EUVIMEDEUVIMED entdecken
Crossref

A Federated Learning Framework Based on Distributed Homomorphic Encryption for Cross-Trust Domain Collaboration in Vehicular Networks

Chao Guo, Xin Zhang, Xiaodong Li, Cheng Gong, Lingcui Zhang, Haitao Xu, Zhu Han

IEEE Transactions on Mobile Computing · 2026-12 · 25 (12) · 22536-22553

DOI: 10.1109/tmc.2026.3723868 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Mobile Computing, 2026-12. Verfasst von Chao Guo und 6 weiteren Autor:innen. Die Arbeit nennt 43 Literaturangaben. Gefördert von: National Key Research and Development Program of China, Fundamental Research Funds for the Central Universities, Fundamental Research Funds for Beijing Electrical Science and Technology Institute.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

With the rapid development of artificial intelligence and new energy technologies, intelligent driving has become a core component of modern transportation infrastructure. The open nature and diverse connectivity of vehicular networks raise two coupled requirements: privacy-preserving collaboration across competing vehicle fleets and freedom from single-point-of-failure risks in the aggregation layer. Existing federated learning solutions that rely on a central aggregator and on lattice-based or Paillier-style homomorphic encryption struggle to meet the real-time budget of V2X communication. This paper proposes a distributed homomorphic-encryption federated learning algorithm for cross-trust-domain collaboration in vehicular networks. The algorithm combines three components: a vehicle clustering and ring-formation scheme based on graph attention networks, a lightweight obfuscated modular decomposition homomorphic encryption scheme, and a ring-topology distributed aggregation protocol. The proposed scheme achieves 87.28% test accuracy on CIFAR-10, exceeding the no-encryption FedAvg baseline of 86.78%, and retains this advantage under a strict non-IID partition with Dirichlet $\alpha =0.3$ α = 0 . 3 . Per-round encryption is 7.6 times faster than CKKS-FL and produces 4.4 times less ciphertext. Against gradient-inversion attacks, the aligned structural similarity index drops to 0.027 on CIFAR-10, an order of magnitude below unprotected FedAvg, and remains an order of magnitude below the baseline even when up to three ring members collude. The network-layer simulation confirms that the ring topology meets the V2X channel coherence time with controlled latency and throughput overhead.

Förderung

  • National Key Research and Development Program of China
  • Fundamental Research Funds for the Central Universities
  • Fundamental Research Funds for Beijing Electrical Science and Technology Institute

Zitieren

APA

Guo, C., Zhang, X., Li, X., Gong, C., Zhang, L., Xu, H., & Han, Z. (2026). A Federated Learning Framework Based on Distributed Homomorphic Encryption for Cross-Trust Domain Collaboration in Vehicular Networks. IEEE Transactions on Mobile Computing, 25(12), 22536-22553. https://doi.org/10.1109/tmc.2026.3723868

Vancouver

Guo C, Zhang X, Li X, Gong C, Zhang L, Xu H, et al. A Federated Learning Framework Based on Distributed Homomorphic Encryption for Cross-Trust Domain Collaboration in Vehicular Networks. IEEE Transactions on Mobile Computing. 2026;25(12):22536-22553. doi:10.1109/tmc.2026.3723868

BibTeX
@article{Guo2026,
  author = {Guo, Chao and Zhang, Xin and Li, Xiaodong and Gong, Cheng and Zhang, Lingcui and Xu, Haitao and Han, Zhu},
  title = {A Federated Learning Framework Based on Distributed Homomorphic Encryption for Cross-Trust Domain Collaboration in Vehicular Networks},
  journal = {IEEE Transactions on Mobile Computing},
  year = {2026},
  volume = {25},
  number = {12},
  pages = {22536--22553},
  doi = {10.1109/tmc.2026.3723868}
}
Kostenlos starten →