EUVIMEDEUVIMED entdecken
Crossref

A Novel Cross-Chain Hierarchical Federated Learning Framework for Enhancing Service Security and Communication Efficiency

Li Duan, He Huang, Chao Li, Wei Ni, Bo Cheng

IEEE Transactions on Services Computing · 2025-05 · 18 (3) · 1199-1212

DOI: 10.1109/tsc.2025.3562329 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Services Computing, 2025-05. Verfasst von Li Duan und 4 weiteren Autor:innen. Die Arbeit wurde bisher 10-mal zitiert (Crossref) und nennt 41 Literaturangaben. Gefördert von: Beijing Nova Program, National Natural Science Foundation of China.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Traditional federated learning (FL) uploads local models to a central server for model aggregation and suffers from server centralization. While blockchain-based FL addresses the issue of centralization, new challenges arise, including limited scalability of a single chain, expensive overhead of blockchain consensus, and inconsistent quality of uploaded models. This article proposes a new cross-chain-based FL (CBFL) framework. Specifically, we propose a three-layer cross-chain FL architecture consisting of a task-releasing chain, a relay chain, and local model uploading chains. The task-releasing chain is used for task issuers to release FL tasks and global model aggregation. The local model uploading chain manages local devices, stores local models and aggregates these local models. To verify the quality of local models, we propose a dual-criteria model quality inspection method based on cross entropy and cosine similarity to exclude substandard local models. We also propose hierarchical FL before global model aggregation to further reduce the communication overhead. Moreover, multi-signature is used to ensure the consistent transmission of models in the cross-chain process. Experiments corroborate that the proposed CBFL improves performance by about 50% compared to the existing BFL framework. Moreover, the proposed dual-criteria model quality inspection method has better robustness than Krum and Trimmed Mean.

Förderung

  • Beijing Nova Program
  • National Natural Science Foundation of China

Zitieren

APA

Duan, L., Huang, H., Li, C., Ni, W., & Cheng, B. (2025). A Novel Cross-Chain Hierarchical Federated Learning Framework for Enhancing Service Security and Communication Efficiency. IEEE Transactions on Services Computing, 18(3), 1199-1212. https://doi.org/10.1109/tsc.2025.3562329

Vancouver

Duan L, Huang H, Li C, Ni W, Cheng B. A Novel Cross-Chain Hierarchical Federated Learning Framework for Enhancing Service Security and Communication Efficiency. IEEE Transactions on Services Computing. 2025;18(3):1199-1212. doi:10.1109/tsc.2025.3562329

BibTeX
@article{Duan2025,
  author = {Duan, Li and Huang, He and Li, Chao and Ni, Wei and Cheng, Bo},
  title = {A Novel Cross-Chain Hierarchical Federated Learning Framework for Enhancing Service Security and Communication Efficiency},
  journal = {IEEE Transactions on Services Computing},
  year = {2025},
  volume = {18},
  number = {3},
  pages = {1199--1212},
  doi = {10.1109/tsc.2025.3562329}
}
Kostenlos starten →