P2P网络中信任管理研究

发布时间：2018-10-16 12:45

【摘要】：P2P网络(peer-to-peer)的出现,有效的解决了传统网络模式的单点失效、性能瓶颈等问题,将人们更直接的通过网络联系起来,人们可以通过对等网络方便的交互信息和资源,自己可以存储自己相关的数据,而不是集中依赖于少数的几台中央服务器。对等网络(peer-to-peer)的开放性、可扩展性、匿名性等特点,使得P2P网络已经在协同工作、大规模并行计算、文件共享、分布式计算、远程协作、搜索引擎、即时通信、移动设备等多种领域得到广泛应用,然而也是这些特点使得对等网络的信任性和安全性没有保障,存在很多欺诈性的恶意节点,另外由于缺乏或者没有完善的激励机制,有四分之一的节点是自由搭便车节点,它们只享受其他节点提供的服务,而从不向其他节点提供服务。这些对于P2P网络的安全和发展都存在着威胁。在P2P网络中建立信任管理模型可以有效的解决系统中的信任和安全问题,并且可以很好的解决系统中的推荐问题。信任管理模型主要分为基于声誉的信任管理模型、基于推荐的信任管理模型和基于PKI的信任管理模型,基于PKI的信任模型中节点都是经过CA认证的可信节点,这使得系统的可扩展性不好和存在单点失效的问题,因此目前基于声誉和基于推荐的信任模型是研究热点。本论文主要围绕P2P网络中的信任管理和激励机制进行了研究,主要成果如下： (1)研究了国内外学者已经提出的各种基于不同角度的信任机制,提出了单次好评度的概念,基于信任的主观性提出用模糊综合评判的方法来计算单次好评度。 (2)从资源评价的角度出发提出了一种新的基于资源评价的信任管理模型。每次交易后的交易记录表由提供资源的节点的母节点进行管理和存储；请求节点不但综合考虑对目标节点的直接信任度和此次交易资源的总好评度,而且还要考虑风险值来选择目标节点。考虑了资源评价的计算方法可以有效的抑制恶意节点的恶意行为。 (3)针对搭便车现象、不可靠服务、不真实反馈三种现象,提出了一个基于信任管理的动态的激励机制。首先提出了上传因子、下载因子等相关概念,用单次好评度来计算贡献值,然后提出了一系列的包括奖惩的规则,最后用博弈论证明了节点更倾向于遵守规则和上传资源。仿真结果表明,提出的激励机制可以促进节点适当上传资源和有效防止恶意行为。
[Abstract]:The emergence of P2P network (peer-to-peer) effectively solves the problems of single point failure, performance bottleneck and so on of traditional network mode. It connects people more directly through the network, and people can exchange information and resources conveniently through peer-to-peer network. You can store your own relevant data instead of concentrating on a few central servers. Peer-to-Peer Network (peer-to-peer) has the characteristics of openness, scalability, anonymity and so on, which makes P2P network already working in collaboration, large-scale parallel computing, file sharing, distributed computing, remote cooperation, search engine, instant messaging, etc. Mobile devices and other fields have been widely used. However, these characteristics make the trust and security of peer-to-peer network not guaranteed, there are many malicious nodes, and because of the lack of or lack of perfect incentive mechanism, 1/4 of the nodes are free-rider nodes. They only enjoy the services provided by other nodes and never provide services to other nodes. These are threats to the security and development of P2P networks. The establishment of trust management model in P2P network can effectively solve the trust and security problems in the system, and can solve the recommendation problem in the system. Trust management model is mainly divided into reputation based trust management model, based on recommended trust management model and trust management model based on PKI, the trust model based on PKI is a trusted node authenticated by CA. This makes the scalability of the system poor and the problem of single point failure, so the trust model based on reputation and recommendation is a hot research topic at present. This thesis mainly focuses on the trust management and incentive mechanism in P2P network. The main results are as follows: (1) the trust mechanism based on different angles has been proposed by domestic and foreign scholars. In this paper, the concept of single praise degree is proposed. Based on the subjectivity of trust, the method of fuzzy comprehensive evaluation is proposed to calculate the single high praise degree. (2) A new trust management model based on resource evaluation is proposed from the point of view of resource evaluation. The transaction record table after each transaction is managed and stored by the parent node of the node providing the resource; the requesting node not only considers the direct trust degree of the target node and the total praise of the transaction resource, And also consider the risk value to select the target node. The calculation method of resource evaluation can effectively suppress the malicious behavior of malicious nodes. (3) aiming at the phenomenon of free rider, unreliable service and untrue feedback, A dynamic incentive mechanism based on trust management is proposed. The related concepts such as upload factor, download factor and so on are put forward first, then a series of rules including rewards and punishments are put forward. Finally, the game theory is used to prove that nodes are more inclined to obey the rules and upload resources. Simulation results show that the proposed incentive mechanism can promote the nodes to upload resources properly and effectively prevent malicious behavior.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP393.08

【参考文献】