Identifying Peer-to-Peer(P2P) traffic accurately has important influence on network flow control. A new P2P traffic identification method with high accuracy is proposed. This method calculates the frequency of 256 ASCII bytes occuring in packet header and turns it into a 256 dimensional statistical feature. Combining transport layer features and packet header statistical feature, this method identifies P2P traffic by means of decision tree algorithm. Data deblocking is proposed to maintain high accuracy and collect port numbers that relate to P2P traffic. The experimental results demonstrate that this method can distinguish P2P traffic from non-P2P traffic in different situations with high accuracy.