Adaptive Gradient Clipping for Robust Federated Learning

TL;DR

This paper introduces Adaptive Robust Clipping (ARC), a dynamic gradient clipping method that improves robustness and convergence in federated learning, especially under adversarial attacks and data heterogeneity.

Contribution

The paper proposes ARC, an adaptive gradient clipping strategy that maintains theoretical robustness guarantees and enhances empirical performance in federated learning.

Findings

ARC improves robustness against adversarial attacks.

ARC enhances convergence in heterogeneous settings.

Experimental results confirm ARC's effectiveness on image classification tasks.

Abstract

Robust federated learning aims to maintain reliable performance despite the presence of adversarial or misbehaving workers. While state-of-the-art (SOTA) robust distributed gradient descent (Robust-DGD) methods were proven theoretically optimal, their empirical success has often relied on pre-aggregation gradient clipping. However, existing static clipping strategies yield inconsistent results: enhancing robustness against some attacks while being ineffective or even detrimental against others. To address this limitation, we propose a principled adaptive clipping strategy, Adaptive Robust Clipping (ARC), which dynamically adjusts clipping thresholds based on the input gradients. We prove that ARC not only preserves the theoretical robustness guarantees of SOTA Robust-DGD methods but also provably improves asymptotic convergence when the model is well-initialized. Extensive experiments on benchmark image classification tasks confirm these theoretical insights, demonstrating that ARC significantly enhances robustness, particularly in highly heterogeneous and adversarial settings.