




已阅读5页,还剩24页未读, 继续免费阅读
版权说明:本文档由用户提供并上传,收益归属内容提供方,若内容存在侵权,请进行举报或认领
文档简介
Stacey Greenaway,Managing Reputation and Trust in Peer-to-Peer Networks.,CP4022 Research Topics in Networks and Distributed Systems. Assessment 2Stacey Greenaway 0487622,IntroductionDefinitionsResearch 1 - XRepResearch 2 Bayesian NetworkResearch 3 - Trust Vectors Research 4 EigenTrustProblemsConclusions,Introduction,Managing Reputation and Trust in Peer-to-Peer Networks. Stacey Greenaway,What is a peerto-peer network?A decentralized networkAll nodes in the network act as both clients and serversPowered by the bandwidth of all peersAd hoc connectionsTypes of P2P Network:Filesharing e.g. Gnutella, Kazaa, BitTorrentInstant Messaging, EcommerceExisitng Trust and Reputation systems:Ebay Feedback System,Introduction,Managing Reputation and Trust in Peer-to-Peer Networks. Stacey Greenaway,Trust - A peers trust in other peers based on his own past experience.Reputation - A peers trust in another peer based on the experiences of other peers.File Provider a peer providing a file for downloadServent a peer who is both client and server.Free Rider - A peer who only downloads and does not share any files. Inauthentic files viruses, corrupt, unreadable, wrong file type, content not what it claimed to be.,Definitions,Managing Reputation and Trust in Peer-to-Peer Networks. Stacey Greenaway,These threats provide the requirements for a trust and reputation system.Decoy files - A malicious peer will respond to any query with a copy of the requested file, but will deliver a file that has been tampered with or contains a virus at the point of download. Malicious peer - A peer who either belongs to one of the groups below or will provide an inauthentic file for every request. Malicious collective - A group of malicious peers who know each other and collaborate to subvert a P2P system.Self Replication virus such as Gnutella vbs.worm poses as a peer and then creates a copy of itself for download.Pseudospoofing - malicious peers control multiple identities, false pseudonyms are used to give good reputation to other pseudonyms controlled by the same malicious peer.,Definitions Attacks and Threats,Managing Reputation and Trust in Peer-to-Peer Networks. Stacey Greenaway,A basic explanation of a Trust and Reputation system:Peers store opinions on their experiences at downloading files.They store an opinion about the file provider and the file.These opinions are computed either into binary or using another mathematical probability.Peers share their opinions providing recommendations for file providers and files.A peers opinion can be weighted based on how much the querying peer trusts them.The aim of the system is to eliminate malicious peers and inauthentic files.,A Trust and Reputation System,Managing Reputation and Trust in Peer-to-Peer Networks. Stacey Greenaway,A Reputation Based Approach for Choosing Reliable Resources inPeer-to-Peer Networks (Damiani et al.) 1Cited by the other 3 research papers. 2 3 4Propose a protocol called XRep:A peer, p, queries the network for other peers opinions (votes) on resources and servents.Resource repository - records an ID for each file downloaded and whether it is good(+) or bad(-) Servent repository - stores the number of successful and unsuccessful downloads by each peer. Votes are converted to binary, where a positive (+) = 1 and negative (-) = 0.,Research 1 - XRep,Managing Reputation and Trust in Peer-to-Peer Networks. Stacey Greenaway,A Reputation Based Approach for Choosing Reliable Resources in Peer-to-Peer Networks (Damiani et al.) 1XRep has six phases: Resource Searching and Resource Selection:query network, retrieving list of files, select one to download based on trust and reputation. Vote Polling:peer (p) asks the other peers opinions (poll request) about the resource (r) it is about to download or on the servent (s) offering the resource. Poll responses encrypted using a public key called “pkpoll” - contains the responding peers vote, IP Address and Port. Vote Evaluation:“pkpoll” decrypted. p clusters the votes, which allows it to detect those sharing the same IP address. (pseudospoofing)An average value of all votes in the cluster is calculated and returned to the querying