Classification Of Network Traffic Based On Wavelat Kernel Extreme Learning Machine

In recent years,with the rapid development of the Internet,the scale of the network continues to expand,and various new kinds of network applications are increasing,and the network composition becomes more and more complicated.Network traffic classification,as one of the basic technologies to enhance network controllability,is of great help to network security management and network structure improvement.The development of network traffic classification technology can be summarized into three stages:the first stage is recognition method based on port number,but with the rapid development of network application,the method of classification rate gradually decreases;the second stage is based on the classification method of payloads.Considering the disadvantages of this method,researchers have gradually reduced the use of this method.Finally,the third stage is developed based on the recognition method of machine learning.This paper introduces Extreme Learning Machine(Extreme Learning Machine,ELM)method for flow classification,Extreme Learning Machine is a kind of contains only one hidden layer forward neural network structure,the traditional multi-layer forward neural network is implemented based on gradient descent,so in the iterative process,naturally increased the Learning time and power consumption,but also often trapped in local optimum.By contrast,ELM greatly reduces learning time and improves the accuracy of classification.In order to solve the problem of randomly generated input layer weight and implicit node bias in the algorithm process,the Kernel ELM(KELM)is adopted.At the same time,considering the network traffic itself has the characteristics of the length of similarity,the introduction of wavelet function,classification of wavelet function transformation,this is bad for the characteristics of decomposition,so as to improve the classification accuracy of the overall algorithm,finally constructs the wavelet kernel extreme learning machine(WK-ELM).By constructing wavelet kernel extreme learning machine(WK-ELM)a traffic classification model,and in a public data set and the local collection on the traffic data of experiment and analysis,from the point of the experimental results,this article constructed the classification method has achieved a higher recognition rate on the flow classification,and greatly reduce the execution time.Therefore,based on the wavelet kernel limit learning machine,this paper is a real and feasible classification model of network traffic.