The Research Of Intrusion Detection System Based On LM-BP Neural Network

With the fast development of Internet, security problems become worse. Traditional network security technologys(such as VPN, Encryption and Firewall) can't fit the development of network technology, so comes the intrusion detection technology. It has not only detected intrusion out of network, but also surveillance the inner users'actions which didn't get authorization. And it becomes important more and more. But on the upgrade of intrusion technology, the intrusion detection system is facing great challenges now. In this paper by probing into the combining of intrusion detection and neural network, the main work are as follows:Firstly, this paper analyses the merit and flaw of intrusion detection system existingly, points out the shortcoming and challenge faced, and concludes its developing direction. At the same time, the structure and learn methods of neural network is studied. Because of the merit of self-studing, self-adaption and parallel_computing, a new intrusion detection model which combining misuse detection and neural network is considered important by author.Secondly, in order to solve the low speed in constringency, getting into local minimum easily and the heavy calculating problem of traditional error back propagation neural network(BP NN), an optimized Levenberg-Marquardt(LM)arithmetic named LM-BP NN by author is presented. By using Matlab, a best network parameter is finded, and at the same time, Experiments prove the LM-BP NN is better than traditional BP NN.At last, for a studying on neural network and intrusion detection system, a intrusion detection system(IDS) based on LM-BP neural network is presented. Experiments prove the new IDS has lower rate of false alert and omitting alert, it could recognise unknown attacks rapidly.