peer (p). A random selection of “voters” from each cluster is contacted for confirmation of their vote using the IP and Port encypted in “pkpoll”. Best Servent Check:Choose the most reputable servent to download file from. Resource Downloading:After download, p updates his repositories with his opinion of both the servent and resource.,Research 1 - XRep,Managing Reputation and Trust in Peer-to-Peer Networks. Stacey Greenaway,Trust and Reputation Model in Peer-to-Peer Networks (Wang et al.) 2 Propose a Trust and Reputation Model using Bayesian Networks to build a profile of each peers opinions based on different contexts of trust.An Analogy:Mike has two friends John who is a mechanic and Bob who is a Doctor. Miketrusts Bob with a medical complaint but not to fix his car and respectively, trusts John to fix his car but not to diagnose a medical condition. So in the context of fixing a car John is trustworthy, but Bob is untrustworthy. What one peer may consider a good file is not what another peer would considergood. For instance peer As priority in a good file is its content regardless of its quality.,Research 2 - Bayesian Network,Managing Reputation and Trust in Peer-to-Peer Networks. Stacey Greenaway,Trust and Reputation Model in Peer-to-Peer Networks (Wang et al.) 2Bayesian network - A Bayesian Network is a graph consistingof nodes and arcs. Nodes represent variables and the arcs represent the relationships and dependencies between thevariables. (5)Fig 1 - basic Bayesian network consisting only of one parent and several child nodes. FP - file providerT - % of all positive interactions. Trust is dependant on Download Speed (DS), FileQuality (FQ) and File Type (FT). More user preferences can be added to the Bayesian Network, e.g. copyright.Trust in a condition only calculated once.,Research 2 - Bayesian Network,Managing Reputation and Trust in Peer-to-Peer Networks. Stacey Greenaway,Fig 1 Nave Bayesian Network,Trust and Reputation Model in Peer-to-Peer Networks (Wang et al.) 2The Trust and Reputation Model:A peers reliability is measured as a file provider and a referee.Queries are issued about the reliability of a file and its provider.A Bayesian Network contains a peers opinions on all past interactionsBayesian Networks are exchanged and compared to form groups of trusted peers. (The similarity of nodes is calculated, nodes with higher value (between 0 and 1) indicate peer preferences) The recommendations of trusted peers are weighted more heavily than unknown peers as they share similar preferences.After every interaction the BNs of the file provider and the referees will be updated to reflect the peers trust in them.Combat pseudospoofing or malicious collective attacks. (false recommendation values will be obvious when compared to trusted peers and counteracted),Research 2 - Bayesian Network,Managing Reputation and Trust in Peer-to-Peer Networks. Stacey Greenaway,Trust and Reputation Model in Peer-to-Peer Networks (Wang et al.) 2Experiments:Simulated a file sharing system based on Gnutella.Each node is either a file provider or peer at one time.Each peer only knows its direct neighbour and a few fileproviders.1000 interactions between 40 peers and 10 file providers.Each of ten runs is evaluated by taking averages of the results.4 systems compared:Trust and Reputation with BNTrust and Reputation without BNTrust with BN Trust without BNresults: Systems where opinions are exchanged perform better.Using a BN only gives marginal increase in performance.,Research 2 - Bayesian Network,Managing Reputation and Trust in Peer-to-Peer Networks. Stacey Greenaway,A Reputation-Based Trust Management System for P2P Networks (Selcuk et al.) 3 Propose a protocol to control the amount of inauthentic files a malicious peer can disperse through the P2P network.