First Provable Guarantees for Practical Private FL: Beyond Restrictive Assumptions

TL;DR

This paper introduces Fed-α-NormEC, a federated learning framework that provides provable differential privacy guarantees under realistic assumptions, supporting practical features like multiple local updates and partial client participation.

Contribution

It is the first to offer theoretical convergence and privacy guarantees for private federated learning with practical features such as partial participation and multiple local updates.

Findings

Fed-α-NormEC achieves convergence with differential privacy under standard assumptions.

Supports partial client participation, enhancing real-world applicability.

Experimental results validate theoretical guarantees on deep learning tasks.

Abstract

Federated Learning (FL) enables collaborative training on decentralized data. Differential privacy (DP) is crucial for FL, but current private methods often rely on unrealistic assumptions (e.g., bounded gradients or heterogeneity), hindering practical application. Existing works that relax these assumptions typically neglect practical FL features, including multiple local updates and partial client participation. We introduce Fed-$\alpha$-NormEC, the first differentially private FL framework providing provable convergence and DP guarantees under standard assumptions while fully supporting these practical features. Fed-$\alpha$-NormE integrates local updates (full and incremental gradient steps), separate server and client stepsizes, and, crucially, partial client participation, which is essential for real-world deployment and vital for privacy amplification. Our theoretical guarantees are corroborated by experiments on private deep learning tasks.