Paper 2024/1347

Secure Multiparty Computation with Lazy Sharing

Abstract

Secure multiparty computation (MPC) protocols enable $n$ parties, each with private inputs, to compute a given function without leaking information beyond the outputs. One of the main approaches to designing efficient MPC protocols is to use secret sharing. In general, secret sharing based MPC contains three phases: input sharing, circuit evaluation, and output recovery. If the adversary corrupts at most $t$ parties, the protocol typically uses $(t,n)$ threshold secret sharing to share the inputs. In this work, we consider a weaker variant of threshold secret sharing called lazy threshold secret sharing (or simply lazy sharing) and show that - Lazy sharing can serve as a viable alternative to threshold secret sharing in MPC without compromising security. - Lazy sharing could be generated more efficiently than threshold secret sharing. As a result, replacing threshold secret sharing with lazy sharing can lead to a more efficient input sharing phase. Moreover, we propose that the efficiency of the circuit evaluation phase can also be further improved. To support this claim, we apply lazy sharing to several state-of-the-art MPC protocols and analyze the efficiency gain in various settings. These protocols include the GMW protocol (Goldreich et al., STOC 1987), the AFLNO protocol (Araki et al., CCS 2016), and the SPDZ protocol (Damg{\aa}rd et al., CRYPTO 2012). By doing so, we analyze the efficiency gains in various settings and highlight the advantages of incorporating lazy sharing into MPC protocols.