A Robust and Efficient Reputation System for Active Peer-to-Peer Systems Dominik Grolimund, Luzius Meisser, Stefan Schmid, Roger Wattenhofer Computer Engineering.

Slides:



Advertisements
Similar presentations
Ion Stoica, Robert Morris, David Karger, M. Frans Kaashoek, Hari Balakrishnan MIT and Berkeley presented by Daniel Figueiredo Chord: A Scalable Peer-to-peer.
Advertisements

Cope with selfish and malicious nodes
Evaluation of a Scalable P2P Lookup Protocol for Internet Applications
Peer-to-Peer Systems Chapter 25. What is Peer-to-Peer (P2P)? Napster? Gnutella? Most people think of P2P as music sharing.
Cloudifying Source Code Repositories: How much does it cost? LADIS 2009 Big Sky, Montana Michael Siegenthaler Hakim Weatherspoon Cornell University.
A Folder Tree Structure for Cryptographic File Systems Dominik Grolimund, Luzius Meisser, Stefan Schmid, Roger Wattenhofer Computer Engineering and Networks.
Incentives Build Robustness in BitTorrent Bram Cohen.
Clayton Sullivan PEER-TO-PEER NETWORKS. INTRODUCTION What is a Peer-To-Peer Network A Peer Application Overlay Network Network Architecture and System.
© 2005 Andreas Haeberlen, Rice University 1 Glacier: Highly durable, decentralized storage despite massive correlated failures Andreas Haeberlen Alan Mislove.
On the Economics of P2P Systems Speaker Coby Fernandess.
A Blueprint for Constructing Peer-to-Peer Systems Robust to Dynamic Worst-Case Joins and Leaves Fabian Kuhn, Microsoft Research, Silicon Valley Stefan.
Fabian Kuhn, Microsoft Research, Silicon Valley
Incentives-Compatible Peer-to-Peer Multicast Tsuen-Wan “Johnny” Ngan with Dan Wallach and Peter Druschel Rice University.
Denial-of-Service Resilience in Peer-to-Peer Systems D. Dumitriu, E. Knightly, A. Kuzmanovic, I. Stoica and W. Zwaenepoel Presenter: Yan Gao.
Modelling and Performance Analysis of BitTorrent-Like Peer-to-Peer Networks.
An Overview of Peer-to-Peer Networking CPSC 441 (with thanks to Sami Rollins, UCSB)
Peer-to-Peer Networks as a Distribution and Publishing Model Jorn De Boever (june 14, 2007)
CSCE 715 Ankur Jain 11/16/2010. Introduction Design Goals Framework SDT Protocol Achievements of Goals Overhead of SDT Conclusion.
On the Topologies Formed by Selfish Peers Thomas Moscibroda Stefan Schmid Roger Wattenhofer IPTPS 2006 Santa Barbara, California, USA.
Network Coding for Large Scale Content Distribution Christos Gkantsidis Georgia Institute of Technology Pablo Rodriguez Microsoft Research IEEE INFOCOM.
An Authentication Service Based on Trust and Clustering in Wireless Ad Hoc Networks: Description and Security Evaluation Edith C.H. Ngai and Michael R.
Taming Dynamic and Selfish Peers “Peer-to-Peer Systems and Applications” Dagstuhl Seminar March 26th-29th, 2006 Stefan Schmid Distributed Computing Group.
FRIENDS: File Retrieval In a dEcentralized Network Distribution System Steven Huang, Kevin Li Computer Science and Engineering University of California,
Spotlighting Decentralized P2P File Sharing Archie Kuo and Ethan Le Department of Computer Science San Jose State University.
Service Differentiated Peer Selection An Incentive Mechanism for Peer-to-Peer Media Streaming Ahsan Habib, Member, IEEE, and John Chuang, Member, IEEE.
Dept. of Computer Science & Engineering, CUHK1 Trust- and Clustering-Based Authentication Services in Mobile Ad Hoc Networks Edith Ngai and Michael R.
Peer-to-Peer Based Multimedia Distribution Service Zhe Xiang, Qian Zhang, Wenwu Zhu, Zhensheng Zhang IEEE Transactions on Multimedia, Vol. 6, No. 2, April.
1 Denial-of-Service Resilience in P2P File Sharing Systems Dan Dumitriu (EPFL) Ed Knightly (Rice) Aleksandar Kuzmanovic (Northwestern) Ion Stoica (Berkeley)
Exploiting Content Localities for Efficient Search in P2P Systems Lei Guo 1 Song Jiang 2 Li Xiao 3 and Xiaodong Zhang 1 1 College of William and Mary,
Aleksandar Kuzmanovic & Edward W. Knightly A Performance vs. Trust Perspective in the Design of End-Point Congestion Control Protocols.
An Authentication Service Against Dishonest Users in Mobile Ad Hoc Networks Edith Ngai, Michael R. Lyu, and Roland T. Chin IEEE Aerospace Conference, Big.
