acceptodds
Under review as a conference paper at ICLR 2027

Secure-CUA: Controlling Untrusted Influence in Computer-Use Agents

Abstract

Computer-use agents (CUAs) perform tasks across applications (such as desktops, mobile apps, and web browsers) by observing graphical interfaces and issuing commands such as clicks and keystrokes. These interfaces combine trusted controls and content with untrusted content needed for legitimate tasks. An adversary controlling this untrusted content can embed instructions or misleading visual cues to change the agent's intended action or redirect its commands to the wrong interface target. We formalize security requirements for both the agent's decisions and their execution through GUI commands. Using an ideal execution model, we show that enforcing these requirements at each step protects complete execution traces. We instantiate this model in , our system for secure CUA execution. Its key idea is to commit to an explicit per-action program, called an *action transaction*, before accessing untrusted content. Each transaction fixes its queries to untrusted content and the permitted uses of their responses. The system masks untrusted regions and evaluates each transaction to produce the next action, using an isolated query model to answer its queries. It then locates the intended interface target using the masked interface. Under the model's assumptions, is secure by design, while generating a new transaction at each step enables adaptation to changing interfaces. We evaluate under benign conditions on 400 WebArena tasks using three frontier models across seeds, yielding 6,000 execution traces. achieves an average task success rate of %, compared with % for Vanilla-CUA and % for CaMeL-CUA.

Primary area:
alignment, fairness, safety, privacy, and societal considerations
Keywords:
Computer-Use Agents, Information-flow control, Agentic Security, Prompt Injection

Reviews

Among:
Shape:
1234this paper, Top 49%
Figure 1. Mean review rating of the reviewed papers at ICLR 2027. This paper averages 2.33 over 3 reviews, above 51% of the others.
Reviewer_2Y5xOpenReview ↗
Rating
1
Confidence
4
Informativeness
2
Presentation
2
Soundness
1
Read review
Paper summary

The paper presents Secure-CUA a react based computer use agent that holds security guarantees under an assumed trusted/untrusted observation partition. Untrusted regions are masked and before any action is executed a one step program is established from this masked content. This content is then accessed through an isolated query model, whose responses may influence the action only as specified by the committed program. The paper formalizes generation integrity and grounding integrity and finally evaluates benign utility of their CUA on 400 WebArena tasks, achieving 53.55% average task success compared with 55.12% for the unprotected baseline.

Critical strengths
  • It is interesting to try to understand how can we try to establish system security guarantees in a react based model. While I see major flaws in terms of establishing this system as system secure, this may be a practical variant that is more usable.
  • The separation between action generation and visual grounding is useful. It highlights that correctly selecting a semantic action does not ensure that GUI commands execute it on the intended target.
  • The benign utility results look great.
Critical weaknesses
  • New terminology, generation integrity and grounding integrity, is introduced, however, its relationship to the existing notions of control-flow integrity (CFI) and data-flow integrity (DFI) is not sufficiently explained. It is not quite clear why these more functional security terms were introduced instead of arguing with CFI and DFI. What additional security property does this formulation establish? They seem to describe contracts for different functional stages of the system, rather than clearly distinct security properties.

  • It is unclear that generation integrity and grounding integrity establish full system security guarantees, particularly against branch steering and DFI violations. The isolated query model processes untrusted content, and its Boolean, numeric, or enumeration responses can determine branches within the committed transaction. For example, a third-party seller on a marketplace could put an injection into a correctly masked ticket listing description. When the isolated query model reads this description to answer whether the ticket is refundable, the injection could cause it to return an incorrect Boolean. The committed transaction could then select the non-refundable ticket while both integrity properties still hold: the interpreter follows a permitted branch, and grounding correctly executes the selected action.

  • It is unclear that generation integrity and grounding integrity can even be correctly established in the implementation. The paper should distinguish proving that these properties compose from demonstrating that its components satisfy them. For example, the grounding guarantee assumes accurate grounding. If the intended action is to type into the Description field, but the grounding model returns coordinates inside the Title field, both fields may pass an editability check. Which checks establish that the selected element is the intended semantic target, rather than merely a valid target for that operation? Similarly, how is it established that untrusted content cannot reach the planner through later observations or runtime feedback? Adaptive attacks should test the limits of this.

  • It is unclear that all content on a website can be clearly marked as untrusted or trusted, and that this distinction is sufficient for security. First, the paper should explain more precisely how DOM labeling translates into masking of the visually rendered website as HTML and JavaScript dynamically change the page. It describes remasking and stability checks, but their effectiveness needs evaluation. The paper should also clarify whether website vulnerabilities such as XSS are excluded: if attacker-controlled content can execute JavaScript and modify DOM elements outside the labeled region, the assumed boundary between trusted and untrusted content may no longer hold.

  • Second, a website owner’s incentives are not necessarily aligned with the user’s agent. A retailer may want to upsell a more expensive product or preselect an upgrade. Secure-CUA masks some marketing content, but what justifies trusting the remaining owner-controlled content? This is why DFI defenses introduce policy constraints on the resulting actions. However, policies need to change over state changes: they either continue holding, need to be combined with new policies, or expire. For example, permission to change a payment recipient before confirmation should not remain available afterward. This is why previous work first establishes CFI by committing to a plan ahead of time, and then addresses DFI through policies governing data flows and action parameters within that plan. The plan provides the execution context needed to determine which policies apply at each stage. In the current implementation, it is unclear how equivalent guarantees can be established: each transaction commits only to the next action, and untrusted query responses can select its branches, without an explicit specification enforcing security constraints across subsequent state transitions.

