Vision language action models improved for stronger visual attack redirection

Enhancing Visual Task-Redirection Attacks on Vision-Language-Action Models via Closed-Loop Optimization

Robotics

Summary

Some robots can follow instructions using both visuals and language, but they can be tricked by altered images to behave differently than intended. The authors found that previous tricks didn't always work well because the robot's changing actions could confuse the attack. They created a new approach, called TRACO, that adjusts its strategy based on what the robot actually does, making these tricks more successful. This shows a potential security risk for robots that rely on vision and language to follow tasks.

What this means in practice

Authors

Chunjin Ye, Guanying Chen, Jianjie Huang, Jianhao Chen, Quanlin Chen, Hanjie Zhao, Haozhe Liang, Tieyun Qian, Xiaochun Cao

Abstract

Vision-language-action (VLA) models enable language-guided robot manipulation, but remain vulnerable to visual attacks that undermine user control over robot behavior. Prior studies optimize visual perturbations using fixed target supervision. However, the perturbation can change earlier actions and, in turn, the physical states reached later. As the attacked trajectory evolves, this supervision may no longer match the behavior needed at the states actually reached. Accurately matching these targets can therefore fail to advance the attacker-specified task. Even with suitable supervision, reducing average matching error does not ensure sustained task progress. The robot may still fail at a critical stage, and further updates can disrupt previously successful behavior. To address these limitations, we propose TRACO, a framework that enhances existing targeted visual attacks through closed-loop optimization. Its Execution-Driven Supervision rebuilds supervision for the states reached under attack to guide subsequent updates. Within this framework, Progress-Guided Selection focuses updates where progress stalls while maintaining supervision on the earlier behavior needed to reach that point. Experiments across VLA models and simulation environments show improved task-redirection success with TRACO, highlighting the security implications of closed-loop attack optimization.