.png){kind=icon}
.png){kind=icon}
Return
Robust Quantum Federated Learning Against Colluding and Non-Colluding Byzantine Attacks
.png){kind=icon}
OA
AI
J
X
T
S
X
DOI：10.1109/tifs.2026.3714274.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Quantum federated learning may suffer performance degradation when Byzantine nodes are involved. Previous defense measures often focus on either non-colluding or colluding Byzantine attacks. As a result, the performance of models trained using quantum federated learning may be affected in scenarios where both colluding and non-colluding Byzantine attacks coexist. This paper introduces a robust quantum federated learning method to tackle the Byzantine problem that encompasses both colluding and non-colluding attacks. First, an adaptive clustering-based defense algorithm is proposed by extending DBSCAN and designing an adaptive weight allocation algorithm, which instantiates a server-side robust aggregation mechanism for quantum federated learning under the considered mixed attack settings. Second, a robust quantum federated learning architecture is constructed based on the proposed adaptive clustering-based defense algorithm to defend against Byzantine attacks, and we design a corresponding image classification scheme. Finally, experimental results demonstrate the robustness of the proposed method across different datasets and Byzantine attack scenarios, including label-flipping attacks on MNIST and Fashion-MNIST. It achieves high performance, with MNIST accuracy reaching 98%, representing a 38-percentage-point improvement over the undefended baseline. It also reduces attack-induced performance degradation in the MNIST backdoor attack setting. The method provides a robust aggregation mechanism for improving the reliability of quantum federated learning under mixed Byzantine attack scenarios.
Keywords:
Quantum federated learning
quantum neural network
robust aggregation algorithm
Byzantine attack
Journal
IF:
8
Papers:
5.2K
Citations:
2.3W
Organization
Cited Papers