Peer-to-Peer Reputations Prashant Dewan Advisor: Dr. Partha Dasgupta 1 January 2019 1/1/2019 Peer-to-Peer Reputations
Peer-to-Peer Reputations Goals To facilitate ‘secure’ transactions among peers in a completely decentralized system Secure implies Peers can be authenticated The chance that a peer gets cheated in a transaction is low A peer is either forced or motivated against cheating 1/1/2019 Peer-to-Peer Reputations
Peer-to-Peer Reputations Possible Techniques Micropayments Expensive Waste resources A peer with large volume of resources can break it Other Distributed Security Systems SDSI, Dynamic Trust Management, Groove,Publius All of them use a central component All of them are dependent on some form of ‘Authority’ 1/1/2019 Peer-to-Peer Reputations
Peer-to-Peer Reputations What is a reputation? The general estimation in which a person is held by the public – Oxford Dictionary A Recommendation is the ‘certificate of satisfaction’ given by one entity (requestor) to the other (provider) after a transaction ( eg. A File Download) The Reputation of a person ( or an entity) is the sum of the ‘recommendations’ received by the person (or an entity) from others (people or entities) with whom it has transacted in the past. 1/1/2019 Peer-to-Peer Reputations
Calculation of Reputation Recommendations B A = 5 C A = 3 D A = 4 Alice’s Reputation 5+3+4 = 12 1/1/2019 Peer-to-Peer Reputations
Peer-to-Peer Reputations Challenges 1/1/2019 Peer-to-Peer Reputations
Peer-to-Peer Reputations Benefits 1/1/2019 Peer-to-Peer Reputations
Peer-to-Peer Reputations Self-Certification Certificate Authority User does not have to trust a third party certificate provider which might be a vulnerability User can separate transactions by having multiple identities 1/1/2019 Peer-to-Peer Reputations
Peer-to-Peer Reputations Liar Farms 5 10 Certificate Authority 2 5 4 8 4 2 4 1/1/2019 Peer-to-Peer Reputations
Peer-to-Peer Reputations IP Based Safeguard d Rep 1 29 2 12.5 3 1/1/2019 Peer-to-Peer Reputations
Peer-to-Peer Reputations IP Based Safeguard Most peers are interested in relative ranks of other peers Adjustment of ranks has a ‘small’ effect on the relative ranks Network Size=1000 peers Mean Rank Difference= 13.246±0.81 (95% confidence) 1/1/2019 Peer-to-Peer Reputations
Elicitation-Storage Protocol The “requestor” selects the highest-reputation provider from a list of possible providers The requestor verifies the recommendations of the provider Checks if the signature on the recommendations match the identity certificate sent along with the recommendation Once verified, the “provider” provides the content The requestors issues the recommendation to the provider The provider stores the recommendation and the identity of the requestor 1/1/2019 Peer-to-Peer Reputations
Peer-to-Peer Reputations Application Areas eBay, Amazon, Monster… School Friends, Family… CORE, CONFIDANT & SPRITE… SMTP, Web Services… Gnutella, Kazaa & Morpheus… 1/1/2019 Peer-to-Peer Reputations
Peer-to-Peer Reputations P2P & NAT Port Session Initiation Protocol Needs Servers TCP Splicing Very complicated UDP Solutions Waste of Bandwidth UPnP – A standard being developed by Microsoft and allies. 1/1/2019 Peer-to-Peer Reputations
Peer-to-Peer Reputations Name Service Server: Domain1: superpeer1 Domain2: superpeer2 Bob ASU: David is Superpeer 1/1/2019 Peer-to-Peer Reputations
Peer-to-Peer Reputations Name Service .. Alice Logs in in New York Connects to the Domain of NYU Informs ASU super peer of current location Bob contacts ASU superpeer and gets the address of Alice 1/1/2019 Peer-to-Peer Reputations
Peer-to-Peer Reputations Name Service .. David Logs out Asks all domain members for the next superpeer Carol becomes the superpeer David Transfers current locations to Carol and Logs out Carol Informs the public server Bob tries to reach David and Fails and gets to Carol via public server 1/1/2019 Peer-to-Peer Reputations