EUVIMEDEUVIMED entdecken
Crossref

Efficient Privacy-Preserving Federated Learning for Consumer Electronics Using Proxy Re-Encryption

Jianhong Zhang, Chuming Shi

IEEE Transactions on Consumer Electronics · 2025-11 · 71 (4) · 10220-10234

DOI: 10.1109/tce.2025.3596793 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Consumer Electronics, 2025-11. Verfasst von Jianhong Zhang und 1 weiteren Autor:innen. Die Arbeit wurde bisher 4-mal zitiert (Crossref) und nennt 36 Literaturangaben. Gefördert von: Natural Science Foundation of Beijing Municipality.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The rapid evolution of consumer electronics, driven by advancements in big data and deep learning, has significantly enhanced user experiences. However, limited datasets frequently lead to suboptimal model performance and overfitting. Federated learning mitigates this challenge by utilizing distributed datasets from consumer devices. To resist membership inference and model inversion attacks, various privacy-preserving federated learning (PPFL) schemes have been proposed, incorporating techniques such as homomorphic encryption and differential privacy to protect local model privacy. Nevertheless, these schemes often expose the global model, which makes participants’ sensitive data susceptible to reconstruction. To address these challenges, we propose a novel secure federated learning framework that enhances data privacy in consumer electronics through proxy re-encryption. Specifically, our approach adopts an ECC-based proxy re-encryption mechanism that allows participants to independently decrypt the global model during training while preventing the parameter server from accessing it. This mechanism not only prevents unauthorized access by the parameter server but also reduces the computational and communication overhead on participants. Moreover, it effectively prevents collusion attacks between malicious participants and the parameter server. Comprehensive comparative analysis demonstrates that our framework outperforms existing schemes, achieving a 38.6% reduction in communication overhead and a 27.4% decrease in computational complexity. Under the setting of 40 training iterations with a batch size of 64, it reaches an accuracy of 84.10%. This balanced integration of security and efficiency makes our framework particularly suitable for privacy-sensitive consumer electronics applications.

Förderung

  • Natural Science Foundation of Beijing Municipality

Zitieren

APA

Zhang, J., & Shi, C. (2025). Efficient Privacy-Preserving Federated Learning for Consumer Electronics Using Proxy Re-Encryption. IEEE Transactions on Consumer Electronics, 71(4), 10220-10234. https://doi.org/10.1109/tce.2025.3596793

Vancouver

Zhang J, Shi C. Efficient Privacy-Preserving Federated Learning for Consumer Electronics Using Proxy Re-Encryption. IEEE Transactions on Consumer Electronics. 2025;71(4):10220-10234. doi:10.1109/tce.2025.3596793

BibTeX
@article{Zhang2025,
  author = {Zhang, Jianhong and Shi, Chuming},
  title = {Efficient Privacy-Preserving Federated Learning for Consumer Electronics Using Proxy Re-Encryption},
  journal = {IEEE Transactions on Consumer Electronics},
  year = {2025},
  volume = {71},
  number = {4},
  pages = {10220--10234},
  doi = {10.1109/tce.2025.3596793}
}
Kostenlos starten →