P2P traffic identification based on transfer learning

With the rapid development of Internet, a large number of peer networks (Peer-to-Peer) applications rise and are widely used. Because of this, it is more difficult for network operators to manage and monitor their networks in a proper way. To identify the peer networks applications generating the traffic traveling through networks is necessary and if we can identify them sooner, we control them better. In this work, we use the machine learning-based classification method to identify the classes of the flows. According to previous work, we choose transfer learning algorithm to classify the traffic, and improve classified results. Finally we compare and evaluate the classification results in terms of the two metrics such as true positive ratio and time expense. Our experiments show that the machine learning algorithm is an efficient algorithm for traffic identification and is able to build a quick identification system.