  • A major problem is that a lot of the security hinges on manually labeling the right website elements as content to be masked, and this labeling problem is underplayed. The paper assumes complete labels, but how can correctness be ensured when deployment may require labeling hundreds of websites and maintaining those labels as websites change? Fatigue can cause misses, and complicated website structures may make the origin of an element unclear. Even if an external attacker controls only a plain-text issue title or comment, the website may render it in multiple places, such as previews, notifications, or accessible button labels, any of which could be missed during labeling.

  • Say we give up on making humans manually label websites and instead use an LLM for this. We then need a separate security argument for the labeling process. First, directly exposing the labeller to untrusted content reintroduces the prompt-injection problem. Second, sanitizing the DOM before labeling, as explored by UCM (Nikolić et al.), may remove context useful for distinguishing trusted from untrusted content. At the same time, the remaining parts of the DOM may not reveal which data could be supplied by a third party. For example, an owner-written heading and a seller(third-party)-supplied title may have identical DOM structures, without their provenance being recoverable from the sanitized DOM. Automatic labeling therefore cannot simply replace manual labels while inheriting the same guarantee.

  • On that note, the contribution beyond Nikolić et al. could be made clearer. It seems currently the contribution is adding per-action transactions that constrain how query responses influence actions, as well as handling of free-form strings?

  • The overhead of this defense is surprisingly large in terms of tokens.

  • The RQs focus on benign utility, access to untrusted information, and overhead, without sufficiently evaluating whether Secure-CUA’s implementation satisfies its security assumptions under attack. The claim that the system is secure by design does not establish complete labeling, correct grounding, or safe influence from endorsed query responses. The evaluation should include the cases above, with precise attacker permissions, adaptive attacks, and ablations showing which mechanism prevents each failure. It should report both attack success and whether legitimate tasks can still be completed under attack.

Questions for authors

See Weaknesses.

Reviewer_Wk6VOpenReview ↗
Rating
2
Confidence
3
Informativeness
2
Presentation
2
Soundness
2
Read review
Paper summary

The paper argues that a CUA must be protected at two stages: action generation and visual grounding. It defines two matching properties, generation integrity and grounding integrity. It proves that enforcing both at every step protects the whole execution trace. The proposed system, Secure-CUA, masks untrusted page regions with manual CSS rules. The evaluation covers 400 WebArena tasks, 3 models and 5 seeds, all without attacks. Task success is 53.55% for Secure-CUA, 55.12% for Vanilla-CUA and 13.17% for CaMeL-CUA.

Critical strengths
  1. Splitting security into generation and grounding is a useful insight.

  2. The system is well engineered. It keeps most of Vanilla's task success and far outperforms CaMeL-CUA.

Critical weaknesses

The paper calls Secure-CUA "secure by design", but all experiments are benign. The guarantee relies on untested assumptions.

Questions for authors

How do the results change if all agents use the same grounder? And if CaMeL-CUA uses the same planner as the others?

1 comment

Thank you for the new attack results and the grounding control — I consider W2 resolved. My concern on W1 stands, however: for a paper whose central claim is security, the experimental section should be organized around demonstrating how secure the framework is, and the current evidence — a small set of known attack scenarios appended in the rebuttal — does not carry that weight. I maintain my rating.

Reviewer_5euJOpenReview ↗
Rating
4
Confidence
4
Informativeness
2
Presentation
3
Soundness
3
Read review
Paper summary

Secure-CUA is a new secure-by-design method for computer use agents (CUAs). The core idea is to create an explicit per-action program that the model must follow, before the model is given access to untrusted data. This program is called "transaction", and it puts restrictions on the queries the model can ask about untrusted content, as well how how the outputs of such queries can be processed. Secure-CUA relies on masking untrusted regions and using quarantined model to process queries. The method has by construction security, only minor ~2% drop in utility. It is a strict improvement over the previously proposed CaMeL-CUA.

Critical strengths
  • The paper makes a major step from the previous (and concurrent) secure-by-design methods by protecting not only generation integrity, i.e., not allowing untrusted content to influence CUA's semantic actions, but also grounding integrity, i.e., making sure the agent faithfully does the semantic action it was instructed to do. They proposal two methods (observation isolation a.k.a. masking) and bounded endorsement (i.e. policing untrusted data -> action influence) to enforce it. Overall, very good conceptual contribution to the field.

  • The proposed method almost does not lose utility, which is a strong achievement for security-by-design method.

Critical weaknesses

None

Questions for authors

I believe this paper is a clear accept and do not require clarifications on this work.


Our of curiosity, how would the guarantees the method provides compare to concurrent UCM? My guess is UCM has generation integrity, but not grounding integrity.

From OpenReview, CC BY 4.0. Authors as signed there; shown unedited.

Related papers

Loading the map…

Discussion (0)

Sign in to comment.