Chord-over-Chord Overlay Sudhindra Rao Ph.D Qualifier Exam Department of ECECS.
Freenet A Distributed Anonymous Information Storage and Retrieval System I Clarke O Sandberg I Clarke O Sandberg B WileyT W Hong.
1 CS 194: Distributed Systems Distributed Hash Tables Scott Shenker and Ion Stoica Computer Science Division Department of Electrical Engineering and Computer.
Wide-area cooperative storage with CFS
Improving Data Access in P2P Systems Karl Aberer and Magdalena Punceva Swiss Federal Institute of Technology Manfred Hauswirth and Roman Schmidt Technical.
SocialFilter: Introducing Social Trust to Collaborative Spam Mitigation Michael Sirivianos Telefonica Research Telefonica Research Joint work with Kyungbaek.
Introduction Widespread unstructured P2P network
A Game Theoretic Framework for Incentives in P2P Systems --- CS. Uni. California Jun Cai Advisor: Jens Graupmann.
1 Speaker : 童耀民 MA1G Authors: Ze Li Dept. of Electr. & Comput. Eng., Clemson Univ., Clemson, SC, USA Haiying Shen ; Hailang Wang ; Guoxin.
Free-riding and incentives in P2P systems name:Michel Meulpolder date:September 8, 2008 event:Tutorial IEEE P2P 2008.
Michael Sirivianos Xiaowei Yang Stanislaw Jarecki Presented by Vidya Nalan Chakravarthy.
Yitzchak Rosenthal P2P Mechanism Design: Incentives in Peer-to-Peer Systems Paper By: Moshe Babaioff, John Chuang and Michal Feldman.
Peer to Peer Research survey TingYang Chang. Intro. Of P2P Computers of the system was known as peers which sharing data files with each other. Build.
A P2P file distribution system ——BitTorrent Pegasus Team CMPE 208.
2: Application Layer1 Chapter 2: Application layer r 2.1 Principles of network applications r 2.2 Web and HTTP r 2.3 FTP r 2.4 Electronic Mail  SMTP,
Do incentives build robustness in BitTorrent? Michael Piatek, Tomas Isdal, Thomas Anderson, Arvind Krishnamurthy, Arun Venkataramani.
Andreas Larsson, Philippas Tsigas SIROCCO Self-stabilizing (k,r)-Clustering in Clock Rate-limited Systems.
The EigenTrust Algorithm for Reputation Management in P2P Networks
1 Distributed Hash Tables (DHTs) Lars Jørgen Lillehovde Jo Grimstad Bang Distributed Hash Tables (DHTs)
Arun Venkataramani Donald Towsley Presented by: Shiqi Chen, Ionut Trestian.
Super-peer Network. Motivation: Search in P2P Centralised (Napster) Flooding (Gnutella)  Essentially a breadth-first search using TTLs Distributed Hash.
Adapted from the original presentation made by the authors Reputation-based Framework for High Integrity Sensor Networks.
1 Maze A Hybrid P2P file sharing system Design by Networking and distributed System lab at Peking University Presenter:Elaine.
FastTrack Network & Applications (KaZaA & Morpheus)
A Passive Approach to Sensor Network Localization Rahul Biswas and Sebastian Thrun International Conference on Intelligent Robots and Systems 2004 Presented.
Distributed Architectures for Medical Systems Andrew A. Kitchen Computer Integrated Surgery 8 March 2001.
Efficient P2P Search by Exploiting Localities in Peer Community and Individual Peers A DISC’04 paper Lei Guo 1 Song Jiang 2 Li Xiao 3 and Xiaodong Zhang.
High-integrity Sensor Networks Mani Srivastava UCLA.
A P2P-Based Architecture for Secure Software Delivery Using Volunteer Assistance Purvi Shah, Jehan-François Pâris, Jeffrey Morgan and John Schettino IEEE.
Distributed Quota Enforcement for Spam Control Jee Whan Choi Chaoting Xuan.
ADVANCED COMPUTER NETWORKS Peer-Peer (P2P) Networks 1.
Peer to Peer Network Design Discovery and Routing algorithms
Algorithms and Techniques in Structured Scalable Peer-to-Peer Networks
Peer-to-Peer Systems: An Overview Hongyu Li. Outline  Introduction  Characteristics of P2P  Algorithms  P2P Applications  Conclusion.
FairTorrent: BrinGing Fairness to Peer-to-Peer Systems
Determining the Peer Resource Contributions in a P2P Contract
Simplified Explanation of “Do incentives build robustness in BitTorrent?” By James Hoover.
Binghui Wang, Le Zhang, Neil Zhenqiang Gong
Presentation transcript:

