Paper 2023/1159

Semi-Honest 2-Party Faithful Truncation from Two-Bit Extraction

Abstract

As a fundamental operation in fixed-point arithmetic, truncation can bring the product of two fixed-point integers back to the fixed-point representation. In large-scale applications like privacy-preserving machine learning, it is essential to have faithful truncation that accurately eliminates both big and small errors. In this work, we improve and extend the results of the oblivious transfer based faithful truncation protocols initialized by Cryptflow2 (Rathee et al., CCS 2020). Specifically, we propose a new notion of two-bit extraction that is tailored for faithful truncation and demonstrate how it can be used to construct an efficient faithful truncation protocol. Benefiting from our efficient construction for two-bit extraction, our faithful truncation protocol reduces the communication complexity of Cryptflow2 from growing linearly with the fixed-point precision to logarithmic complexity. This efficiency improvement is due to the fact that we reuse the intermediate results of eliminating the big error to further eliminate the small error. Our reuse strategy is effective, as it shows that while eliminating the big error, it is possible to further eliminate the small error at a minimal cost, e.g., as low as communicating only an additional 160 bits in one round.