EUVIMEDEUVIMED entdecken
Crossref

Privacy Preserving Decentralized Online Federated Learning for Consumer Electronics

Mengli Wei, Zhifei Liu, Lunchao Xia, Zhongyuan Zhao

IEEE Transactions on Consumer Electronics · 2026-02 · 72 (1) · 848-860

DOI: 10.1109/tce.2026.3653552 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Consumer Electronics, 2026-02. Verfasst von Mengli Wei und 3 weiteren Autor:innen. Die Arbeit wurde bisher 3-mal zitiert (Crossref) und nennt 46 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Basic Research Program of Jiangsu, Open Fund.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Modern Consumer Electronics (CE) ecosystems, including smartphones, wearables and home IoT devices, generate rich local data that are privacy sensitive and seldom shareable, which makes centralized aggregation impractical. In this paper, a setting is considered in which each agent is a consumer device that collaboratively trains a global model while keeping raw data local and communicating over directed networks. Accordingly, a differential privacy-based distributed online one-point residual bandit optimization (DP-DORO) algorithm is proposed for constrained multi-agent optimization, where the cost is unknown a priori and accessible only through function-value queries. Specifically, row and column stochastic matrices are employed, which are more amenable to directed graphs than doubly stochastic designs. Moreover, privacy leakage during information exchange is mitigated by injecting single-state randomized perturbations that deliver differential privacy guarantees. In addition, for time-varying and unknown costs, a local one-point gradient estimator is constructed at each device and decisions are updated using function values only, thereby avoiding gradient queries per iteration. Under mild conditions, it is established that agents consensus-track a common trajectory, and that the dynamic regret grows sublinearly with time $T$ . Finally, three experiments with common datasets are conducted to verify the effectiveness of DP-DORO for privacy-preserving collaborative learning in CE networks.

Förderung

  • National Natural Science Foundation of China
  • Basic Research Program of Jiangsu
  • Open Fund
  • Postgraduate Research and Practice Innovation Program of Jiangsu Province

Zitieren

APA

Wei, M., Liu, Z., Xia, L., & Zhao, Z. (2026). Privacy Preserving Decentralized Online Federated Learning for Consumer Electronics. IEEE Transactions on Consumer Electronics, 72(1), 848-860. https://doi.org/10.1109/tce.2026.3653552

Vancouver

Wei M, Liu Z, Xia L, Zhao Z. Privacy Preserving Decentralized Online Federated Learning for Consumer Electronics. IEEE Transactions on Consumer Electronics. 2026;72(1):848-860. doi:10.1109/tce.2026.3653552

BibTeX
@article{Wei2026,
  author = {Wei, Mengli and Liu, Zhifei and Xia, Lunchao and Zhao, Zhongyuan},
  title = {Privacy Preserving Decentralized Online Federated Learning for Consumer Electronics},
  journal = {IEEE Transactions on Consumer Electronics},
  year = {2026},
  volume = {72},
  number = {1},
  pages = {848--860},
  doi = {10.1109/tce.2026.3653552}
}
Kostenlos starten →