In this paper, RunTrust, a trust model based on Run-length coding algorithm, performs trust evaluation by compressing records of peers’ behaviors which contain more information including time dimension, so it can exhibit good performance in trust evaluation and detecting malicious especially oscillatory behaviors. Moreover, the capability of filtering false feedbacks is also improved. The simulating results indicated that RunTrust has significantly raised the efficiency of trust management system only at a very low cost. The work on RunTrust has paved for the research on trust data compacting.