Research Of Network-based Intelligent Intrusion Detection System Model

At present, the technology of network security mainly includes firewall, access control, VPN, the data encryption etc. And they are all belong to passive defend measure which stands on protection. Intrusion detection is a dynamic security protection technology, which can settle the issues that the traditional technologies couldn't handle, and is a hotspot in network security at present. But the conventional intrusion detection has the problem of being lack of efficiency, which restricts the practical application of intrusion detection technology.This thesis mainly studies the method of intrusion detection, and applies some intelligent algorithms in artificial network and data mining subjects to intrusion detection. We improve the algorithm that we used to enhance the accuracy and reliability of intrusion detection, and design an intelligent intrusion detection classifier that has a good detection effect.This thesis firstly introduces the background, development of intrusion detection, the concept and theory of intrusion detection system, and compares each kind of detection technology to point out their advantages and disadvantages. Then the thesis presents a new intrusion detection method that combines artificial neural network and data mining according to the problems of current intrusion detection system, designs and implements an intrusion detection model. And the previous technology adopts the BP algorithm to train the neural network to determine its structure, so as to detect intrusions; the latter technology is based on decision tree algorithm in data mining. We improve the C4.5 algorithm, and adopt this algorithm to build a decision tree that reflects the intrusion's character, then use the intrusion detection regulation deriving from the decision tree to detect the intrusion on network. Finally we implement the intelligent intrusion detection classifier model that combines neural network and data mining.This thesis using DARPA'98 data which is authoritative in intrusion detection field to evaluate the intrusion detection model that we presents, and the experiment shows that the intelligent model's detection effect is high.