Research On Generation And Distribution Optimization Of Immune Detector Based On Deep Belief Network

The artificial immune system can accurately recognize the external invasion by simulating the biological immune system.It is widely used in network intrusion detection.However,due to the increasing traffic of network data,the detection accuracy and efficiency of the algorithm based on negative selection,which is often applied,are low.In this paper,the dimensionality reduction of the initial data and the distribution of the detector are improved respectively,and a method for generating and optimizing the distribution of the immune detector based on the deep belief network is proposed.The main contents is as follows:First against the Negative Selection algorithm(Negative Selection algorithm,NSA)in detector using the set of initial self generated,because of the characteristics of the data set is very complex,result in the problem of low efficiency of detector,this paper puts forward a immune detector generation algorithm based on depth belief network,the depth of belief network applied to feature extraction and optimization to pretreatment dimensionality reduction of high-dimensional data,greatly reduce the data dimension of the original data set,removing a large amount of redundancy,the algorithm ensures the raw data before and after the dimension reduction of high-dimensional feature reservation to the greatest extent,The generation efficiency of the mature detector is improved by using the negative selection algorithm.After dimensionality reduction,when the detector is randomly generated by negative selection algorithm,the detector's utilization rate is low because the randomly generated detector cannot be evenly distributed in the non-self space.A distribution optimization algorithm for immune detector based on particle swarm optimization is proposed.The algorithm will initially generated by the detector to optimize distribution and variation process of using the clonal selection algorithm increases the diversity of antibodies,the evolution equation of particle swarm optimization algorithm is used to guide the direction of the antibody mutation,improve the convergence speed,covered with a detector of autologous density calculation of fitness,make all detector are gathered in the abnormal samples density larger area,and handle conflicts between detector,detector can uniform distribution in the space of the self,utilization was improved.The NSL-KDD data set was used for the simulation experiment.The results show that the method performs well in the detection efficiency,accuracy and false positives of the detector.