The Research Of Intrusion Detection Technology Based On Artificial Neural Network

Internet has played a more and more important position in people's lives. With the development of it, business or personal affairs have become more convenient. Thereby, it gives rise to the importance of the security for network, and it enhances the more importance of network-based security technology. Because of this, the intrusion detection technology, which has its characteristic of intelligent monitoring, real-time detection, and dynamic response, becomes an inevitable complement to the traditional security mechanism. It now has become a direction of researching and developing dynamic security tools. On point of the intrusion detection technology, the soft computing is the trend for detecting technique, including neural networks. With the capacities of self-adaption, self-organization and self-learning, Neural networks can conduct some background knowledge that is not so clear through training. Based on the artificial neural network, this thesis goes further to make a study and discussion on intrusion detection techniques.The paper briefly describes the common network attacks and network security technology. It then introduces the basic concepts of the intrusion detection system (IDS), and makes an analysis of the basic principles of intrusion detection. With the comparison of the ever existing intrusion detection models and intrusion detection techniques, it goes ahead to have a discussion on the development and challenges of intrusion detection technology. While the main point of this paper is to describe the advantages of intrusion detection technology, which is based on methods of the artificial neural network.The artificial neural network methods mentioned in this thesis is based on BP neural network model. On the study of the basic concepts, standardized BP algorithm and the improved BP conjugate gradient algorithm, this paper has constructed a IDS model that is based on BP neural network. According to the model of the design goals, the basic framework of the system has given out in this paper. And key researches and elaboration have been given out by way of the implementation of the system's technologies. Finally, this paper focuses on the feature selection and normalization of the selected data from the simulation network packets which contains the KDD99 data. and it goes further to enhance a test in RMATLAB R2009 which can validate the system's effective performance by way of the experimental data .