Generation Algorithm And Application Of Artificial Immune-based Detectors

The development of modern network technology brings human lives great convenience, but it also generates large number of new type network attacks. The traditional means of prevention are difficult to deal with these new network attacks, so intrusion detection technology has developed rapidly as an emerging network security technology. As the intrusion detection systems have surprisingly consistent functional requirements with biological immune systems, numbers of scholars try to build artificial immune-based intrusion detection systems, in order to achieve advantages of biological immune systems such as distributivity, robustness, self-adaptive.In the artificial immune based intrusion detection systems, mature detectors complete the tolerance and generation process mainly by negative selection algorithm currently. The mature detector's selection criteria are affinity. The traditional R contiguous match based negative selection algorithm emphasis on local exact match but can not accurately describe the affinity between antibodies, so it lead to missed and false detection easily and need to be improved. This paper proposes two modified negative selection algorithm through the reference of fuzzy mathematics, applied to different intrusion detection environment. In addition, there are many adaptability and completeness shortcomings in the intrusion detection model LYSIS proposed by Hofmeyr. Narrow range of the test objects, and the resulting memory detectors have to antigen is not high, seriously affecting the efficiency of the system. To complete these problems this thesis initial validate the model LYSIS by using negative selection algorithm and the immune evolutionary algorithm.This thesis contains the following works:1. Analysis the average time cost of self detectors and non self detectors.2. Use fuzzy mathematics method to improve the negative selection algorithm based on r-contiguous matching rule, make the detector generated farther away from the self,and shape of detectors irregular,in order to reduce the black holes and improve detection rate and to lower the false alarm rate. The simulation experiments based on IRIS data sets show that the algorithm improved the detection rate and false alarm rate performance superior to traditional algorithms.3.Improve the model LYSIS by using Negative selection algorithm and the immune evolutionary algorithm,and propose a new detection algorithm,Increased affinity of the memory detector generated,pre-estimated the black holes use priori knowledge reducing the incidence of undetected. The evaluation of the algorithm based on DARPA 98 sets shows that the new algorithm can generate efficient memory detector.