A traffic identification based on PSO-RBF neural network in peer-to-peer network

Abstract

To identify and control the peer-to-peer P2P traffic accurately, this paper proposes a novel classification method of peer-to-peer network traffic identification based on machine learning. The method is based on the conventional machine learning to identify the network traffic, in the data specimen collection stage, adding the particle swarm optimisation algorithm to collect the data. In the classification tool building stage, a radial basis function neural network is used, which is very suitable for specify the data category. In the experiment, collecting three kinds of typical traffic P2P traffic, HTTP traffic, and game traffic on the current internet, identification and classification, results show that the method has a higher precision rate and recall rate two evaluation indexes of three kinds of flow rate with more than 90%. The use of particle swarm optimisation feature selection algorithm reduces the training time. On the whole, the method has good classification effect.