are mutually independent, In the current NB structure,the NB acquires the network parameter only through the training sample computing the probability of root node and the condition probability of the sub-nodes in the root node,in the same time it raises efficiently the detection rate. Although the NB has the characteristics of higher detection rate,it exists the limits of same function among all attributes.Therefore,the paper proposes a new revision for NB,which is the model based on the Chi-square test.The model considers differences of attributes fuction, it deletes the redundancy attributes and does not concern them,so can simply the NB structure, the efficiency is raised in detection system. The Chi-square test can promise to acquire the exact characteristic, and reduce the calculation quantity, in the meantime, the model raises the classified efficiency,and makes the intrusion detection have a better efulness and adaptability.In the end, this paper carries on the analysis of an experiment to the improved NB system. The result expresses:The improved system has the advantage of high dependability and good detection rate.