English

Dynamic Epsilon Scheduling: A Multi-Factor Adaptive Perturbation Budget for Adversarial Training

Computer Vision and Pattern Recognition 2025-12-05 v2 Machine Learning

Abstract

Adversarial training is among the most effective strategies for defending deep neural networks against adversarial examples. A key limitation of existing adversarial training approaches lies in their reliance on a fixed perturbation budget, which fails to account for instance-specific robustness characteristics. While prior works such as IAAT and MMA introduce instance-level adaptations, they often rely on heuristic or static approximations of data robustness. In this paper, we propose Dynamic Epsilon Scheduling (DES), a novel framework that adaptively adjusts the adversarial perturbation budget per instance and per training iteration. DES integrates three key factors: (1) the distance to the decision boundary approximated via gradient-based proxies, (2) prediction confidence derived from softmax entropy, and (3) model uncertainty estimated via Monte Carlo dropout. By combining these cues into a unified scheduling strategy, DES tailors the perturbation budget dynamically to guide more effective adversarial learning. Experimental results on CIFAR-10 and CIFAR-100 show that our method consistently improves both adversarial robustness and standard accuracy compared to fixed-epsilon baselines and prior adaptive methods. Moreover, we provide theoretical insights into the stability and convergence of our scheduling policy. This work opens a new avenue for instance-aware, data-driven adversarial training methods.

Keywords

Cite

@article{arxiv.2506.04263,
  title  = {Dynamic Epsilon Scheduling: A Multi-Factor Adaptive Perturbation Budget for Adversarial Training},
  author = {Alan Mitkiy and James Smith and Myungseo wong and Hana Satou and Hiroshi Tanaka and Emily Johnson},
  journal= {arXiv preprint arXiv:2506.04263},
  year   = {2025}
}

Comments

arXiv admin note: This version has been removed by arXiv administrators as the submitter did not have the right to agree to the license at the time of submission

R2 v1 2026-07-01T02:59:40.853Z