Proximal Splitting Adversarial Attack for Semantic Segmentation - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2023

Proximal Splitting Adversarial Attack for Semantic Segmentation

Résumé

Classification has been the focal point of research on adversarial attacks, but only a few works investigate methods suited to denser prediction tasks, such as semantic segmentation. The methods proposed in these works do not accurately solve the adversarial segmentation problem and, therefore, overestimate the size of the perturbations required to fool models. Here, we propose a white-box attack for these models based on a proximal splitting to produce adversarial perturbations with much smaller \ell_\infty norms. Our attack can handle large numbers of constraints within a nonconvex minimization framework via an Augmented Lagrangian approach, coupled with adaptive constraint scaling and masking strategies. We demonstrate that our attack significantly outperforms previously proposed ones, as well as classification attacks that we adapted for segmentation, providing a first comprehensive benchmark for this dense task.

Dates et versions

hal-04382554 , version 1 (09-01-2024)

Licence

Paternité

Identifiants

Citer

Jérôme Rony, Jean-Christophe Pesquet, Ismail Ben Ayed. Proximal Splitting Adversarial Attack for Semantic Segmentation. CVPR 2023 - IEEE/CVF Conference on Computer Vision and Pattern Recognition, Jun 2023, Vancouver (CA), Canada. pp.20524-20533, ⟨10.1109/CVPR52729.2023.01966⟩. ⟨hal-04382554⟩
32 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More