With the rapid development of the Internet, the P2P (Peer-to-Peer) technology which is characterized by no utilization of any servers with centralized functions has kept advancing apace. However, the wide use of the P2P application has lead to the illegal file sharing in addition to the rapid growth of network traffic. Thus, it has become an urgent requirement to discriminate the users of P2P file sharing application from other users. This paper proposes some features including "new connection rate" of P2P application traffic to achieve its discrimination. SVM (Support-Vector-Machine) is applied to the discrimination based on these features in order to automate the processing of the discrimination, and its accuracy is demonstrated as a result of some experiments.