Are GUI Agents Focused Enough? Automated Distraction via Semantic-level UI Element Injection

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Yang, Wenkui, Jin, Chao, Zhu, Haisu, Luo, Weilin, Yuen, Derek, Shao, Kun, Huang, Huaibo, Duan, Junxian, Cao, Jie, He, Ran
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866908949215707136
author Yang, Wenkui
Jin, Chao
Zhu, Haisu
Luo, Weilin
Yuen, Derek
Shao, Kun
Huang, Huaibo
Duan, Junxian
Cao, Jie
He, Ran
author_facet Yang, Wenkui
Jin, Chao
Zhu, Haisu
Luo, Weilin
Yuen, Derek
Shao, Kun
Huang, Huaibo
Duan, Junxian
Cao, Jie
He, Ran
contents Existing red-teaming studies on GUI agents have important limitations. Adversarial perturbations typically require white-box access, which is unavailable for commercial systems, while prompt injection is increasingly mitigated by stronger safety alignment. To study robustness under a more practical threat model, we propose Semantic-level UI Element Injection, a red-teaming setting that overlays safety-aligned and harmless UI elements onto screenshots to misdirect the agent's visual grounding. Our method uses a modular Editor-Overlapper-Victim pipeline and an iterative search procedure that samples multiple candidate edits, keeps the best cumulative overlay, and adapts future prompt strategies based on previous failures. Across five victim models, our optimized attacks improve attack success rate by up to 4.4x over random injection on the strongest victims. Moreover, elements optimized on one source model transfer effectively to other target models, indicating model-agnostic vulnerabilities. After the first successful attack, the victim still clicks the attacker-controlled element in more than 15% of later independent trials, versus below 1% for random injection, showing that the injected element acts as a persistent attractor rather than simple visual clutter.
format Preprint
id arxiv_https___arxiv_org_abs_2604_07831
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Are GUI Agents Focused Enough? Automated Distraction via Semantic-level UI Element Injection
Yang, Wenkui
Jin, Chao
Zhu, Haisu
Luo, Weilin
Yuen, Derek
Shao, Kun
Huang, Huaibo
Duan, Junxian
Cao, Jie
He, Ran
Cryptography and Security
Computation and Language
Computer Vision and Pattern Recognition
Existing red-teaming studies on GUI agents have important limitations. Adversarial perturbations typically require white-box access, which is unavailable for commercial systems, while prompt injection is increasingly mitigated by stronger safety alignment. To study robustness under a more practical threat model, we propose Semantic-level UI Element Injection, a red-teaming setting that overlays safety-aligned and harmless UI elements onto screenshots to misdirect the agent's visual grounding. Our method uses a modular Editor-Overlapper-Victim pipeline and an iterative search procedure that samples multiple candidate edits, keeps the best cumulative overlay, and adapts future prompt strategies based on previous failures. Across five victim models, our optimized attacks improve attack success rate by up to 4.4x over random injection on the strongest victims. Moreover, elements optimized on one source model transfer effectively to other target models, indicating model-agnostic vulnerabilities. After the first successful attack, the victim still clicks the attacker-controlled element in more than 15% of later independent trials, versus below 1% for random injection, showing that the injected element acts as a persistent attractor rather than simple visual clutter.
title Are GUI Agents Focused Enough? Automated Distraction via Semantic-level UI Element Injection
topic Cryptography and Security
Computation and Language
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2604.07831