Research On Intrusion Detection Algorithm Based On Cultural Algorithm

Nowadays,the network security problem is increasing because of the development of Internet technology.The emergence of these problems makes the intrusion detection technology became a hot spot of research scholars.Among them,the immune intrusion detection is of great interest to scholars because of its strong adaptability,robustness and self-repair ability.In the whole process of immune intrusion detection,it mainly include the negative selection algorithm and the clonal selection algorithm.The negative selection algorithm is used to generate the mature detector,and the clonal selection algorithm is used to train the mature detector.Therefore,the clonal selection algorithm plays a vital role because it can improve the performance of the detector.In view of the traditional immune clonal selection algorithm that uses a single mutation,which leads to the problems of falling into local optimization or slowly convergence.By introducing cultural algorithm,this paper realizes the evolution of population space and belief space.This paper proposes clonal selection algorithm of adaptive hybrid mutation,which combines the strong global search ability of cauchy mutation with the strong local search ability of chaos mutation in the mutation.And it utilizes the knowledge of belief space to adaptively determine the time and the proportion of the two kinds of mutations.The algorithm is tested in KDDCUP99 data set,the result shows that the algorithm has good convergence and robustness.Due to the characteristic of bi-level evolution of cultural algorithms,and there is no unified method to optimize the parameters of Support Vector Machine(SVM).The SVM parameter optimization based on cultural algorithm is proposed.Based on Kernel Principal Component Analysis(KPCA),the new algorithm uses KPCA to reduce the dimension of data and then introduces cultural algorithm to optimize SVM's penalty parameters C and Gaussian kernel parameters ?.Finally,the SVM classifier is trained by using the data set after dimensionality reduction.The algorithm is tested in KDDCUP99 data set,the result shows that this algorithm can effectively reduce the parameter optimization time of SVM and reduce the classification error rate.