【作者】 冯艳；

【导师】 曾锃；

【作者基本信息】 湖南大学 ， 计算机应用技术， 2009， 硕士

【摘要】 如今,P2P技术已经广泛应用于网络的各个方面,其信誉问题成为了研究的重点。传统的网络安全机制如数字签名和数字水印技术虽然能保证交易的保密性和完整性,但无法对交易双方的信用度进行评价,因此基于P2P的信任机制能有效的加强系统的可靠性。本文在分析现有的信任模型优缺点的基础上,提出一种SNTrust的超节点的P2P网络信任模型。模型中采取的是组的概念,每个组的节点分为超级节点和普通节点,超级节点负责连接本组中所有普通节点,存放组内每个节点的全局信誉度,并在超级节点之间组成P2P网络。每个普通节点加入P2P网络都要通过相应的超级节点,在本地保存直接交易数据。在信任度的计算方面,该模型改善了以往信任模型中计算信任度的方法,把信任度的范围扩展到(-1,1)。信任度的计算以组为单位,分为两个层次:组内的本地信任关系和以组为单位的全局信任关系。组内部节点的信任关系用直接信任度、间接信任度和总体信任度来计算。其总体信任度由直接信任和间接信任组成,直接信任度由消费节点通过与提供服务的服务节点的历史成功/失败交互记录获取,并且对于失败交易加入了惩罚因子来进行调整;间接信任度的值通过推荐信任与总体信任的综合来获取,而推荐信任中不引入惩罚因子,直接参考相互交易的失败和成功的比例。以组为单位的全局信任关系:取参与交易的两个节点自身的全局信任度和两节点所在的组直接交易信任度的最小值来判断。该实验通过对不同交易量下的交易成功率对比,充分说明了本信任模型能够有效的抑制恶意节点的行为,改善网络性能。同时,通过与其它信任模型进行性能分析,更进一步证明了该模型的可行性和有效性。更多还原

【Abstract】 Today, P2P technology has been widely applied in all aspects of networks, the individual credibility has become a focus of trust model study. Traditional network security mechanisms, such as digital signatures and digital watermarking, can ensure the confidentiality and integrity of the transaction, but they cannot evaluate the trust of both transaction sides. Hence, the trust mechanism based on P2P can enhance system reliability.Through the analysis on the advantages and disadvantages of the existing trust model, we propose a SNTrust P2P network trust model. Using the concept of group, Nodes of each group are separated into super-nodes and normal nodes. Super-nodes are responsible for connecting all the normal nodes in this group, storing the whole trust of each node in the group. P2P network is formed among super-nodes. Normal node must pass through the corresponding super-nodes to join the P2P network, and the direct transaction data are stored locally in order to provide effect protection for the P2P network transactions.This model improves the previous methods of trust calculation in trust models and extends the scope of trust to (-1,1). In terms of cluster, trust calculation falls into two levels: local trust in the cluster and whole trust based on cluster.The node trust in the cluster is calculated by direct trust, indirect trust (which constitute final trust) and whole trust. Direct trust is attained from the ratio of historical success / failure records on the interactions between consumption node and service node,the failed deal is adjusted with the penalty factor; Indirect trust is obtained through integrating recommendation trust and whole trust. Recommendation trust, not affected by penalty factor, refers directly to the rates of transaction failure and success.Whole trust based on cluster is judged by the least value of whole credit worthiness of the two nodes involved in the transaction and direct transaction credit worthiness of the cluster in which the two nodes are included.The experiment, comparing the ratio of successful transaction in various volumes, demonstrates that this trust model is effective in restraining the behavior of malicious nodes and improving network performance. At the same time, the performance analysis on several other trust models further proves the feasibility and effectiveness of this model.更多还原