FLAGuard: Efficient Verifiable Federated LoRA of Large Language Models
IEEE Transactions on Mobile Computing · 2026-05 · 25 (5) · 7182-7195
DOI: 10.1109/tmc.2025.3641570 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Mobile Computing, 2026-05. Verfasst von Tianyou Zhang und 4 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 51 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Beijing Natural Science Foundation.
Abstract (Originalsprache)
Federated fine-tuning efficiently adapts large pre-trained models to new tasks by using additional data while minimizing re-training costs. This approach enhances data privacy and reduces computational demands but relies on a central server, often cloud-based, which is vulnerable to adversarial attacks that can compromise the aggregation process. We propose ${\sf FLAGuard}$ FLAGuard , a novel and efficient verification scheme that specifically addresses these challenges in federated Low-Rank Adaptation (LoRA) settings. ${\sf FLAGuard}$ FLAGuard is the first to introduce a two-stage verification process specifically designed for LoRA-based aggregation. In the first stage, the scheme independently verifies the correctness of the aggregated $A$ A and $B$ B matrices. In the second stage, it verifies the multiplication result of the aggregated $A$ A and $B$ B matrices, ensuring the correctness of the final LoRA parameters. Additionally, we introduce the Iterative Gradient Sampling and Convolutional Compression (IGSCC) technique, which combines probabilistic sampling with convolutional operations to efficiently reduce the dimensionality of gradient matrices. This enables secure verification without sacrificing model performance. Our comprehensive security analysis of ${\sf FLAGuard}$ FLAGuard further establishes its reliability in federated learning environments. Extensive experimental results demonstrate that ${\sf FLAGuard}$ FLAGuard achieves over a $100\times$ 100 × speedup in the aggregation verification phase and reduces communication overhead by more than 50% compared to state-of-the-art methods.
Förderung
- National Natural Science Foundation of China
- Beijing Natural Science Foundation
Zitieren
APA
Zhang, T., Yu, H., Chen, Y., Wang, S., & Yang, Z. (2026). FLAGuard: Efficient Verifiable Federated LoRA of Large Language Models. IEEE Transactions on Mobile Computing, 25(5), 7182-7195. https://doi.org/10.1109/tmc.2025.3641570
Vancouver
Zhang T, Yu H, Chen Y, Wang S, Yang Z. FLAGuard: Efficient Verifiable Federated LoRA of Large Language Models. IEEE Transactions on Mobile Computing. 2026;25(5):7182-7195. doi:10.1109/tmc.2025.3641570
BibTeX
@article{Zhang2026,
author = {Zhang, Tianyou and Yu, Haiyang and Chen, Yuwen and Wang, Shuo and Yang, Zhen},
title = {FLAGuard: Efficient Verifiable Federated LoRA of Large Language Models},
journal = {IEEE Transactions on Mobile Computing},
year = {2026},
volume = {25},
number = {5},
pages = {7182--7195},
doi = {10.1109/tmc.2025.3641570}
}
EUVIMED