“Trust Vectors” are kept locally by peersConsult own “trust vector”, or request a “trust rating” from other peersUses query messages to gain recommendations from other peers“Trust Vectors” are updated after every download with + or opinion.,Research 3 - Trust Vectors,Managing Reputation and Trust in Peer-to-Peer Networks. Stacey Greenaway,A Reputation-Based Trust Management System for P2P Networks (Selcuk et al.) 3 Trust Vectorsare binary consist of 8, 16, or 32 bitslength is stored as an integer variablepositive (1) or negative (0) opinion is represented in the vector as 1 bitupdates recorded at the vectors most significant bit Trust Rating Calculated by dividing the sum of the Trust Vector by the power of 2, then dividing the result by 2 to the power of the number of significant bits in the vector.Distrust RatingHas more weight than a Trust RatingMalicious action hard to recover from,Research 3 - Trust Vectors,Managing Reputation and Trust in Peer-to-Peer Networks. Stacey Greenaway,Fig 2 Trust Vector,Fig 3 Trust Rating,A Reputation-Based Trust Management System for P2P Networks (Selcuk et al.) 3The trust ratings of file providers are evaluated:an average of the trust values of the most trusted peers in the list is calculateddetermined by a set threshold value.If No. Trusted Peers Threshold a random selection of peers are queriedtrust and distrust ratings will be included in the responses.credibility rating gives weight to the opinions.credibility vector - peers opinion truthful (1) untruthful (0)Threshold sets number of responses to evaluate.,Research 3 - Trust Vectors,Managing Reputation and Trust in Peer- to- Peer Networks. Stacey Greenaway,Fig 3 Trust Evaluation,A Reputation-Based Trust Management System for P2P Networks (Selcuk et al.) 3Experiments:Test performance under various malicious attacks:navehypocriticalmalicious collectivepseudospoofingSimulation Spec:1000 peers and 1000 files between 1% and 10% maliciouspeer linked to 3 neighbours, query submitted over these links for 3 hops, specified by TTL,Research 3 - Trust Vectors,Managing Reputation and Trust in Peer- to- Peer Networks. Stacey Greenaway,Research 3 - Trust Vectors,Managing Reputation and Trust in Peer- to- Peer Networks. Stacey Greenaway,F1 - represents the ratio of malicious to all downloads.,Fig 4 Results,A Reputation-Based Trust Management System for P2P Networks (Selcuk et al.) 3,The EigenTrust Algorithm for Reputation Management in P2P Networks (Kamvar et al.) 4 “to decrease the number of inauthentic files in a P2P file sharing system that assigns each peer a unique global trust value based on the peers history of uploads”. 5 Eigenvectors A special set of vectors associated with Linear Algebra, and matrixes, where left eigenvector is a row of the matrix and right eigenvector is a column of the matrix. (6 7)Malicious peers identified rather than the inauthentic files Decoy FilesSelf Replicating WormsMalicious CollectivesMalicious SpiesPseudospoofing,Research 4 - EigenTrust,Managing Reputation and Trust in Peer-to-Peer Networks. Stacey Greenaway,The EigenTrust Algorithm for Reputation Management in P2P Networks (Kamvar et al.) 4 Basic Trust Model “Peer i is more likely to trust the opinions of peers from whom he has had an honest interaction with in the past. ”. 5 Local Trust Value - calculated from peer is experiences downloading from other peers, j. Global Trust Value - calculated from the local trust values assigned to peer i by peers j. based on their experiences downloading from i.Each peer computes its own Global Trust Value and stores it locally.Local Trust Values are normalized.Peers share Trust Values.Trust Values are weighted by the amount of trust peer i places in the referee.Peer i asks the friends of friends for their recommendations until a view of entire network is achieved.Eventually Global Trust Value will represent the trust the entire system holds in peer j.,Research 4 - EigenTrust,Managing Reputation and Trust in Peer-to-Peer Networks. Stacey Greenaway,The EigenTrust Algorithm for Reputation Management in P2P Networks (Kamvar et al.) 4 Basic Trust Model Pre Trusted Peers, PPeers who established the system or first usersTheir distribution across the network =Their trust values are used to break up malicious collectives or when a referee is inactive.,Research 4 - EigenTrust,Managing Reputation and Trust in Peer-to-Peer Networks. Stacey Greenaway,Fig 5 Basic EigenTrust Algorithm,The EigenTrust Algorithm for Reputation Management in P2P Networks (Kamvar et al.) 4 Secure Trust Model Score Managers peers who compute the trust value of another peer in the systemAssigned using Distributed Hash Table (DHT) Set of Daughter Peers, DiDistributed Hash Tablehash functions map a Unique ID for each peer (IP Address and TCP port) into points in a logical coordinate space.coordinate space is partitioned over the network, every peer covers a region of that dynamic space.the peer who covers the region where that ID is hashed becomes that peers score manager.,Research 4 - EigenTrust,Managing Reputation and Trust in Peer-to-Peer Networks. Stacey Greenaway,Fig 6 Can Hash Space(cited by Kamvar et al. (13) 4),The EigenTrust Algorithm for Reputation Management in P2P Networks (Kamvar et al.) 4 Secure Trust Model Score Managers - computing Global Trustcomputes the Global Trust Values of its Daughtersholds an opinion vectorqueried to find the trust worthiness of d, whereThe score manager learns:set of peers who download from d and their opinion of dset of peers who d has downloaded from and its opinion of those interactions.,Research 4 - EigenTrust,Managing Reputation and Trust in Peer-to-Peer Networks. Stacey Greenaway,The EigenTrust Algorithm for Reputation Management in P2P Networks (Kamvar et al.) 4 Secure EigenTrust Algorithm,Research 4 - EigenTrust,Managing Reputation and Trust in Peer-to-Peer Networks. Stacey Greenaway,Fig 7 Secure EigenTrust Algorithm,The EigenTrust Algorithm for Reputation Management in P2P Networks (Kamvar et al.) 4 Experiments,Research 4 - EigenTrust,Managing Reputation and Trust in Peer-to-Peer Networks. Stacey Greenaway,Fig 8 Simulation Settings,The EigenTrust Algorithm for Reputation Management in P2P Networks (Kamvar et al.) 4 ExperimentsThreat Model Aindividual malicious peersThreat Model Bmalicious collectivesAmount of malicious peers is increased by 10% - max. 70%. Results:Inauthentic files make up approximately 10% of the network compared to a maximum of over 90% in a network without the proposed trust model. Using values of Pre Trusted Peers has broken up the malicious collective.,Research 4 - EigenTrust,Managing Reputation and Trust in Peer-to-Peer Networks. Stacey Greenaway,Fig 9 Threat Model A,Fig 10 Threat Model B,The EigenTrust Algorithm for Reputation Management in P2P Networks (Kamvar et al.) 4 ExperimentsThreat Model Cmalicious collectives with camouflage malicious peers who will offer authentic files some of the time in order to gain higher trust values, to increase their chance at being selected as a download source. Amount of authentic files uploaded is increased by 10% - max. 90%. Results:The more authentic files a malicious peer provides, the more impact they have at providing inauthentic files.Too costly in terms of bandwidth and disk space.,Research 4 - EigenTrust,Managing Reputation and Trust in Peer-to-Peer Networks. Stacey Greenaway,Fig 11 Threat Model C,Performance BottleneckMain problem facing Trust and Reputation systemsPeer with highest reputation always selected as download sourceNo suitable solution proposedRandom selectionPriority queue,Managing Reputation and Trust in Peer-to-Peer Networks. Stacey Greenaway,Fig 12 Performance Bottleneck 4,Problems,None of the proposed systems have been implemented in real worldHard to measure trust and reputation when measuring human opinionNo standards that define w
温馨提示
- 1. 本站所有资源如无特殊说明,都需要本地电脑安装OFFICE2007和PDF阅读器。图纸软件为CAD,CAXA,PROE,UG,SolidWorks等.压缩文件请下载最新的WinRAR软件解压。
- 2. 本站的文档不包含任何第三方提供的附件图纸等,如果需要附件,请联系上传者。文件的所有权益归上传用户所有。
- 3. 本站RAR压缩包中若带图纸,网页内容里面会有图纸预览,若没有图纸预览就没有图纸。
- 4. 未经权益所有人同意不得将文件中的内容挪作商业或盈利用途。
- 5. 人人文库网仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对用户上传分享的文档内容本身不做任何修改或编辑,并不能对任何下载内容负责。
- 6. 下载文件中如有侵权或不适当内容,请与我们联系,我们立即纠正。
- 7. 本站不保证下载资源的准确性、安全性和完整性, 同时也不承担用户因使用这些下载资源对自己和他人造成任何形式的伤害或损失。
最新文档
- 宁夏大学新华学院《即兴口语传播一》2023-2024学年第一学期期末试卷
- 武汉生物工程学院《田径运动会编排操作》2023-2024学年第一学期期末试卷
- 西京学院《全科医疗中的医患关系与沟通技巧》2023-2024学年第一学期期末试卷
- 青岛远洋船员职业学院《漫画墨线绘制》2023-2024学年第一学期期末试卷
- 2024-2025学年江苏省扬大附中东部分学校九年级化学第一学期期末监测模拟试题含解析
- 公路货运行业数字化转型2025年与物流政策法规研究报告
- 公路货运行业数字化转型与绿色物流发展报告
- 雁形板在节能建筑中的应用及安装技巧
- 美容整形手术室的感染控制体系
- 长春理工大学《应用统计学B》2023-2024学年第一学期期末试卷
- DeepSeek在银行业务场景的应用
- 后期入股合同协议
- 【信得科技】2025猪腹泻病防控手册
- 江西省吉安市十校联盟学校2024-2025学年七年级下学期4月期中考试数学试卷(无答案)
- 2024年山东夏季高中学业水平合格考地理试卷真题(含答案)
- 广西《沃柑质量分级》编制说明
- 医疗器械从业培训
- 2025年保密观考试题库及答案
- 工地安保合同协议
- DB11 T 640-2009 旅游咨询服务中心设置与服务规范
- 大模型备案-落实算法安全主体责任基本情况
评论
0/150
提交评论