Research On Anomaly Detection Model Of Network Traffic Based On Adaptive Support Vector Machine

The network has created a lot of convenient and quick experience for people,but also become a tool for the illegal elements to abuse.They use the network to attack the private or public system,which has caused huge losses to people.Therefore,the problem of network security is more and more concerned by the public,especially the researchers,and the research of network traffic anomaly detection method is the focus.Through the research of the existing network traffic anomaly detection models,it is found that these models have the problems of low prediction accuracy and high false alarm rate.Therefore,this paper establishes the network traffic anomaly detection model based on adaptive support vector machine algorithm.The main work is as follows.First of all,the network traffic data has the problem of high feature space dimensio n,in which redundant features will increase the computational complexity of traffic anomaly detection and reduce the efficiency of anomaly detection.In order to solve the problem of high feature space dimension of network tra ffic data,we use Relief algorithm to search and promote chromosomes in the population of genetic algorithm,and propose R-GA feature selection algorithm to select features of network traffic data.Secondly,it is found that good feature subset and appropriate kernel function parameters are two important aspects that affect the prediction results of the model.Generally,the kernel function of SVM needs to be set manually.In this paper,an adaptive SVM algorithm is proposed to automatically generate the kernel function parameters,which reduces the impact of human setting on the prediction performance of SVM.Thirdly,because the use of all features of traffic data for anomaly detection will affect the efficiency and effect of detection,this paper uses R-GA feature algorithm for feature selection,and uses adaptive support vector machine as classifier to establish the R-GA-SVM network traffic anomaly detection model.Finally,the prediction performance of the model is tested by UNSW-NB15 dataset,and the validity of the anomaly detection model is proved by the comparative experiment.