The Research Of Intrusion Detection System Based On Artificial Immune

With more and more attention paid on the network security, the traditional securitytechnology, with firewall represented, can not adapt to the development of the times. Asa new type of dynamic protection technology, intrusion detection technology makes upfor its shortcomings very well, and it is an important part of modern network protectionsystem. Inspired by the natural immune system, detection technology based on artificialimmune theory intrusion catches the attention of people. It has many benefits, such asdistributed, diversity, robustness, adaptability, and provides a new direction for thecurrent intrusion detection technology research.In the immune intrusion detection system based on dynamic colonel selection,the quality of the mature detector has an important impact on system performance, and adetector coverage of all non-self set can effectively improve the detection efficiency andaccuracy. First of all, in the existing mature, the detector set not only would be a wasteof valuable detector space, but also can not cover the non-self space very well. Secondly,in order to ensure proper false detection rate, the system requires too much artificialstimulation so it reduces detection efficiency of the system for the performance of theserver will have an impact. Finally, detection system based on artificial immuneintrusion can only detect whether the invasion once happened, but can not identify thetype of abnormal data. However, it is of great significance for the analysis of networkinvading.In order to be able to identify abnormal data types and to solve the problem ofoverlapping of the detector, this paper presents a detector identifying learning andoptimization algorithms. The proposed algorithm sets each non-self antigens as centerclustering, the identity of the detector within each class, the type of data is used toidentify abnormal. Meanwhile, in order to solve the problem of overlapping, afterN-generations mature detector it optimizes the inner portion of the detector for higherconcentrations of class variation and deletes to ensure that the diversity of the detector,and to improve the coverage of the detector pairs non-self antigen.According to too many requests of the artificial stimuli, this paper adds ambulatorymodule based on Bayesian and decision tree to replace the administrator manualintervention.When it needs synergistic stimulatory signals, it will use the module to give feedback signal. If it reaches certain conditions, it will directly send co-stimulatorysignals, or ask the administrator to give an acknowledgment signal, so that the systemcan quickly respond to requests from the stimulation of the detector module. Thereforeit will improve the quality of the detector, and make the intrusion detection systemabnormal react more efficiently.At the end of this paper, it tests the improved dynamic clonal algorithm by puttingforward above through the simulation experiments and experimental results shows thefeasibility and effectiveness of the improved algorithm.