What question did this study set out to answer?

This research focuses on enhancing the transferability of unrestricted adversarial examples using a novel method.

May 6, 2026SensorsOpen Access

DeDiAttack: Enhancing Transferability of Unrestricted Adversarial Examples via Deformation-Constrained Diffusion

View Full Paper

Ask AI

Bookmark

Authors

BQBin QuOsnabrück University APAnjie PengSouthwest University of Science and Technology SZShijie ZhaoSouthwest University of Science and Technology

Discussion

Member takes

Implication

DeDiAttack enhances transferability of adversarial examples in surrogate models by leveraging elastic deformation, indicating better visual fidelity and robustness evaluation.

Key Points

This research focuses on enhancing the transferability of unrestricted adversarial examples using a novel method.
Proposed DeDiAttack based on deformation-constrained diffusion
Utilized elastic deformation during the denoising process
Conducted extensive black-box experiments on UAEs
DeDiAttack improves black-box transferability of generated UAEs by 7.2% on the ViT-B surrogate model
Effectively eliminates unnatural artifacts from adversarial examples
Provides tools for robustness evaluation in vision-based systems

Cite This Study

Qu et al. (2026) studied this question.

synapsesocial.com/papers/69fa989404f884e66b532668 https://doi.org/10.3390/s26092823

View Full Paper

Ask AI

Bookmark

Implication

DeDiAttack enhances transferability of adversarial examples in surrogate models by leveraging elastic deformation, indicating better visual fidelity and robustness evaluation.

DeDiAttack: Enhancing Transferability of Unrestricted Adversarial Examples via Deformation-Constrained Diffusion

Discussion

Member takes

Implication

Key Points

Cite This Study

DeDiAttack: Enhancing Transferability of Unrestricted Adversarial Examples via Deformation-Constrained Diffusion

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider