The Research Of Intrusion Detection Technology Based On Back Propagation Neural Network

With the rapid development of computer technology and the widely applied ofinternet,network intrusion attacks happen frequently.Besides,as intrusion types come outafter the other,it is hardly for the firewall to propect network independently.So experts andorganizations commit themselves to raise more powerful and active protection strategies toenhance the network security.Intrusion detection is one of the effective approaches.This paper starts with the related presentations of intrusion detection technology andit’s development tendency firstly.Then we analyzed and summarized the defects oftraditional intrusion detection,such as low detection capability to unknown networkattack,low real-time detection capability to attracks,high false positive and negativerate.Recently,Neural Network has been applied in intrusion detection which opened an newavenues.Then this paper introduces some related theories of Neural Network and focus onthe research of BP nerve network(Back Propagation Neural Network) which has beenextensively used.This paper disserts the basic idea,principle,algorithm derivation,existing problems andseveral popular improved methods of BP nerve network as well.Then the paper aims atintrusion detection,makes improvement on the basic of Adaptive Learning Rate andproposes a advanced intrusion detection algorithm.On each iteration,it will use differentlearning rate to adjust threshold and connection weights between the various nodesaccording to the change of error and local gradient,the madificiation on weights will bemore focused as well. In addition,we introduce a transfer function with adjustable factor toraise the recognition rate and convergence rate of the algorithm.Lastly,the simulation experiment under Matlab7.0is carried out with the use of partialdata in KDDCUP99database. It makes a comparison of the new improved algorithm thatthis paper proposed and other methods:the Adaptive Learning Rate and the truditionalalgorithm of BP. Through the analysis of the experimental results,it shows that the proposedalgorithm achieves the desired consequent.