Expensive techniques like byzantine agreement or secure multi-party computation with Omega(n^2) communication can be applied to distribute arbiters in a non-autonomous way. Yet we are interested in efficient protocols that can be achieved by keeping the arbiters autonomous (non-communicating), especially for p2p settings in which the arbiters do not even know each other. Avoine and Vaudenay employ multiple autonomous arbiters in their optimistic fair exchange protocol which uses global timeout mechanisms; all arbiters have access to -loosely- synchronized clocks. They left two open questions regarding the use of distributed autonomous arbiters: (1) Can an optimistic fair exchange protocol without timeouts provide fairness (since it is hard to achieve synchronization in a p2p setting) when employing multiple autonomous arbiters? (2) Can any other optimistic fair exchange protocol with timeouts achieve better bounds on the number of honest arbiters required? In this paper, we answer both questions negatively. To answer these questions, we define a general class of optimistic fair exchange protocols with multiple arbiters, called "distributed arbiter fair exchange" (DAFE) protocols. Informally, in a DAFE protocol, if a participant fails to send a correctly formed message, the other party must contact some subset of the arbiters and get correctly formed responses from them. The arbiters do not communicate with each other, but only to Alice and Bob. We prove that no DAFE protocol can meaningfully exist.
Category / Keywords: cryptographic protocols / optimistic fair exchange, distributed cryptography, distributed arbiters, trusted third party. Publication Info: full version of PODC 2009 brief announcement Date: received 10 Feb 2009, last revised 29 Mar 2013 Contact author: kupcu at cs brown edu Available format(s): PDF | BibTeX Citation Note: updated full version Version: 20130329:095128 (All versions of this report) Short URL: ia.cr/2009/069 Discussion forum: Show discussion | Start new discussion