Distributing trusted third parties

Abstract

Trusted Third Parties (TTPs) are widely employed in various scenarios for providing fairness guarantees (e.g., in fair exchange or e-commerce protocols, including secure two-party computation), for distributing secrets (e.g., in authentication or secret-sharing protocols, as well as group signatures), and for creating trust (e.g., as certificate authorities). Such wide use of TTPs, as well as the trust requirement that is put on them make them a prime target for distributed systems and cryptography research. There are some well-known and proven solutions to the problem of distributing the trust put on TTPs: byzantine agreement or secure multi-party computation techniques can be employed to distribute the job of any TTP to multiple parties, tolerating up to half or one-third of those parties being malicious. Such techniques are not widely-employed in practice possibly due to their quadratic complexity or inter-operation requirements. This brings up the question of distributing TTPs in a much more efficient way, possibly via using autonomous agents, who do not directly communicate with each other. In this paper, we present various known techniques for distributing the trust put on TTPs. Then, we concentrate on using multiple autonomous parties, who do not communicate with each other at all, to realize a single TTP. We discuss the role of synchrony in such attempts, and conclude with some open questions.