Saved in:
Bibliographic Details
Main Authors: Liu, Xinlei, Hu, Tao, Xie, Jichao, Yi, Peng, Ma, Hailong, Li, Baolin
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.20308
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911698844123136
author Liu, Xinlei
Hu, Tao
Xie, Jichao
Yi, Peng
Ma, Hailong
Li, Baolin
author_facet Liu, Xinlei
Hu, Tao
Xie, Jichao
Yi, Peng
Ma, Hailong
Li, Baolin
contents Gradient-based attacks are important methods for evaluating model robustness. However, since the proposal of APGD, it has been difficult for such methods to achieve significant breakthroughs. To achieve such an effect, we first analyze the issue of "high-loss non-adversarial examples" that degrades attack performance in previous methods, and prove that this issue arises from inappropriate objectives for adversarial example generation. Subsequently, we reconstruct the objective as "maximizing the difference between the non-ground-truth label probability upper bound and the ground-truth label probability", and proposes a novel and powerful gradient-based attack method named Sequential Difference Maximization (SDM). SDM establishes a three-layer optimization framework of "cycle-stage-step". It adopts the negative probability loss function and the Directional Probability Difference Ratio (DPDR) loss function in the initial and subsequent optimization stages, respectively, and approaches the ideal objective of adversarial example generation via stage-wise sequential optimization. Experiments demonstrate that compared with previous state-of-the-art methods, SDM not only achieves stronger attack performance but also exhibits superior cost-effectiveness. The code is available at https://github.com/X-L-Liu/ICML-SDM.
format Preprint
id arxiv_https___arxiv_org_abs_2605_20308
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle SDM: A Powerful Tool for Evaluating Model Robustness
Liu, Xinlei
Hu, Tao
Xie, Jichao
Yi, Peng
Ma, Hailong
Li, Baolin
Computer Vision and Pattern Recognition
Artificial Intelligence
Machine Learning
Gradient-based attacks are important methods for evaluating model robustness. However, since the proposal of APGD, it has been difficult for such methods to achieve significant breakthroughs. To achieve such an effect, we first analyze the issue of "high-loss non-adversarial examples" that degrades attack performance in previous methods, and prove that this issue arises from inappropriate objectives for adversarial example generation. Subsequently, we reconstruct the objective as "maximizing the difference between the non-ground-truth label probability upper bound and the ground-truth label probability", and proposes a novel and powerful gradient-based attack method named Sequential Difference Maximization (SDM). SDM establishes a three-layer optimization framework of "cycle-stage-step". It adopts the negative probability loss function and the Directional Probability Difference Ratio (DPDR) loss function in the initial and subsequent optimization stages, respectively, and approaches the ideal objective of adversarial example generation via stage-wise sequential optimization. Experiments demonstrate that compared with previous state-of-the-art methods, SDM not only achieves stronger attack performance but also exhibits superior cost-effectiveness. The code is available at https://github.com/X-L-Liu/ICML-SDM.
title SDM: A Powerful Tool for Evaluating Model Robustness
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2605.20308