节点文献

对等网络文件共享系统中的信誉机制研究

Research on Reputation Mechanism in Peer-to-Peer File Sharing System

【作者】 韦冬

【导师】 杨寿保;

【作者基本信息】 中国科学技术大学 , 计算机系统结构, 2009, 博士

【摘要】 互联网的迅速发展给传统的网络应用模式带来了新的挑战,促成了对等网络P2P系统的产生,使得互联网系统的计算模式从客户端/服务器模式向对等模式转变。目前,P2P技术已广泛应用于很多领域,其中文件共享系统可以说始终是P2P技术发展的动力之源。但是,因为动态性和异构性的特点,P2P文件共享系统仍然存在着诸如共享文件的可靠性、节点间的信任等亟待解决的问题。P2P网络的信誉机制通过对系统中的用户或资源进行定性或定量的评价来预测该用户或资源未来的行为,从而起到鼓励良性行为、抑制恶意行为、辅助用户决策的重要作用。因此,对P2P文件共享系统中的信誉机制展开研究,设计和实现具有抵抗各种攻击行为能力的信誉系统,对提高P2P网络的可用性,具有十分重要的现实意义。本文在深入研究和总结相关领域已有成果的基础上,分别针对P2P网络文件共享系统的对象信誉层、节点信誉层、拓扑结构层三部分内容展开研究,设计实现了一个完整的P2P网络上的信誉系统。本文的主要研究工作如下:首先,分析了P2P用户的一般操作流程和节点动作对文件对象扩散的影响,建立了基于文件对象的污染传播模型,描述了系统中的文件传播过程和节点状态转换,揭示了文件共享系统中对象信誉机制对抑制污染的重要作用。在此模型的基础上,针对基于节点信誉机制的不足,建立了基于对象信誉的P2P文件共享防污染系统。利用查询消息来触发节点传播和收集投票,通过计算节点间投票向量的夹角余弦来衡量节点投票的相似性,以此赋予投票权重,运用自适应的信誉阈值来判断文件的可信性。为了实现节点间信任关系的传播和信任信息的有效利用,针对现有的基于最大流算法经典方案的局限,提出了基于基尔霍夫定律的信任传播方案。该方案基于电阻网络理论构建电阻信任网络图,更加精确的刻画了信任传播现象,更为自然的反映了信任关系的实质,同时还对算法时间复杂度进行了优化。最后,从宏观上考虑P2P网络的拓扑结构与信誉机制的相互影响,受到自然免疫原理的启发,采用人工免疫系统方法,研究了P2P网络中基于信誉的邻居选择方法和拓扑构造方案,优化了网络中的节点连接,在覆盖层网络层次上抑制了污染文件的传播。通过仿真实验和实际系统验证,该信誉系统以较低开销有效抑制了污染文件在网络中的传播,并且能够惩罚搭便车行为,应对各种攻击情形。

【Abstract】 The rapid development of Intemet has brought new challenge to the traditional network model,resulting in the emergence of P2P system.It makes the transfer of Internet computing model,from client/server model to peer-to-peer.At present,P2P technology has been widely utilized in many fields,especially the P2P file sharing system,as the power source of the P2P technology development.However,due to the nature of dynamics and heterogeneity,there are many issues worthing to research in P2P file sharing system,such as the reliability of sharing files and the trustness between nodes.The reputation mechanism of P2P network predicts future behavior of users or resources by making qualitative or quantitative evaluation of them.It can encourage benign behavior and inhibit malicious behavior,to play an important role in system. Therefore,it is of a very important and practical significance to carry out the research on the reputation mechanism in P2P file sharing system,and design and implement the attack proof reputation system,to improve the availability of P2P network.After investigating related works deeply,this dissertation focuses on three main parts of P2P file sharing system respectively,the object reputation,the node reputation and the overlay topology.A complete reputation system on the P2P file sharing network is designed and implemented.The main structure of this dissertation is as follows:At first,we analyse the general operation flow of users and the impact of the user action on the document propagation,and establish object-based model of pollution proliferation.The model describes the process of document propagation and state transition diagram of node,and revealed the important role of reputation mechanism in file sharing system.Based on the propagation model above,we establish an object reputation based P2P file sharing anti-pollution system,instead of node reputation.We utilize the query message to trigger the dissemination and collection of votes,and measure the similarity between nodes by calculating the cosine between voting vectors,then weight the votes.Self-adaptive threshold is used to determine the credibility of documents. In order to achieve effective utilization of reputation information,we propose a trust propagation algorithm based on Kirchhoffs law,instead of the classic maximum flow based algorithm.The resistance reputation network is established based on the resistor network theory.It is more precise and more natural to reflect the relationship and transitivity of trust.We also optimize the time complexity of algorithm.Finally,we consider the interactivity between network topology and reputation system in the macroscopical view.Inspired by the natural immune principles,we employ the artificial immune algorithm based reputation system to achieve the neighbor selection,connection optimization,and topology construction.The spread of pollution is inhibited at the level of overlay network.The simulation experiment and practical system show that our reputation system can control the pollution prolification effectively at a low cost.And it can punish the free-riding behavior,and cope with various attack situations.

节点文献中: