First Provable Guarantees for Practical Private FL: Beyond Restrictive Assumptions

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Shulgin, Egor, Malinovsky, Grigory, Khirirat, Sarit, Richtárik, Peter
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908731789279232
author Shulgin, Egor
Malinovsky, Grigory
Khirirat, Sarit
Richtárik, Peter
author_facet Shulgin, Egor
Malinovsky, Grigory
Khirirat, Sarit
Richtárik, Peter
contents 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-$α$-NormEC, the first differentially private FL framework providing provable convergence and DP guarantees under standard assumptions while fully supporting these practical features. Fed-$α$-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.
format Preprint
id arxiv_https___arxiv_org_abs_2512_21521
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle First Provable Guarantees for Practical Private FL: Beyond Restrictive Assumptions
Shulgin, Egor
Malinovsky, Grigory
Khirirat, Sarit
Richtárik, Peter
Machine Learning
Optimization and Control
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-$α$-NormEC, the first differentially private FL framework providing provable convergence and DP guarantees under standard assumptions while fully supporting these practical features. Fed-$α$-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.
title First Provable Guarantees for Practical Private FL: Beyond Restrictive Assumptions
topic Machine Learning
Optimization and Control
url https://arxiv.org/abs/2512.21521