Improving PID Control with Bayesian Optimization for Adversarially Robust Federated Learning

Authors

• Adrian Pena, The University of Texas Rio Grande Valley
• Sergei Chuprov, The University of Texas Rio Grande ValleyFollow
• Raman Zatsarenko
• Leon Reznik

Document Type

Conference Proceeding

Publication Date

2026

Abstract

Federated Learning (FL) often suffers from unstable convergence and reduced robustness under non-IID data and malicious attacks. In this paper, we present FedPIDAvg_tuned, a control-theoretic aggregation framework for improving stability and adversarial robustness in FL. In particular, our approach combines a server-side Proportional–Integral–Derivative (PID) controller with Bayesian optimization to tune controller gains for different data and attack conditions. The PID controller regulates global model updates through feedback on loss dynamics, providing adaptive scaling that improves stability and convergence. The tuned gains are applied within a trust-weighted trimmed-mean mechanism to remove adversarial or outlier updates. Using the Flower framework, we evaluate our method on IID (CIFAR-10) and non-IID (FEMNIST) datasets under clean and adversarial settings. Results show that FedPIDAvg_tuned achieves higher accuracy and stability than FedAvg, Multi-Krum, and Bulyan, showing that feedback control with Bayesian tuning is a practical path toward robust FL.

Comments

© 2026 Copyright held by the owner/author(s).

Recommended Citation

Pena, Adrian, Sergei Chuprov, Raman Zatsarenko, and Leon Reznik. "Improving PID Control with Bayesian Optimization for Adversarially Robust Federated Learning." In Proceedings of the 2026 9th International Conference on Information and Computer Technologies, pp. 149-154. 2026. https://doi.org/10.1145/3803291.380337

Publication Title

Proceedings of the 2026 9th International Conference on Information and Computer Technologies

DOI

10.1145/3803291.380337