【作者】 陶世忠；

【导师】 史清华；

【作者基本信息】 山东大学 ， 计算机应用技术， 2007， 硕士

【摘要】 随着信息技术的迅猛发展,分布式计算架构也在经历着变革,Peer-to-Peer(以下简称P2P)就是其中一种很有前景的技术。P2P技术给我们带来的不仅是机遇,还有挑战,这是因为P2P网络中恶意节点的存在,使得网络中出现的各种攻击成为威胁网络正常运作的主要因素。因此,如何在P2P网络中识别出恶意节点,约束和杜绝恶意节点的恶意行为成为目前的一个热点。这其中,对节点的信誉值的研究越来越引起大家的关注。因为双方是否信任对方是双方进行交易的前提,信誉值可以用来表示这个节点以往的行为是否适合参与某个网络社区的生活。全文首先介绍了网络安全与密码学的相关知识,并分析了P2P技术的特点、优势以及当前的研究现状,并结合了Chord模型和P2P网络的优点,引用了一种双重的结构化P2P模型:Dual-Chord,该模型用于节点的资源定位和查找上。系统分为两层:主干网层和子网层。该拓扑在保留DHT拓扑的良好特征的同时,充分考虑节点可信度差异,即,相对于低可信度节点,高可信度节点占据更重要的位置,从而增强了拓扑的可靠性和稳定性。文章在上述基础上,提出并改进了一个基于推荐值的信任模型。在改进的模型中,充分考虑了给出推荐值的节点其自身的信任度,克服了以往模型中优先设置一些高可信任度节点的不安全做法。在文中,作者还考虑了将节点的信任度异地存储的问题,并引入公钥密码体系,有效防止了节点通过修改自身信任度来实施欺骗的恶意行为。对于节点的恶意行为,文章中进一步进行了细分,这是因为不同的恶意行为对于网络的影响是不一样的,比如发布恶意代码相对于下载速度慢而言带来的威胁更为严重,这样就使得对于信任值的计算就更为科学。通过分析,可以看到该模型对于冒名、协同作弊以及诋毁等行为有较好的抑制作用。实验结果证明,当网络中恶意节点上升到一定比例后,改进后的算法仍然能保持较高的下载成功率。更多还原

【Abstract】 With the development of information technology, the framework of distributed computing is under a new revolution, among which Peer-to-Peer (P2P) technology is seen as a promising one. As there are vicious peers that arouse many attacks preventing network from running normally, Peer-to-Peer technology brings us both opportunities and threats. How to identify, restrict and put end to those vicious peers is becoming a hotspot. And the research on peers’ reputation has brought more and more attention. Because trustworthiness of each other is the precondition of business, reputation can represent whether the peer’s foregone behavior is suitable to participate in the network community.This paper first introduces interrelated knowledge about network security and cryptology, and then analyzes the characteristic, advantage and research actuality of P2P technology. The author combines the advantages of Chord model and P2P network and puts forward a hierarchical P2P model: Dual-Chord. The Chord-model-based system is used for peers’ resource locating and lookup. It consists of main net and subnet. This topology both maintains good quality of DHT and considers the peers’ difference of trustworthiness, which means the more trusty peers in network can get more important topology position, and make more contribution to topology stability and reliability.As a foundation of the above, this paper puts forward a trust model based recommendation value and makes improvement on it. In the improved model, we consider sufficiently about the trustworthiness of a peer itself which affords recommendation value. This model conquers the insecurity of setting up some peers with high trustworthiness in advance. What’s more, we take separate storage of peer’s trustworthiness into account and use Public-Key-Infrastructure (PKI), which can prevent the peer from deceiving by reworking its trustworthiness. The text classifies in detail about peer’s vicious behavior, that’s because different vicious behavior affects differently to the network. For example, putting out vicious codes is more serious than downloading slowly. We can calculate more scientific trustworthiness in this way. Through analysis, we know this model can restrain the conduct of cheating and imitating. Experiment result proves that when the number of vicious peers has reached a definite proportion, the improved algorithm can keep a higher successful download.更多还原