Criticality Leveraged Adversarial Training (CLAT) for Boosted Performance via Parameter Efficiency

TL;DR

CLAT introduces a parameter-efficient adversarial training method that focuses on fine-tuning critical layers, significantly reducing parameters and improving robustness and accuracy.

Contribution

This paper proposes CLAT, a novel approach that enhances adversarial training by selectively fine-tuning critical layers, reducing parameters and boosting robustness.

Findings

Reduces trainable parameters by approximately 95%.

Achieves over 2% improvement in adversarial robustness.

Can be integrated with existing adversarial training methods.

Abstract

Adversarial training enhances neural network robustness but suffers from a tendency to overfit and increased generalization errors on clean data. This work introduces CLAT, an innovative approach that mitigates adversarial overfitting by introducing parameter efficiency into the adversarial training process, improving both clean accuracy and adversarial robustness. Instead of tuning the entire model, CLAT identifies and fine-tunes robustness-critical layers - those predominantly learning non-robust features - while freezing the remaining model to enhance robustness. It employs dynamic critical layer selection to adapt to changes in layer criticality throughout the fine-tuning process. Empirically, CLAT can be applied on top of existing adversarial training methods, significantly reduces the number of trainable parameters by approximately 95%, and achieves more than a 2% improvement in adversarial robustness compared to baseline methods.