| In the age of big data,How to ensure the security of network has become a hot research topic in today’s research.Previous security technologies such as digital encryption,firewall,VPN and so on have effectively improved the network security.However,with the development of intrusion technique,complex intrusion technology can easily crack security technology.In this case,the research of network security intrusion detection(ID)makes network security’s research upsurge.Subsequently,active and intelligent intrusion prevention system(IPS)emerges as the times require based on IDS,which makes up the defect that IDS can not deal with attacks intelligently.Some related literature of intrusion detection algorithm are studied and analyzed which are about shortcomings of current intrusion detection system.It proposes a kernel extreme learning machine and particle swarm optimization algorithm based on intrusion detection in this paper,improving the performance of intrusion detection system.An algorithm based on improved particle swarm optimizing multi kernel Extreme Learning Machine(PKELM)is proposed for the intrusion detection,that based on the single kernel limit learning machine which has high false alarm rate,slow convergence speed and weak generalization ability.In the algorithm,through the kernel function of Mercer to make multi kernel function which solves single kernel machine in poor robust performance and low rate of detection;Then through the Gauss disturbance and other ways to improve the local search ability of particle swarm algorithm,which optimizes kernel parameter and the regularization of the multi kernel extreme learning machine factor,improving the kernel extreme learning machine convergence speed and generalization ability.At the same time,aiming at the huge number of network data and the discrete feature distribution,a clustering algorithm based on Improved Particle Swarm Optimization(K-Means)is proposed(IPMeans).In this algorithm,the clustering center is improved by particle swarm optimization algorithm,to improve the clustering ability of K-Means algorithm,and then it is used to deal with the algorithm of intrusion data,to increase similar data aggregation,making data is more easily identified by the intrusion detection system,improving the ability of processing massive data and the speed of detection system.An intrusion detection algorithm based on IPMeans-PKELM is proposed by combining the optimized multi-kernel learning machine and the optimized K-Means clustering.Thealgorithm adds the kernel parameter optimization and clustering function of intrusion data based on the KELM.Relative to the original intrusion detection algorithm which deals with high-dimensional data in low detection rate,and set hidden layer section randomly,causing the results of testing errors to be big,the IPMeans-PKELM algorithm increases intrusion’s recognition and the speed of the intrusion detection system through introducing IPMeans algorithm to cluster intrusion data.At same time,and the improved particle swarm algorithm is used to optimize kernel parameters,improving the generalization ability and the rate of intrusion detection system.Finally,simulation experiments are carried out in the KDD CUP99 environment,the processed data are split with 10-CV and nine of data are trained for KELM classifier,one of data is used to test.The experiment results shows that the method effectively improves the intrusion detection rate while reducing the false alarm rate. |
