Tag
The paper introduces Daydreaming, an attack that steals hidden agent skills through normal task interactions, demonstrating that hiding skill files does not protect them from reconstruction.
This paper proposes DURA, a diffusion-based unrestricted robotic attack that generates visually natural adversarial patches to disrupt Vision-Language-Action (VLA) models in both white-box and black-box settings, highlighting safety risks for physically deployed robotic systems.