Network Anomaly Traffic Classification and Optimization Based on PSO-SVM

Abstract

A network traffic classification model and optimization method based on PSO-SVM is proposed in this paper to solve the difficulties of traffic classification and its low-performance model in intrusion detection system. Based on the expansion of SVM from the two-category traffic classification structure into the five-category structure, a hybrid kernel function combining Poly and RBF is constructed by model to ensure the generalization ability and model learning; and then after conducting particle swarm optimization on the various parameters of SVM model, the search spaces tablished by nonlinear inertia weight coefficient and learning factor of asynchronous optimization are conducted with fitness evaluation to achieve the optimal solution and enhance the convergence ability of algorithm. The experimental results show that the network traffic classification model and optimization method based on PSO-SVM proposed in this paper can achieve traffic classification and improve the performance of classification model.