针对P2P网络的分布式特点,依据节点在单位时间内连接的目的子网数量(d值)和节点单位时间内连接的目的IP数目与有效连接数目的比值(m值)特性,提出一种基于节点连接特性的P2P节点识别算法。P2P节点的d值和m值都明显大于其他节点(如典型的HTTP节点),并存在一个阈值区间,据此可高效识别P2P节点。在清华大学校园网上的实验结果显示,该算法比当前主流的算法识别效果更好,节点误识别率和丢失率都小于5%。 According to the distributed characteristics of P2P networks, a method is proposed based on the number of destination subnets (d value) and the ratio of the number of destination IPs to the number of valid connections (m value) per node in a unit time P2P Node Identification Algorithm for Node Connection Characteristics. P2P nodes d and m values ​​are significantly greater than other nodes (such as a typical HTTP node), and there is a threshold interval, which can effectively identify P2P nodes. Experimental results on Tsinghua University campus network show that this algorithm has better performance than the current mainstream algorithms, and the error recognition rate and loss rate of nodes are all less than 5%.