Research And Realization On Intrusion Detection Systems Based On Neural Networks

With the Internet developing. Electronic government and E-business applying, as closing data shows that annual loss caused by the security system of the computer was destroyed reaches several ten billion Yuan on the whole world. And the trend of loss is still very ascendant obviously. So the question of security has already presented in front of us.Intrusion detection is defined as "the problem of identifying individuals who are using or probing a computer system without authorization and those who have legitimate access to the system but are abusing their privileges" . This concept was expression since by James P in April of 1980. Through development of more than 20 years, it became focused research in security field with intrusion happening frequently.This paper analyses the merit and flaw of intrusion detection system existingly, points out the shortcoming and challenge faced, and concludes its developing direction. To form node intrusion detection system by combining the character of host intrusion detection system and net intrusion detection system.Node intrusion detection designed in this paper can detect the inner intrusion action and out intrusion action through analyzing the data source from BIDS and NIDS. Giving full play to HIDS and NIDS each merit and making them supplement each other, this detection system reaches better detecting result. We select neural network as application to detect, for considering the influence of large numbers of uncertain factor and noise, and the neural network just has the ability of self-study and fault-tolerant. In inputting and exporting layer, We uses degree of membership to fuzzy input or output data, better to solve the question of sharp border effecting problem.This system regards WindowsNT as the development platform, utilizes VC6.0 language to develop and uses Matlab to train the test data. According to the software engineering principle, this system adopts modular design ideals and parallel development tactics to develop and finishes an experimental intrusion detection system.