Research On Intrusion Detection Based On Improved Crow Search Algorithm And Optimized Support Vector Machin

With the advent of the Internet era,computer network has become an important part of society,affecting all aspects of people’s life and work.The large amount of information stored in the network not only brings convenience to people’s life,but also poses a serious threat to people’s information security caused by malicious attacks in the network.Intrusion detection,as a security defense mechanism,detects and responds to various attacks in the network to improve the security of the network space and prevent various vulnerabilities and threats.In order to improve the classification performance of intrusion detection,this paper addresses the problems of unbalance of data categories in intrusion detection data set and improper selection of parameters of support vector machine(SVM)classifier.An intrusion detection model based on improved crow search algorithm(ICSA)was proposed to optimize SVM parameters and an improved KADASYN algorithm was used to balance intrusion detection data sets to further improve the classification performance of the model.The main research work of this paper is as follows:(1)Improved scheme based on crow search algorithm(CSA).In order to solve the problem of uneven distribution of crow population caused by random initial population in crow search algorithm(CSA),the stability and uniformity of Latin hypercube sampling(LHS)were used to initialize crow population and improve the diversity of crow initial population.Aiming at the problems of insufficient search ability and slow convergence of CSA algorithm,dynamic perception probability is used to replace fixed perception probability,and local search and global search are adjusted to accelerate the convergence rate of the algorithm.Aiming at the problem that CSA algorithm is easy to fall into local optimization,the weight values of Levy flight and entropy weight method are introduced into the algorithm formula to give different weights to individual crows,so as to help CSA algorithm jump out of the local optimal extreme value.The test results show that the proposed ICSA algorithm has certain advantages in parameter optimization,convergence speed and robustness,which verifies the effectiveness of the proposed ICSA algorithm.(2)ICSA algorithm was used to optimize the intrusion detection model of SVM parameters.The reasonable selection of penalty factor and kernel function parameters in SVM has an extremely important impact on the classification performance and generalization ability of SVM.In this paper,the ICSA algorithm is used to find the optimal combination of SVM parameters,and then the ICSA＿SVM model is used to classify the NSL＿KDD dataset.The results indicate that the accuracy of the ICSA＿SVM model on intrusion detection datasets is as high as 92.43%.(3)An improved scheme of K-means clustering algorithm based on adaptive Synthetic sampling(ADASYN)algorithm.KADASYN algorithm based on K-means and ADASYN hybrid algorithm is proposed to solve the problem that ADASYN algorithm does not consider the characteristic information between minority samples when generating minority samples.KADASYN algorithm is applied to NSL＿KDD data set.While increasing minority samples,random under-sampling is used to reduce majority samples to balance the data.The balanced data set is tested on ICSA＿SVM intrusion detection model.The experimental results show that the addition of the proposed KADASYN algorithm results in a detection rate of 92.48%and 88% for the ICSA＿SVM intrusion detection model on a few categories of R2L and U2R,respectively.