Intrusion Detection Model Based On Dimension Reduction And Improved MEA-SKohonen Neural Network

Intrusion detection system can detect and prevent attacks in time and ensure network security effectively.This paper focuses on the related algorithms of intrusion detection and artificial intelligence.Aiming at the problems of U2 R and R2 L classes in KDD Cup-99 data set,the amount of data is too small,and lead to low detection accuracy,an intrusion detection model based on genetic algorithm dimensionality reduction and improved MEA-SKohonen neural network is proposed to improve the detection accuracy of U2 R and R2 L classes.The main work of this paper is as follows:(1)In order to solve the problem of intrusion detection,this paper chooses the Kohonen neural network as the main algorithm,and selects the Elman neural network as the contrast experiment.By adding a output layer,the Kohonen neural network is replaced by SKohonen neural network,so that it can be changed from an unsupervised network to a supervised network.,the neighborhood function is improved and the detection performance of the SKohonen neural network is improved.Finally,the neural network is optimized using the mind evolutionary algorithm to determine the initial weight and threshold of the network.(2)Because the data amount of U2 R and R2 L categories in KDD Cup-99 data sets is too small,it is difficult for neural network algorithm to learn the characteristics of these two data categories,so the accuracy of detection is too low to meet the relevant requirements of intrusion detection system.Therefore,the genetic algorithm is used to reduce the dimensionality of the data set,and some redundant features are removed,so that the neural network model can better learn the characteristics of the two types of data and improve the accuracy of detection.(3)Using genetic algorithm to reduce the dimension of the dataset,finally,three different data are selected,the MEA-SKohonen neural network model and the MEA-Elman neural network model are trained and verified.The experimental results show that the accuracy of U2 R and R2 L classes can be effectively improved by using data dimension reduction method.Overall,using the KDD Cup-99 dataset to construct an intrusion detection model,the MEA-SKohonen neural network is superior to the MEA-Elman neural network.The detection accuracy of intrusion detection model based on dimension reduction and improved MEA-SKohonen neural network can meet the relevant requirements,and significantly improve the problem of low detection accuracy of U2 R and R2 L categories in KDD Cup-99 dataset,although there are still some shortcomings,it has better accuracy,robustness and generalization ability.