Intrusion Detection Technology Based On Fast Neural Network

Computer resources and internet safety absolutely is one of the most attractive study areas in the world because of the wide spread of the computer, the high development of the internet and the fragility of the system. Attackers make use of the leaks, such as the leaks of the system software, the flaw caused by inappropriate management deployment and the weakness of the network protocol, to intrude the system. Recently, new intrusion named Phishing, Botnet and waste information are emerging. Network attacks have become more complex, flexible and extensive than ever before, there are a lot of challenges and dangers in network security now.Intrusion Detection as a proactive security protection technology provides an internal attacks, external attacks and misuse of real-time protection. Neural network has the ability of generalization and abstract; can deal with uncompleted information and adapt network environment; possess the high learning ability and adaptive ability; can identify the characteristics of completely new intrusion; can overcome the limitations come from the expert system detection technology. By putting the neural network learning method in the intrusion detection, it can effectively improve the detection rate, reduce missed alarm rate. The innovation of the dissertation are as follows.To solve the shortcomings of the traditional BP neural network algorithm, such as local minimum in the study, studying too long time and converging hard, after learning hard about the traditional intrusion detection and neural network technologies. Put the Extreme Learning Machine (ELM) in the intrusion detection, use the KDDCUP99 dataset to compare BP neural network and ELM neural network separately by simulation experiment. The experiment results show that this algorithm can improve the learning speed, reduce learning time and improve the detection rate.This paper researches and analyses the deficiencies of the currently existing intrusion systems, then construct an intrusion detection system based on the Extreme Learning Machine. This paper realizes this model by presenting the processes of implementations for main modules, including data collection module, data analyzing module, pre-treat module, neural network module, responding module and rule set module. Then tell on the structure diagram of the module and the principle and realizing way for each module. Lastly, the paper analyses the problem of the realization and the difficulties which the intrusion system will be confronted with.