Annealing Self-Distillation Rectification Improves Adversarial Training
This addresses adversarial robustness for machine learning models, offering an incremental improvement through a plug-and-play method.
The paper tackles robust overfitting in adversarial training by proposing Annealing Self-Distillation Rectification (ADR), which uses soft labels to reflect distribution shifts under attack, resulting in improved model robustness across datasets.
In standard adversarial training, models are optimized to fit one-hot labels within allowable adversarial perturbation budgets. However, the ignorance of underlying distribution shifts brought by perturbations causes the problem of robust overfitting. To address this issue and enhance adversarial robustness, we analyze the characteristics of robust models and identify that robust models tend to produce smoother and well-calibrated outputs. Based on the observation, we propose a simple yet effective method, Annealing Self-Distillation Rectification (ADR), which generates soft labels as a better guidance mechanism that accurately reflects the distribution shift under attack during adversarial training. By utilizing ADR, we can obtain rectified distributions that significantly improve model robustness without the need for pre-trained models or extensive extra computation. Moreover, our method facilitates seamless plug-and-play integration with other adversarial training techniques by replacing the hard labels in their objectives. We demonstrate the efficacy of ADR through extensive experiments and strong performances across datasets.