attack arXiv Mar 21, 2026 · 16d ago
Aarush Aggarwal, Akshat Tomar, Amritanshu Tiwari et al. · Indian Institute of Technology Roorkee
Universal adversarial patch attack achieving cross-architecture transfer from ViT to CNN segmentation models for autonomous driving
Input Manipulation Attack vision
Robust semantic segmentation is crucial for safe autonomous driving, yet deployed models remain vulnerable to black-box adversarial attacks when target weights are unknown. Most existing approaches either craft image-wide perturbations or optimize patches for a single architecture, which limits their practicality and transferability. We introduce OmniPatch, a training framework for learning a universal adversarial patch that generalizes across images and both ViT and CNN architectures without requiring access to target model parameters.
cnn transformer Indian Institute of Technology Roorkee
ML Security Papers