Known By Their Actions: Fingerprinting LLM Browser Agents via UI Traces
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866917495759175680 |
|---|---|
| author | Lugoloobi, William Marro, Samuelle Magomere, Jabez Wright, Joss Russell, Chris |
| author_facet | Lugoloobi, William Marro, Samuelle Magomere, Jabez Wright, Joss Russell, Chris |
| contents | As LLM-based agents increasingly browse the web on users' behalf, a natural question arises: can websites passively identify which underlying model powers an agent? Doing so would represent a significant security risk, enabling targeted attacks tailored to known model vulnerabilities. Across 14 frontier LLMs and four web environments spanning information retrieval and shopping tasks, we show that an agent's actions and interaction timings, captured via a passive JavaScript tracker, are sufficient to identify the underlying model with up to 96\% F1. We formalise this attack surface by demonstrating that classifiers trained on agent actions generalise across model sizes and families. We further show that strong classifiers can be trained from few interaction traces and that agent identity can be inferred early within an episode. Injecting randomised timing delays between actions substantially degrades classifier performance, but does not provide robust protection: a classifier retrained on delayed traces largely recovers performance. We release our harness and a labelled corpus of agent traces \href{https://github.com/KabakaWilliam/known_actions}{here}. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_14786 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Known By Their Actions: Fingerprinting LLM Browser Agents via UI Traces Lugoloobi, William Marro, Samuelle Magomere, Jabez Wright, Joss Russell, Chris Cryptography and Security Artificial Intelligence Human-Computer Interaction Machine Learning As LLM-based agents increasingly browse the web on users' behalf, a natural question arises: can websites passively identify which underlying model powers an agent? Doing so would represent a significant security risk, enabling targeted attacks tailored to known model vulnerabilities. Across 14 frontier LLMs and four web environments spanning information retrieval and shopping tasks, we show that an agent's actions and interaction timings, captured via a passive JavaScript tracker, are sufficient to identify the underlying model with up to 96\% F1. We formalise this attack surface by demonstrating that classifiers trained on agent actions generalise across model sizes and families. We further show that strong classifiers can be trained from few interaction traces and that agent identity can be inferred early within an episode. Injecting randomised timing delays between actions substantially degrades classifier performance, but does not provide robust protection: a classifier retrained on delayed traces largely recovers performance. We release our harness and a labelled corpus of agent traces \href{https://github.com/KabakaWilliam/known_actions}{here}. |
| title | Known By Their Actions: Fingerprinting LLM Browser Agents via UI Traces |
| topic | Cryptography and Security Artificial Intelligence Human-Computer Interaction Machine Learning |
| url | https://arxiv.org/abs/2605.14786 |