基于节点行为的信任管理机制应用研究

【作者】 赵爽；

【导师】 李凤银；

【作者基本信息】 曲阜师范大学 ， 工程硕士（专业学位）， 2016， 硕士

【摘要】 P2P(Peer-to-Peer)网络中节点地位是同等的,没有服务者和被服务者之分,节点在被其他节点服务的同时也在向其他节点提供服务。在目前互联网的重心从以网络为核心转变为以移动终端为核心的背景下,P2P技术凭借其部署灵活和价格低廉等特点得到广泛应用。当前P2P网络中的信任管理模型也存在着很多的问题,例如:没有在计算信任值时考虑到信任会随着时间衰减,没有考虑节点更相信与自己看法更相似的节点的推荐,以及网络规模的扩大引起的模型性能降低、计算开销大的问题,因此,针对上述问题展开了如下工作:(1)对于P2P网络中信任管理模型未考虑信任会随着时间衰减以及节点更愿意相信看法与自己更相似的节点推荐的问题,提出了基于节点评分行为的信任管理模型(NRBTrust)。由于信任随着时间的衰减特性,模型引入信任的时间衰减函数在计算信任值时作为调节因子,由于节点更倾向于相信与自己看法相似的节点,所以在计算推荐信任度时把节点评价相似度考虑在内。(2)对于P2P网络中信任管理模型中节点规模的扩大导致系统计算开销增大的问题,提出了基于兴趣簇的信任管理模型(IC-Trust)。该模型通过节点间相似度的比较,当达到特定阈值时则可以加入这一兴趣簇内。同一兴趣簇内节点的交易相对于兴趣簇间节点的交易更加频繁。一个兴趣簇内有且只有一个超级节点,超级节点是在簇内选取的综合能力最高者担任,它负责兴趣簇内节点的信任信息的管理、节点交易信息的管理和节点的加入和退出操作。如果交易双方在同一兴趣簇内,通过整合兴趣簇内其他节点对目标节点的历史交易记录和交易节点自己与目标节点之间交易的经验,再结合相关的参数得出节点的信任值。如果交易双方不在同一兴趣簇内,在节点直接交易的基础上再融合兴趣簇间信任度得到节点的信任度。(3)设计并实现了一个信任管理原型系统来验证提出的两个信任管理模型,从系统的运行来看,本文所提出的信任管理模型能够识别出恶意节点的各种恶意行为并且大大的提高了系统的成功交易率,提高了分布式环境下节点交互的安全性。更多还原

【Abstract】 In P2P network, all nodes have equal status, the node provide services for other nodes when enjoying services provided by them, therefore, the node not only can be a service provider but a consumer. Nowadays, the core of the Internet has changing from network into mobile terminal, and P2P technology has obtained extensive and successful applications on the strength of its flexible deployment and low-cost.Many problems exist in the existing trust models in P2P network, such as ignoring that trust value decay with time when computing trust value, ignoring to considering the nodes tend to trust their similarity node’s recommendation, and reduced performance and computational overhead with the growing network size. To solve the above problems, the main works of this paper are as follows:(1) As for the problems of ignoring that trust value decay with time when computing trust value and ignoring to considering the nodes tend to trust their similarity node’s recommendation, a trust model based on nodes rating behavior(NRBTrust) is proposed.NRBTrust takes the time decay function as regulatory factor when computing direct trust,because trust value decays with the passage of time. And NRBTrust also takes nodes evaluation similarity into consideration when computing recommender trust, because nodes tend to believe the nodes, which is more similar with them.(2) As for the problem of the continuous expansion of nodes in trust management model leading to computational cost, a trust management model based on interest clusters(IC-Trust)is proposed. IC-Trust adds nodes with higher similarity that achieves a similarity threshold into a cluster. The trade rate among nodes in a cluster is more frequent than the nodes in different clusters. Each interest cluster will select a node with highest comprehensive performance as the super node, which is responsible for managing trust information, trading information and the dynamic adding and exiting of the other nodes. If two trading nodes are in the same interest cluster, the trust of the nodes can be obtained by integrating the trading records among other nodes in the same cluster and the target node and the trading experience between the trading node and the target node, and several related parameters should be considered. If two trading nodes are in different clusters, the trust of the node can be obtained by integrating the trust rating among the interest clusters and the direct trading of the node.(3) A prototype system of trust management is designed to validate the above two models.The performance of the system shows that the proposed models can effectively restrain manykinds of malicious behaviors of some nodes, improve the successful trade rate,improve the security of node trading in distributed network environment.更多还原