Concrete Efficiency Improvements for Multiparty Garbling with an Honest Majority. (Latincrypt 2017)
Aner Ben-Efraim and Eran Omri.
Latincrypt 2017
Abstract
Secure multiparty computation is becoming a necessary component in many real-world systems. The efficiency of secure two-party protocols has improved tremendously in the last decade, making such protocols efficient enough for many real-world applications. Recently, much attention is being diverted to making secure {\em multiparty} computation (for more than two parties) truly practical as well. In particular, the last couple of years saw a resurgence of interest in constant round secure protocols, based on the multiparty garbling paradigm of Beaver et al.~(STOC 1990). Such protocols generally offer improved performance in high latency networks, such as the internet.
In this paper we consider the case where a majority of the parties are honest, and construct highly efficient constant round protocols for both the semi-honest setting and the malicious setting. Our protocols in the semi-honest setting significantly improve over the recent multiparty garbling protocols for honest majority of Ben Efraim et al. (ACM CCS 2016), both in asymptotic complexity and in concrete running time.
In the malicious setting, we consider security with abort when assuming more than $2/3$ of the parties are honest. We show that by assuming the existence of simple preprocessing primitives, which do not require knowledge of the computed function, we get malicious security at almost the same cost as semi-honest security. I.e., the function dependent preprocessing and the online phase are almost identical to the semi-honest setting.
We ran experiments to measure the effect of our optimizations and to show that our protocols compete with the state-of-the-art constant round protocols.