Network signal processing and intrusion detection by a hybrid model of LSSVM and PSO

Abstract

In the paper, the hybrid model of particle swarm optimization and least square support vector machine is proposed to network signal processing and network intrusion detection, and PSO is utilized to select the parameters of support vector machine simultaneously. In the study, KDDCUP99 datasets are adopted to research the network intrusion detection performance of the hybrid model of particle swarm optimization and least square support vector machine. The detection accuracies for DOS, R2L, U2R and Probing of the hybrid model of particle swarm optimization and least square support vector machine are 96.7, 95.0, 95.0 and 95.0 respectively, the detection accuracies for DOS, R2L, U2R and Probing of least square support vector machine are 83.3, 82.5, 80.0 and 82.5 respectively, which indicates that the accuracies of the hybrid model of particle swarm optimization and least square support vector machine are higher than those of least square support vector machine. It is indicated that that the hybrid model of particle swarm optimization and least square support vector machine has a higher detection ability than least square support vector machine.