AI Insight
Researchers have developed UniTexture, a universal adversarial attack that uses a single textured 3D object to manipulate Vision-Language-Action (VLA) robotic systems across multiple tasks simultaneously. By optimizing texture patterns through gradient-based methods and differentiable rendering, the attack causes robots to deviate from intended actions when the adversarial object is present in their visual field. Testing on VLA models OpenVLA and π0.5 showed that task success rates dropped from 90.0% to 48.4%, with the attack transferring across different task sets and model architectures without requiring re-optimization.
Why it matters
This research exposes a critical security vulnerability in multitask robotic systems that could lead to unsafe physical behaviors in real-world deployments. The finding that a single adversarial pattern can compromise performance across diverse tasks suggests that current VLA models share exploitable weaknesses that need to be addressed before widespread adoption in safety-critical applications.
Understand the Science
⚠️ Preprint – Noch nicht peer-reviewed
Dieser Artikel wurde noch nicht von unabhängigen Experten begutachtet. Die Ergebnisse sind vorläufig und sollten mit Vorsicht interpretiert werden.
Abstract: Vision-Language-Action (VLA) models have emerged as generalist robotic policies capable of following diverse language instructions and performing a wide range of manipulation tasks. However, their direct control over embodied agents also exposes them to adversarial interference that may cause unsafe physical behaviors. Existing attacks on robotic policies are typically optimized for a single task or instruction, leaving the cross-task vulnerabilities of multitask VLAs largely unexplored. We introduce UniTexture, a cross-task universal adversarial texture attack that uses a single textured 3D object to induce targeted deviations in VLA action predictions across multiple tasks. UniTexture backpropagates gradients from the policy’s action outputs to surface texture parameters through a differentiable renderer. It jointly optimizes the shared texture over a distribution of tasks, instructions, states, and viewpoints using a targeted action-space objective, steering predicted actions toward attacker-defined targets without optimizing a separate texture for each task. We evaluate UniTexture on OpenVLA and $pi_{0.5}$ across diverse manipulation tasks and multiple evaluation settings. UniTexture reduces the mean task success rate from 90.0% under benign conditions to 48.4% under attack, induces target-aligned action shifts, and further exhibits cross-suite and cross-model transfer without re-optimization. Together, these findings reveal shared cross-task vulnerabilities in multitask VLAs that can be systematically exploited through a single adversarial surface texture.
Source: UniTexture: Cross-Task Universal Adversarial Textures for Vision-Language-Action Models