One-shot Empirical Privacy Estimation for Federated Learning

TL;DR

This paper introduces a novel one-shot method for empirically estimating privacy loss in federated learning models, enabling efficient privacy auditing during training without extensive retraining or prior knowledge.

Contribution

The work presents a scalable, provably correct one-shot privacy estimation technique applicable during a single training run in federated learning, overcoming limitations of previous methods.

Findings

Provides provably correct privacy estimates under Gaussian mechanism

Demonstrates effectiveness on federated learning benchmarks

Works across various adversarial threat models

Abstract

Privacy estimation techniques for differentially private (DP) algorithms are useful for comparing against analytical bounds, or to empirically measure privacy loss in settings where known analytical bounds are not tight. However, existing privacy auditing techniques usually make strong assumptions on the adversary (e.g., knowledge of intermediate model iterates or the training data distribution), are tailored to specific tasks, model architectures, or DP algorithm, and/or require retraining the model many times (typically on the order of thousands). These shortcomings make deploying such techniques at scale difficult in practice, especially in federated settings where model training can take days or weeks. In this work, we present a novel "one-shot" approach that can systematically address these challenges, allowing efficient auditing or estimation of the privacy loss of a model during the same, single training run used to fit model parameters, and without requiring any a priori knowledge about the model architecture, task, or DP training algorithm. We show that our method provides provably correct estimates for the privacy loss under the Gaussian mechanism, and we demonstrate its performance on well-established FL benchmark datasets under several adversarial threat models.