P2p Network Security Trust Model

In peer to peer network entities can join and quit dynamically, the user management isn't central, closed and controllable with absence of a generally accepted trusted center' Traditional security mechanisms such as authentication and authorization usually need identity information of entities in advance. However the anonymity and dynamical properties of the node in P2P network, Traditional security mechanisms lose effectiveness. Trust mechanism introduce the trust concept in society, evaluation of the node's behavior can reflect its reliability. This dissertation studies how to build trust model in P2P network in order to separate malicious nodes out of the network environment.In this dissertation a Bayesian -based trust model is presented in P2P network, history interaction information of nodes takes as sample information in this trust model: Combining prior information of nodes makes inference the node further behavior. According to the difference of the node's trust relationship, distinguish direct trust and recommended trust, and respectively give two kinds of calculation method. The confidence of direct trust takes as the basis whether need to evaluate recommended trust. In the recommended trust computation, the different weight is assigned to nodes according to the level of direct trust. In this dissertation also studies the unfair recommend in recommended trust. An unfair recommend filtering algorithm is presented. By calculating recommend trust of the node whether to fall within the confidence interval, determine whether to recommend information is false. And according to the level of the trust recommended standard deviation give an order of filtering malicious nodes. Finally in this dissertation has confirmed the model validity through virtual scene and has made the comparison with Liu M's model. Then two kind of unfair recommendfiltering algorithm validity has made the analysis.