A Robust and Efficient Reputation System for Active Peer-to-Peer Systems Dominik Grolimund, Luzius Meisser, Stefan Schmid, Roger Wattenhofer Computer Engineering and Networks Laboratory (TIK), ETH Zurich NetEcon’06 June 10, Ann Arbor, Michigan, USA Havelaar Distributed Computing Group

2 / 26 Talk Outline  Environment  Existing Solutions  Principles of Havelaar  Evaluation  Conclusions

3 / 26 Do You Know YouTube?  Very popular online video platform  > 30 mio. users, growing rapidly  >> 1 mio. watched every day  >> 10,000 uploaded every day  very active!

4 / 26 Guess What: YouTube is Centralized  Hosted on servers  Simple, but: huge costs  1 mio. $ / month for bandwidth and storage  Low quality  Limited (10-minute clips)

5 / 26 Imagine YouTube Being Decentralized  Files stored in a distributed storage system  Resources provided by the users  Uncontrollable environment:  unreliable, ordinary desktop computers  private users  turn computer on and off at any time  can leave the system forever at any time  open, attracts malicious agents, attacks  rational agents, free-riders

6 / 26 Three Key Problems 1. Availability How can the data be made immediately accessible when requested, although users can turn off their computer at any time? 2. Reliability How can the data be stored persistently, despite the inherent dynamics, node departures, and malicious nodes? 3. Incentives (focus of this talk) How can rational agents be encouraged to provide their resources without free-riding?

7 / 26 Kangoo – A Distributed Storage System  Research at ETH Zurich  Availability achieved with redundancy:  A file is divided into ~100 blocks, which are then encrypted and encoded into ~500 redundant fragments using erasure codes  Any 100 are sufficient to reconstruct the file  Lots of transactions necessary!  Usage of YouTube would result in tens of thousands of transactions per peer and week  Not ready yet, but you can subscribe for the beta:

8 / 26 This Talk: Havelaar  How to encourage peers to provide their upload bandwidth? (storage and online time are handled by Kangoo itself)  Havelaar is independent of Kangoo  can be used for other systems as well.  Robust to attacks  Efficient, scalable in the number of transactions

9 / 26 Talk Outline  Environment  Existing Solutions  Principles of Havelaar  Evaluation  Conclusions

10 / 26 Existing Solutions  Direct reciprocity (e.g. BitTorrent)  Tit-for-tat, iterated prisoner dilemma  Works for content distribution, but not for a system where interactions are too infrequent  Monetary-based (e.g. Karma)  Economic theory  But: centralized or else inefficient, market regulations,...  Reputation systems (e.g. eBay)  Service differentiation: The higher your reputation, the better your service  Good, but how...

