Paper 2024/1771

PRIME: Differentially Private Distributed Mean Estimation with Malicious Security

Abstract

Distributed mean estimation (DME) is a fundamental and important task as it serves as a subroutine in convex optimization, aggregate statistics, and, more generally, federated learning. The inputs for distributed mean estimation (DME) are provided by clients (such as mobile devices), and these inputs often contain sensitive information. Thus, protecting privacy and mitigating the influence of malicious adversaries are critical concerns in DME. A surge of recent works has focused on building multiparty computation (MPC) based protocols tailored for the task of secure aggregation. However, MPC fails to directly address these two issues: (i) the potential manipulation of input by adversaries, and (ii) the leakage of information from the underlying function. This paper presents a novel approach that addresses both these issues. We propose a secure aggregation protocol with a robustness guarantee, effectively protecting the system from "faulty" inputs introduced by malicious clients. Our protocol further ensures differential privacy, so that the underlying function will not leak significant information about individuals. Notably, this work represents the first comprehensive effort to combine robustness and differential privacy guarantees in the context of DME. In particular, we capture the security of the protocol via a notion of "usefulness" combined with differential privacy inspired by the work of Mironov et al. (CRYPTO 2009) and formally analyze this security guarantee for our protocol.