Paper 2020/1577

Multi-Party Replicated Secret Sharing over a Ring with Applications to Privacy-Preserving Machine Learning

Alessandro Baccarini and Marina Blanton and Chen Yuan

Abstract

Secure multi-party computation has seen significant performance advances and increasing use in recent years. Techniques based on secret sharing offer attractive performance and are a popular choice for privacy-preserving machine learning applications. Traditional techniques operate over a field, while designing equivalent techniques for a ring can boost performance. In this work we develop a suit of multi-party techniques for a ring in the honest majority setting starting from elementary operations to more complex with the goal of supporting general-purpose computation. We demonstrate through empirical evaluation that our techniques can be several times faster than their field-based equivalents and up to two orders of magnitudes faster for certain operations such as matrix multiplication. We also evaluate our techniques on machine learning applications and show that the resulting performance is on par with that of most recent custom protocols for these applications.