Paper 2021/1395

Homomorphic Secret Sharing for Multipartite and General Adversary Structures Supporting Parallel Evaluation of Low-degree Polynomials

Reo Eriguchi and Koji Nuida

Abstract

Homomorphic secret sharing (HSS) for a function $f$ allows input parties to distribute shares for their private inputs and then locally compute output shares from which the value of $f$ is recovered. HSS can be directly used to obtain a two-round multiparty computation (MPC) protocol for possibly non-threshold adversary structures whose communication complexity is independent of the size of $f$. In this paper, we propose two constructions of HSS schemes supporting parallel evaluation of a single low-degree polynomial and tolerating multipartite and general adversary structures. Our multipartite scheme tolerates a wider class of adversary structures than the previous multipartite one in the particular case of a single evaluation and has exponentially smaller share size than the general construction. While restricting the range of tolerable adversary structures (but still applicable to non-threshold ones), our schemes perform $$ parallel evaluations with communication complexity approximately $$ times smaller than simply using $$ independent instances. We also formalize two classes of adversary structures taking into account real-world situations to which the previous threshold schemes are inapplicable. Our schemes then perform $$ parallel evaluations with almost the same communication cost as a single evaluation, where $$ is the number of parties.