Most wireless sensor networks（ WSNs） are deployed in outdoor environment

malicious node is a major threat to the safety of WSNs. Credit threshold model has a good performance in the field to identify the malicious node

but with the attack technology upgrade

the malicious node become more conceal and difficult to be found.The credit threshold model is improved by introducing the concept of time sequence and similarity through analysis

using Euclidean metric and K-mediods algorithm can distinguish this kind of malicious node effectively.Through simulation test

verify the improved algorithm effectively identify undistinguishable malicious node.