Study On Network Intrusion Detection Based On Fuzzy Neural Networks

With the development of the application of internet,corporations,goverments and universities all have build their own websites and provided some services,which brings us more convenience and benefits.But there are so many secuity problems rising during the development of the Internet.If we do not resovle them,these security problems would take Internet into hell.As the footstone of the Internet and IT development,the network security technology has been one of the most important research subject in computer realm,and been continuously expanded and enriched from practice of solution to the network security problem.Except for multi-network security technology(such as VPN,encryption,firewall),Network Intrusion Detection is another important method in network security.Firstly, this papaer outlines some research present situation about network security and intrusion detection, introduces concept,function,method,work principlend and evaluate criterion in relation to intrusion detection, and analyzes handicap facing with traditional intrustion detection in new network environment.Secondly,through titanic research and experiment on Self-Organizing Map(SOM) neural network,and fuzzy improvement on learning algorithm of SOM network by fuzzy logical thought,this paper has attained an more fast and efficient fuzzy neural network.And then,based on the exception detection research on network intrusion,we apply SOM neural network on intrusion detection, and analyze it within intrusion detection model,to construct a neural network-based detection model.Specifying its realization,considering character of network intrusion detection,we present relevant detection rule and clustering analytical method aimed at intrusion detection,then proceed to the next step to raise a new detection algorithm - SOM clustering-based network intrusion detection one.This algorithm makes use of SOM to cluster object,and devides object character space to distinguish normal and intrusion behavior. For more efficiently decrease false positive rate and increase detection rate,and taking fuzziness that intrusion detection has itself into consideration,we improve SOM clustering-based network intrusion detection method,and present fuzzy SOM network intrusion detection method. In the end,we use MATLAb to proceed emulation experiment on fuzzy SOM-based and SOM clustering-based network intrusion detection.By comparing the results of these two kind of experiment,conclusions can be acquired that the correctness rate of former is slightly high than that of latter.Thus there is practice sense in fuzzy improvement on SOM.It makes SOM network abstract stastics feature of data detected more correctly and can increase some edge detection rate of intrusion detection.Meanwhile,results show that it is feasible,efficient,and good-expansionary that fuzzy neural network-based network intrusion detection method this paper presents are faced of unknown intrusion detection.This method can reduce false positive rate in effect,and elevate detection rate simultaneously.