11 / 26 Reputation Systems How to keep track of the contribution of each peer?  Client (e.g. Kazaa)  Simple to subvert, as it has been shown with Kazaa Lite  Centralized (e.g. eBay)  Many many more transactions if used for fairness in a p2p system  server cluster would be needed  Decentralized  Good, but how...

12 / 26 Decentralized Reputation Systems  Direct observations do not scale to large networks with infrequent interactions  We need to incorporate second-hand observations  Big new problem: false reports

13 / 26 Coping with False Reports How to defend against false reports?  Max-flow  Maximum likelihood estimation  Bayesian approach  Transitivity of trust  weigh the voting by the reputation of the sender  Most systems are designed for a decentralized „pure reputation system“ (e.g. eBay), but not meant for a fairness system where we need to track the contribution of each peer with lots of transactons

14 / 26 Storing Contribution Values Where to store the contribution value of each peer?  Flood in the system (e.g. EigenRep)  Request from peers before transaction  Store in a DHT: „DHT-based approach“  store and update contribution value of peer u at h(u) in a DHT  Scales linearly in the number of transactions

15 / 26 Talk Outline  Environment  Existing Solutions  Principles of Havelaar  Evaluation  Conclusions

16 / 26 Introducing Havelaar  Approximation is good enough!  If peer u provides three times more than v, u should get about three times a better download bandwidth than v  Track contribution value C: bandwidth b, size s  If locally computed contribution value is close to the global / real one for all peers, that‘s fine

17 / 26 Local Vector  Every peer has a local observation vector o  After u downloads from v (bandwidth of 5, size 3), u will increase the entry of v by 5 * 3 (C v += 15)  Only after complete transaction

18 / 26 Send Local Vector To Successors h 1 (w) h 2 (w) h 3 (w) h 4 (w) observation vector o o o o o w once a round (~ week) k successors: determined by hash functions on the sender id w same successors in every round can only send to its k successors  limited influence can only send once per round „self-observation“ of the sender is dropped  cannot praise itself defend against attacks:

19 / 26 Aggregation: Need More Observations  Need more observations for an accurate approximation  Aggregate exponentially more: o0o0 use all for contribution update c O O O = [o0,o1+o1+o1,o2+o2+o2] [o1,o2,o3] own observations o 3 dropped defend against attacks: for each entry, outliers are detected and dropped praise or accusation „within bounds“ will be smoothed out (lots of observations aggregated) distribution of a vector can be analyzed  if spiked, then it is most likely an attack  drop, maybe even decrease the trust value of that peer

20 / 26 Rewarding  Always allocate full bandwidth  No artificial limits  Contention: Two or more want to download from a third node at the same time  allocate according to the contribution values  Different resource allocation algorithms possible. We chose an algorithm similar to: An Incentive Mechanism for P2P Networks, R. B. Ma et al., ICDS 2004

21 / 26 Talk Outline  Environment  Existing Solutions  Principles of Havelaar  Evaluation  Conclusions

22 / 26 Evaluation bootstrapping  We have analyzed and simulated Havelaar  5 successors and a matrix with four vectors is already enough for huge networks with more than 100,000 nodes and 5,000 transactions per peer and round.

23 / 26 Communication Costs  Need to send a huge matrix, but: it does not depend on the number of transactions!  The more transactions, the higher the accuracy!

24 / 26 Talk Outline  Environment  Existing Solutions  Principles of Havelaar  Evaluation  Conclusions

25 / 26 Conclusions  Havelaar for active, long-term peer-to-peer systems  Robust against attacks, false reports  Low communication costs: scalable in the number of transactions  Churn: not an issue because the local vector can be sent at any time in a round  Kangoo takes care about other attacks (sybil attacks, white washing) and has strong identifiers

26 / 26 Thank you for your attention!