Paper 2022/1207

Attaining GOD Beyond Honest Majority With Friends and Foes

Abstract

In the classical notion of multiparty computation (MPC), an honest party learning private inputs of others, either as a part of protocol specification or due to a malicious party's unspecified messages, is not considered a potential breach. Several works in the literature exploit this seemingly minor loophole to achieve the strongest security of guaranteed output delivery via a trusted third party, which nullifies the purpose of MPC. Alon et al. (CRYPTO 2020) presented the notion of Friends and Foes ($$) security, which accounts for such undesired leakage towards honest parties by modelling them as semi-honest (friends) who do not collude with malicious parties (foes). With real-world applications in mind, it's more realistic to assume parties are semi-honest rather than completely honest, hence it is imperative to design efficient protocols conforming to the $$ security model. Our contributions are not only motivated by the practical viewpoint, but also consider the theoretical aspects of $$ security. We prove the necessity of semi-honest oblivious transfer for $$-secure protocols with optimal resiliency. On the practical side, we present QuadSquad, a ring-based 4PC protocol, which achieves fairness and GOD in the $$ model, with an optimal corruption of $$ malicious and $$ semi-honest party. QuadSquad is, to the best of our knowledge, the first practically efficient $$ secure protocol with optimal resiliency. Its performance is comparable to the state-of-the-art dishonest majority protocols while improving the security guarantee from abort to fairness and GOD. Further, QuadSquad elevates the security by tackling a stronger adversarial model over the state-of-the-art honest-majority protocols, while offering a comparable performance for the input-dependent computation. We corroborate these claims by benchmarking the performance of QuadSquad. We also consider the application of liquidity matching that deals with highly sensitive financial transaction data, where $$ security is apt. We design a range of $$ secure building blocks to securely realize liquidity matching as well as other popular applications such as privacy-preserving machine learning (PPML). Inclusion of these blocks makes QuadSquad a comprehensive framework.