The Research On Intrusion Detection System Based On Artificial Neural Networks

As our demand for Internet grows in the recent years, many organizations and enterprises have established their own internal networks and have linked it with the Internet. While people enjoy the advantages of the network, the problem of computer security becomes more and more seriously. Therefore, how to detect, defend and suppress the various kinds of illegal activities in the network becomes urgency ,we must deal with at present for computer security.Hence, the intrusion detection technology occurs. The intrusion detection is a reasonable supplement to the traditional measures, such as firewall, data encryption, which can provide real-time protection against the internal attack, exterior attack and the misoperation. The intrusion detection work belongs to the classified recognition, which has massive non-linear problem. Therefore, It is essential to develop the ideal non-linear integration processing method. The artificial neural networks are a simulation of physiological structure and mechanism of human brain. They are good at processing these imperfect association memories, pattern recognition of characteristic damage as well as self-study of the rules. They also play an irreplaceable role in solving complex problems which are difficult to model. The neural network's learning and adaptation to this uncertainty can satisfy the needs of classified recognition in intrusion detection exactly.There are several hundred kinds of the artificial neural networks model. Among them, the radial basis function neural network is a newly-born and excellent one. The aim of this research is to introduce the radial basis function neural network into the intrusion detection,After analyzing the advantages and disadvantages of the entire raid of study algorithm and the on-line study algorithm respectively,on the basis of on-line study algorithm RANEKF (Resource Allocating Network Via Extended Kalman Filter), the author put forward an effective method to improve its deficiency. The result of the experiment indicates that this method has solved accurate classified problem in the intrusion detection better, and reduce obviously the rate of failing to report and misreport in intrusion detection rate obviously.The first chapter is an introduction, which tells the history and the present situation of artificial neural networks, and analyzes the advantages and disadvantages of the...