Abstract: With the continuous development of wireless sensor networks, malicious nodes was a major threat to its security. The traditional credit threshold model could not accurately identify the malicious nodes, such as sub attack node. And there will be a low recognition rate and high false positive rate and other issues. To solve these problems, a new recognition method based on DPAM-MD algorithm was introduced, which was based on the traditional credit threshold model, combining manhattan metric with DPAM algorithm to distinguish the sub attack node. A new clustering algorithm based on density was proposed, which was based on the objective function of the distance equalization between the intra-cluster and the inter-cluster. The proposed algorithm can improve the quality of clustering, shorten the time of clustering, and improve the efficiency of recognition of malicious nodes .Through simulation results ,verify the improved algorithm effectively identify undistinguishable node.

Key words: wireless sensor networks; malicious node; DPAM-DM