Preference Redirection via Attention Concentration: An Attack on Computer Use Agents
This addresses a critical security problem for companies using open-weight CUAs, as it exploits vision modality vulnerabilities, though it is incremental by focusing on a specific attack method.
The paper tackles the security vulnerability of Computer Use Agents (CUAs) by introducing PRAC, an attack that manipulates model preferences via attention redirection to steer product selection on an online shopping platform, demonstrating generalization to fine-tuned versions.
Advancements in multimodal foundation models have enabled the development of Computer Use Agents (CUAs) capable of autonomously interacting with GUI environments. As CUAs are not restricted to certain tools, they allow to automate more complex agentic tasks but at the same time open up new security vulnerabilities. While prior work has concentrated on the language modality, the vulnerability of the vision modality has received less attention. In this paper, we introduce PRAC, a novel attack that, unlike prior work targeting the VLM output directly, manipulates the model's internal preferences by redirecting its attention toward a stealthy adversarial patch. We show that PRAC is able to manipulate the selection process of a CUA on an online shopping platform towards a chosen target product. While we require white-box access to the model for the creation of the attack, we show that our attack generalizes to fine-tuned versions of the same model, presenting a critical threat as multiple companies build specific CUAs based on open weights models.