A New Method of Trust Inference Based on Markov Model for Peer-to-Peer Network

Trust is applied to different aspects in a variety of systems. In computing systems, it is not well understood from computational perspective. Knowing how much a person trusts the others could be helpful in obtaining the trusted group in peer-to-peer network and trusted computing is a good way to measure the trust. Trust inference is essential in decision making. Conditional probability is used to quantize the trust relationship. This paper proposes a new algorithm based on the improvements of Markov model and adopts the level factor and confidence to compute the indirect trust inference. Through replicating cross-nodes, the trust network could consist of several independent paths without duplicate nodes. This algorithm describes two facts that with the increase of indirect relationship, the trust value should be lower, and each person has his own confidence in trust probability. In our experiment on the dataset Epinion, compared with the well-known trust inference algorithms Average and Multiplication, our method computes a more accurate and objective trust inference value, demonstrating its effectiveness.