Rational Proofs with Multiple Provers

TL;DR

This paper introduces multi-prover rational interactive proofs (MRIP), demonstrating they are more powerful than single-prover systems and characterizing their capabilities under various utility gap conditions.

Contribution

The paper defines MRIP protocols, analyzes their power relative to existing proof systems, and establishes tight characterizations for different utility gap classes.

Findings

MRIP is more powerful than RIP and MIP under standard assumptions.

Two provers and three rounds suffice to achieve full MRIP power.

Constant, noticeable, and negligible utility gaps are all effectively characterized.

Abstract

Interactive proofs (IP) model a world where a verifier delegates computation to an untrustworthy prover, verifying the prover's claims before accepting them. IP protocols have applications in areas such as verifiable computation outsourcing, computation delegation, cloud computing. In these applications, the verifier may pay the prover based on the quality of his work. Rational interactive proofs (RIP), introduced by Azar and Micali (2012), are an interactive-proof system with payments, in which the prover is rational rather than untrustworthy---he may lie, but only to increase his payment. Rational proofs leverage the provers' rationality to obtain simple and efficient protocols. Azar and Micali show that RIP=IP(=PSAPCE). They leave the question of whether multiple provers are more powerful than a single prover for rational and classical proofs as an open problem. In this paper, we introduce multi-prover rational interactive proofs (MRIP). Here, a verifier cross-checks the provers' answers with each other and pays them according to the messages exchanged. The provers are cooperative and maximize their total expected payment if and only if the verifier learns the correct answer to the problem. We further refine the model of MRIP to incorporate utility gap, which is the loss in payment suffered by provers who mislead the verifier to the wrong answer. We define the class of MRIP protocols with constant, noticeable and negligible utility gaps. We give tight characterization for all three MRIP classes. We show that under standard complexity-theoretic assumptions, MRIP is more powerful than both RIP and MIP ; and this is true even the utility gap is required to be constant. Furthermore the full power of each MRIP class can be achieved using only two provers and three rounds. (A preliminary version of this paper appeared at ITCS 2016. This is the full version that contains new results.)