Fair Coin Flipping: Tighter Analysis and the Many-Party Case

TL;DR

This paper introduces a new multi-party coin-flipping protocol with significantly reduced bias, improving fairness guarantees over previous protocols, especially as the number of parties increases, and provides a novel analysis framework for fairness.

Contribution

It presents a new t-party, m-round coin-flipping protocol with improved bias bounds and a novel paradigm for analyzing fairness in multi-party settings.

Findings

Achieves bias of O(t^4 * 2^t * sqrt(log m) / m^{1/2+1/(2^{t-1}-2)}) for t ≤ 1/2 log log m.

Improves bias bounds for three-party protocols to O(√log m / m).

Generalizes previous protocols by including a recovery mechanism for aborting or cheating parties.

Abstract

In a multi-party fair coin-flipping protocol, the parties output a common (close to) unbiased bit, even when some adversarial parties try to bias the output. In this work we focus on the case of an arbitrary number of corrupted parties. Cleve [STOC 1986] has shown that in any such $m$-round coin-flipping protocol, the corrupted parties can bias the honest parties' common output bit by $\Theta(1/m)$. For more than two decades, the best known coin-flipping protocol was the one of Awerbuch et al. [Manuscript 1985], who presented a $t$-party, $m$-round protocol with bias $\Theta(t/\sqrt{m})$. This was changed by the breakthrough result of Moran et al. [TCC 2009], who constructed an $m$-round, two-party coin-flipping protocol with optimal bias $\Theta(1/m)$. Haitner and Tsfadia [STOC 2014] constructed an $m$-round, three-party coin-flipping protocol with bias $O(\log^3m / m)$. Still for the case of more than three parties, the best known protocol remained the $\Theta(t/\sqrt{m})$-bias protocol of Awerbuch et al. We make a step towards eliminating the above gap, presenting a $t$-party, $m$-round coin-flipping protocol, with bias $O(\frac{t^4 \cdot 2^t \cdot \sqrt{\log m}}{m^{1/2+1/\left(2^{t-1}-2\right)}})$ for any $t\le \tfrac12 \log\log m$. This improves upon the $\Theta(t/\sqrt{m})$-bias protocol of Awerbuch et al., and in particular, for $t\in O(1)$ it is an $1/m^{\frac12 + \Theta(1)}$-bias protocol. For the three-party case, it is an $O(\sqrt{\log m}/m)$-bias protocol, improving over the $O(\log^3m / m)$-bias protocol of Haitner and Tsfadia. Our protocol generalizes that of Haitner and Tsfadia, by presenting an appropriate recovery protocol for the remaining parties to interact in, in the case that some parties abort or are caught cheating. We prove the fairness of the new protocol by presenting a new paradigm for analyzing fairness of coin-flipping protocols.