Network Intrusion Detection System Research Based On Intelligence Algorithm

With the development of the computer and network technology, it has melt into every aspect of our society, it promotes the progress of our society, economy and culture. At the same time, it brings a serious problem——network security. The shortcoming of traditional safeguard technology is remedied by intrusion detection technology, so IDS is an active safeguard measure and brings a main trend research into network security. Unfortunately, there are many problems existing in most intrusion detection system at present, such as low detection rate, high false alarm rate and bad integrated performance.To find some effective methods for resolving these problems, this dissertation studies the structure of NIDS(network intrusion detection system), intelligence algoritm for detection and interactive technology. According to the further study on the correlation theories and methods, an interactive network system between linkage defense and multi-agent IDS is designed, multi-sensor information fusion and projection pursuit algorithm are proposed in this paper. The main research works of this dissertation can be summarized as follows:(1)To resolve the shortcoming of network intrusion detection and firewall, the interactive technology is introduced into the multi-agent IDS, and we design a kind of linkage model. Then we build data base, strategy control centre and network agent,as well as, has laid the foundation for the intrusion detection's related intelligent algorithm at the next step.(2)The multi-sensor information theory is introduced into the network intrusion detection, and make a strategic decision based on D-S evidence theory is proposed. In the system, each sensor is regard as a independent classifier which sorts the data by grey relational analysis. And then the result of class from each sensor are fused by the fusion centre. The simulation shows that intrusion detection performancecan be improved by information fusion.(3)An intrusion detection based on optimization projection pursuit algorithm is proposed. Aimed at the optimization problem of calculation the best projected vector of projection pursuit, a method is used based on QPSO. Optimization projection pursuit algorithm is implemented to project the data of feature from high dimension to low dimension, then the category of the network data is determined by the clustering algorithm. The algorithm include data preconditioning, data normalized, the best projected vector calculation, data project to low dimension, and finally inspection data categorization. The experimental results show that the performance of the present method is feasible and effective.