P2p Network Trust Management Research

With the rapid development of P2P technology, network applications based on P2P technology is becoming a hot research topic. However, the autonomy, distributed, dynamic and large-scale nature of P2P network makes it more difficult to keep safe and stable than traditional network. Trust management, which imitates the trust mechanism in human society, is able to establish trust relationships between peers, and solve some trust issues. Therefore, the research on trust management of P2P network is very meaningful.In this thesis, cloud model is introduced into traditional reputation-based trust model, to express the uncertainty of trust completely. Cloud model proposed by academician Deyi Li, is a theory dealing with the convertion between qualitative and quantitative models. Cloud model mainly deals with two kinds of uncertainty:fuzziness and randomness, and can combine them together. When compute the trust of peers, this thesis considers both the average trust of peers and the stability of peers' behaviour, then a decision-making method based on the change rate of trust value is proposed.To decrease the computation and communication overhead in the computation of the peers'global trust in large-scale P2P network, taking the idea of trust in human society, we propose a trust model based on interest group. The peers with semantic similarity are clustered, in which a large network is divided into small interest groups. The peers in the same group are equal, autonomous and have frequent transactions with each other. To calculate the trust value, computations were firstly done within each group, then the results were aggregated to gain the global trust value. Simulation results show that our model can effectively reduce the computation and communication overhead, and can effectively guarantee the transaction success ratio and the resistance against attacks.Trust value in existing trust models is usually partial or one-sided, and the computation and communication overhead of global trust are relatively large. The proposed trust model in this thesis can express the uncertainty of trust completely, as well as effectively reduce the computation and communication overhead. Moreover, it enhances the security and stability of P2P networks, and has important reference value to promote P2P applications.