Traffic Identification Method for Specific P2P Based on Multilayer Tree Combination Classification by BP-LVQ Neural-Network

P2P has become an important traffic form of current Network, and P2P identification has been a hot research in Network monitoring & management area. In order to tackle the problem of P2P data encryption persecuting P2P identification, in this paper, a traffic identification method for specific P2P based on multilayer tree combination classification BP-LVQ Neural-Network was proposed, baseing itself on traffic characters of P2P. This method improved the P2P identification with BP Neural-Network, by abstracting attributes of P2P flow statistics, selecting the optimal attribute subset, establishing a P2P classifier through the multilayer combination with BP Neural-Network and LVQ Neural-Network. After associating with the confidence level of specific P2P, the identification result was acquired. Experimental results showed the validity of the proposed algorithm which performed better considering identification accuracy and time consuming.