EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

CAEBA: A Dynamic Hidden Backdoor Attack Framework in Federated Learning

Xiaojun Guo, Guoliang Li, Yun Hu

ACM Transactions on Privacy and Security · 31.08.2026 · 29 (3) · 1-21

DOI: 10.1145/3831698 ↗

Kurz gefasst

Fachartikel in der Zeitschrift ACM Transactions on Privacy and Security, online erschienen am 05.08.2026. Verfasst von Xiaojun Guo und 2 weiteren Autor:innen. Die Arbeit nennt 55 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Federated Learning (FL) is a privacy-preserving distributed learning framework, but its distributed data collection and client-side training pipeline expose the global model to backdoor attacks. Existing FL backdoor attacks often depend on fixed triggers or on update manipulations that are easier to isolate under robust aggregation and model-inspection defenses. We propose CAEBA (Conditional AutoEncoder Backdoor Attack), a dynamic hidden backdoor framework that uses a conditional autoencoder to generate target-aware and visually stealthy triggers while progressively implanting the backdoor through federated optimization. CAEBA separates the clean classifier from the trigger generator, formulates the attack as a dual-constrained objective, and updates the generator at a lower frequency than the classifier to stabilize local optimization. We evaluate CAEBA on MNIST, FashionMNIST, CIFAR-10, CIFAR-100, and Tiny-ImageNet under representative aggregation rules and defenses, including FL-Detector, RFLBAT, DeepSight, FoolsGold, and FLAME. The results show that CAEBA preserves main-task accuracy while maintaining persistent backdoor effectiveness.

Zitieren

APA

Guo, X., Li, G., & Hu, Y. (2026). CAEBA: A Dynamic Hidden Backdoor Attack Framework in Federated Learning. ACM Transactions on Privacy and Security, 29(3), 1-21. https://doi.org/10.1145/3831698

Vancouver

Guo X, Li G, Hu Y. CAEBA: A Dynamic Hidden Backdoor Attack Framework in Federated Learning. ACM Transactions on Privacy and Security. 2026;29(3):1-21. doi:10.1145/3831698

BibTeX
@article{Guo2026,
  author = {Guo, Xiaojun and Li, Guoliang and Hu, Yun},
  title = {CAEBA: A Dynamic Hidden Backdoor Attack Framework in Federated Learning},
  journal = {ACM Transactions on Privacy and Security},
  year = {2026},
  volume = {29},
  number = {3},
  pages = {1--21},
  doi = {10.1145/3831698}
}
Kostenlos starten →