EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Optimizing Utility–Privacy Tradeoff in Federated Learning Over Fading Channels

Yujia Mu, Xizixiang Wei, Yonina C. Eldar, Cong Shen

IEEE Transactions on Machine Learning in Communications and Networking · 2026 · 4 · 1058-1075

DOI: 10.1109/tmlcn.2026.3706257 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Machine Learning in Communications and Networking, 2026. Verfasst von Yujia Mu und 3 weiteren Autor:innen. Die Arbeit nennt 87 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: U.S. National Science Foundation, Tom and Mary Beck Center for Renewable Energy as part of the Institute for Environmental Sustainability (IES) at the Weizmann Institute of Science.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This paper develops a utility-privacy optimized federated learning (FL) framework that aggregates local updates transmitted over fading channels using only channel state information at the receiver (CSIR). We first establish a rigorous convergence analysis of distributed stochastic gradient descent (SGD) under arbitrary aggregation weights and channel fading, deriving an upper bound that explicitly captures the combined effects of channel quality and SGD noise. Using this bound, we formulate and solve an optimization problem that yields closed-form or efficiently computable aggregation weights. The resulting method provably achieves the standard ${\mathcal {O}}(1/t)$ convergence rate for strongly convex objectives, even under arbitrarily varying fading channels. We further characterize the differential privacy (DP) induced by channel noise and develop a principled mechanism that enables explicit utility-privacy tradeoffs through weight selection. Extensive experiments on MNIST, CIFAR-10, Shakespeare, and F-EMNIST demonstrate that the proposed fading-aware aggregation eliminates the training failures suffered by FedAvg under deep fading and approaches the performance of ideal error-free communication. The results provide a unified convergence-communication-privacy framework for robust and privacy-aware FL over wireless fading channels.

Förderung

  • U.S. National Science Foundation
  • Tom and Mary Beck Center for Renewable Energy as part of the Institute for Environmental Sustainability (IES) at the Weizmann Institute of Science

Zitieren

APA

Mu, Y., Wei, X., Eldar, Y. C., & Shen, C. (2026). Optimizing Utility–Privacy Tradeoff in Federated Learning Over Fading Channels. IEEE Transactions on Machine Learning in Communications and Networking, 4, 1058-1075. https://doi.org/10.1109/tmlcn.2026.3706257

Vancouver

Mu Y, Wei X, Eldar YC, Shen C. Optimizing Utility–Privacy Tradeoff in Federated Learning Over Fading Channels. IEEE Transactions on Machine Learning in Communications and Networking. 2026;4:1058-1075. doi:10.1109/tmlcn.2026.3706257

BibTeX
@article{Mu2026,
  author = {Mu, Yujia and Wei, Xizixiang and Eldar, Yonina C. and Shen, Cong},
  title = {Optimizing Utility–Privacy Tradeoff in Federated Learning Over Fading Channels},
  journal = {IEEE Transactions on Machine Learning in Communications and Networking},
  year = {2026},
  volume = {4},
  pages = {1058--1075},
  doi = {10.1109/tmlcn.2026.3706257}
}
Kostenlos starten →