With the rapid development of Internet technology,the attention of network security issues continues to rise,and intrusion detection technology has become the hot point of research.This paper mainly studied the gaussian mixture model and K-means in clustering algorithm,corresponding improvements were proposed to be applied to network intrusion detection,the detection effect was verified by experimental simulation.The main research contents and contributions of this article were as follows:(1)Aiming at the sensitivity of gaussian mixture model to initial value,this paper proposed a clustering algorithm of gaussian mixture model based on improved density peaks(DP-GMMC).First,it calculated the local density and distance by similarity matric,improved the density formula using ? law 15 fold line method,determined the number of clustering and corresponding clustering center according to the decision graph.Secondly,the gaussian mixture model was used to calculate the logarithmic likelihood function of sample points.Then,the maximum expectation algorithm iteratively updated the mean,variance and mixing coefficient,and completed the division of the sample points according to the probability.Simulation experiments show that,compared with GMM algorithm,DP algorithm and AHC-EM algorithm,the clustering accuracy of DP-GMMC algorithm could increase the percentage points of 33.6,17.34 and 1.53 on Iris data set,and could obtain accurate clustering number.(2)Aiming at the defect that K-means algorithm randomly selected cluster center,this paper proposed a K-means algorithm based on improved flower pollination(IFPK-means).First,the algorithm used chaotic strategy to increase the diversity of the population.Secondly,the flower pollination algorithm was iterated to get the optimal value.Then,tabu search was introduced for optimization.Finally,the K-mean algorithm completed the final clustering.Simulation results show that,compared with ABC algorithm,PSO algorithm,DE algorithm,FPA algorithm and DEFPA algorithm,the average clustering accuracy of IFPK-means algorithm could respectively increase percentage points of 13.26,5.41,6.59,12.7 and 4.33.(3)In order to verify the performance of the improved clustering algorithm in intrusion detection,this paper firstly performed numerical,standardization,dimensionality reductionand and other processing on the intrusion datasets.Then,the DP-GMMC algorithm and IFPK-means algorithm were used to perform comparative experiments on the data set.Simulation results show that,on the NSL-KDD dataset,compared with GMM algorithm and DP algorithm,the detection rate of the DP-GMMC algorithm could increase the percentage points of 8.14 and 3.94;compared with K-means algorithm and DEK-means algorithm,the detection rate of the IFPK-means algorithm could respectively increase the percentage points of 7.31 and 4.71.The paper has 38 figures,20 tables and 62 references. |