Abstract: Various methods are capable of classifying Peer to Peer (P2P) traffic in coarse grained way efficiently and accurately. However, few papers focus on the fine grained identification of P2P traffic. Affinity Propagation (AP) algorithm is brought to P2P traffic identification field for the first time and a novel method of identifying P2P traffic finely based on Hierarchical Weighted Semi-supervised AP (Hi-WSAP) algorithm is proposed, which is on the foundation of the Hi-WAP algorithm and absorbs the semi-supervised thought. The proposed method identifies P2P traffic in semi-supervised way by application employing only 10 fast computed traffic features. Experimental results using two datasets show this method achieves high identification accuracy and low time complexity, which provides a path to finely identify P2P traffic in real-time.