1 / 22

Detecting Reputation Variations in P2P Networks

Detecting Reputation Variations in P2P Networks. Theodora Dariotaki & Alex Delis Deprt. of Informatics & Telecommunications The University of Athens (th.dariotaki, ad)@di.uoa.gr. Basic Questions. How do reputation schemes work?. C. A. B. Why we might want to detect reputation variations?.

tyler
Télécharger la présentation

Detecting Reputation Variations in P2P Networks

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Detecting Reputation Variations in P2P Networks Theodora Dariotaki & Alex Delis Deprt. of Informatics & Telecommunications The University of Athens (th.dariotaki, ad)@di.uoa.gr The University of Athens

  2. Basic Questions • How do reputation schemes work? C A B • Why we might want to detect reputation variations? The University of Athens

  3. Reputation Monitoring Mechanism (RMM) • Monitors the reputation variations of offerers • Limits abrupt changes of reputation values • New concepts • RVM peers (Reputation Variation Monitor) • Epoch • Storage Structures: • DPE: DirectPeerExperience Table • DRE: DirectResourceExperience Table • RT: Reputation Table (RVM Only) The University of Athens

  4. Phase I - Resource Request Q: Where is the resource located? • Requester q dispatches an AskResource message asking for resource s • Offerers reply with a HoldResource message - Dispatch: The message is forwarded in a scope of h hops from q or until answered by a resource holder - Cycles: Messages received more than once, are discarded The University of Athens

  5. Phase II - Recommendation Request Q: Are the offerers trustworthy? • Requester q dispatches an AskRecom message for all offerers’ reputation and resource requested. • First-Line (FL) recommenders respond with PostRecom messages. - Prevention of blacklisting The University of Athens

  6. Phase III - Evaluation of Offerer/Resource Reputation Q: Are FL-recommenders reputable? • If Direct Peer Experience > θrecommendation accepted • If qhas never communicated with a FL-recommender, qdispatches an AskRecom message for FL’s reputation.Second-Line (SL) recommenders respond with PostRecom Q: Are SL-recommenders reputable? • Only if qhas direct experience with the SL-recommender and Direct Peer Experience > θthe recommendation is accepted q may rely - on its own opinion - on the recommenders’ opinion - on both The University of Athens

  7. Phase IV – Offerer SelectionBaseline Reputation Scheme (BRS) A peer is candidate for resource downloading if both: reputation level of the resource holder reputation level of the hosted resource exceed a threshold θ. • Candidate peers are sorted in a list • Random selection of one-of-top reputable peers to prevent overloading of most reputable peers • Challenge-response handshake between requester and resource offerer to ensure resource possession • Download initiation The University of Athens

  8. Phase IV – Offerer Selection (1/4)RMM Scheme • Step a: Early Offerer Selection • Offers with Peer/Resource reputation level < θ are discarded • Step b: Reputation Variation Request • Requester qdispatches anonymous AskRVMmessages • RVMs respond with RVMReplymessages - reputation levels of offerers during last λ epochs The University of Athens

  9. y x Phase IV – Offerer Selection (2/4) Step c: Evaluation of Offerer Reputation Variation • Relative Reputation Variation (V) is computed for all offerersas the fraction x/y x: difference between a previous and the last observed reputation level (0.67-0.84=-0.17) y: difference between the perfect reputation and the lowest of the two reputation levels (1.00-0.67=0.33) The University of Athens

  10. Step d: Reputation Update RRMM=0.75 RRMM=0.79 y y' x' x Dependence on 2 epochs General case: λ epochs Phase IV – Offerer Selection (3/4) • Requester q re-evaluates offerers’ reputation Dependence on 1 epoch The University of Athens

  11. Step e: Final Offerer Selection Phase IV – Offerer Selection (4/4) • Candidate peers are sorted in a list • Random selection of one-of-top reputable peers • Challenge-response handshake between requester and resource offerer • Download initiation The University of Athens

  12. Phase V – Resource Download & Experience Updates • The requester qasks for resource s from the selected offerer pw by sending a DownloadReq message • pwsends the resource • qrecords its satisfaction in both DPE & DREtables -DPE: satisfaction concerning selected offerer pw FL-recommenders for pw SL-recommenders for every FL-recommender of pw -DRE: satisfaction concerning downloaded resource The University of Athens

  13. Forwarding continues until max hops h are exceeded or the resource is found Resource holders reply with HoldResource Messages received twice are discarded Peers forward the query Requester broadcasts an AskResource (h=3) Example 13 4 12 found 3 5 10 11 1 2 7 6 AskResource HoldResource 8 9 found h = 3 1: requester 8&10: resource holders The University of Athens

  14. 11 replies for 8 with PostRecom 6 & 13reply for 10 11 & 13are unknown to requester 6 has been proven trustworthy Requester broadcasts AskRecom for both 8 & 10 and resource s Example found 13 4 12 3 5 10 11 1 2 found 7 6 found AskRecom PostRecom 8 9 1: requester 8&10: resource holders The University of Athens

  15. 12 replies for 11 with PostRecom 9replies for 13(but 9 is unknown to 1) Requester dispatches an AskRecom for 11 & 13 Assume that 12 claims that 11 is trustworthy. Then 11’s recomme-ndation for 8 is accepted. Example 13 4 found 12 3 5 10 11 1 2 7 6 AskRecom PostRecom 8 9 found 1: requester 8&10: resource holders The University of Athens

  16. Requester considers 8 to be more reputable than 10 and downloads the resource from 8 (BRS) Example 13 4 12 3 5 10 11 1 2 7 6 DownloadReq 8 9 1: requester 8&10: resource holders The University of Athens

  17. RVMs respond with RVMReply sending the reputation values for 8&10 during previous λ epochs Requester sends anonymous AskRVM asking for reputation variations on 8&10 Requester computes the Relative Variation Values for 8&10 and detects abrupt changes in 8’s reputation Example(RMM) 13 4 12 3 5 10 11 1 2 7 6 AskRVM RVMReply 14 8 9 15 16 RVM peers 1: requester 8&10: resource holders The University of Athens

  18. Requester downloads the resource from 10 Example 13 4 12 3 5 10 11 1 2 7 6 DownloadReq 8 9 1: requester 8&10: resource holders The University of Athens

  19. BRS vs. RMM λ = 3 The University of Athens

  20. Discussion (1/3) • Pseudospoofing & Shilling Attacks • Smooth out abrupt changes • Challenge-response handshake • Bind with real-world identities • Man-in-the-middle • Message Authentication/Integrity Check • RVM • anonymity • impersonation • failure The University of Athens

  21. Discussion (2/3) • Number and Duration of Epochs • Average frequency fx of download requests in popular peers • Network population N • Space Overhead of RVM • λxNp(Np:average # of peers assigned to a RVM) The University of Athens

  22. Discussion (3/3) • Communication Cost • (k: average # of neighboring nodes, h: max # of hops) • More efficient solutions: • Select kr most reputable neighbors • Use a P2P routing protocol (e.g. Chord with O(logN) messages, N: network population) • Anonymity cost for RVMs (e.g. Tarzan with O(N) messages) The University of Athens

More Related