Support Vector Machine-based Network Intrusion Detection System

Network Intrusion Detection System as an actives security tool provides immediate protection against internal attack, misoperation and external attack. With the development of network, network intrusion detect system is challenged by the new situation. The current products of NIDS are mostly on the accuracy of detection not high, the phenomena of net data lost becoming worse. Support Vector Machine based on statistical learning theory is a machine learning method, it has a good many advantages in intrusion detection field, such as high processing speed, power extensibility. That can mostly improve the performance of IDS, and enhance the adaptability to high-speed network environment.The thesis mainly discusses the application of SVM in NIDS. To begin with, the thesis summarizes some related knowledge about intrusion detection and Support Vector Machine, a NIDS model based on SVM is designed on the basis of great of experiments. Then, the thesis introduces a heuristic fast learning algorithm of SVM, and uses it in intrusion detection system. In order to make support vector classifier obtain better performance, this thesis uses heuristic rules to select the most advantageous samples, and determine the parameters of SVM by train them. It is used to solve the problem of doing it with experience. In order to improve the learning speed, it uses the inner product matrix decomposing algorithm to improve the classifying speed.Network intrusion detect system is an essential component of whole network information security system. According to the research on NIDS based on SVM, it is an important area in NIDS research. Although there are many problems to solve, and have a relatively large gap from the practical applications, it does not affect the bright prospects. Sooner or later, it will be applied in practice.