Research Of Network Traffic Protocol Identification Algorithm Base On Markov Random Field

Traditional network devices for each network packets "equally, sharing resources"service form can’t meet the diverse needs of the user. In this case, the Internet serviceproviders hope the router can treat network data flow and the charge of network flowgradient or other methods intelligently to earn more network resources use cost withoutincreasing costing of upgrading the hardware. These requirements need network canrecognize different kinds of data traffic. Along with the network service diversification,the types of network traffic protocol and the complexity of network traffic increase rapidly,which leads to a lot of difficulty to identify network traffic protocol.Through the analysis of the various applications of network traffic protocol, wepropose the traffic data of network session header information as basic elements ofstatistics and obtain the characteristics of the network traffic. According to the MIV valueof BP neural network, we evaluate characteristics influence on the results, and ultimatelydetermine the optimal network traffic statistics characteristic collection. Basing on thecurrent research of some network traffic protocol identification technology, we proposeHidden Markov Model (HMM) to identify the protocol of network traffic.In the design of experiments, we use the software college of Huazhong university ofscience and technology, Intel multi-core to collect the data sets, and then applycharacteristics of optimal message statistics collection to train Markov model, whichmakes accuracy of the model achieve90%.After that the results will be compared withother classification methods, further verify the accuracy, briefness and efficiencies of thenetwork traffic protocol recognition method based on Markov random field.