Exposing LLM User Privacy via Traffic Fingerprint Analysis: A Study of Privacy Risks in LLM Agent Interactions
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866916997365760000 |
|---|---|
| author | Zhang, Yixiang Deng, Xinhao Gu, Zhongyi Chen, Yihao Xu, Ke Li, Qi Wu, Jianping |
| author_facet | Zhang, Yixiang Deng, Xinhao Gu, Zhongyi Chen, Yihao Xu, Ke Li, Qi Wu, Jianping |
| contents | Large Language Models (LLMs) are increasingly deployed as agents that orchestrate tasks and integrate external tools to execute complex workflows. We demonstrate that these interactive behaviors leave distinctive fingerprints in encrypted traffic exchanged between users and LLM agents. By analyzing traffic patterns associated with agent workflows and tool invocations, adversaries can infer agent activities, distinguish specific agents, and even profile sensitive user attributes. To highlight this risk, we develop AgentPrint, which achieves an F1-score of 0.866 in agent identification and attains 73.9% and 69.1% top-3 accuracy in user attribute inference for simulated- and real-user settings, respectively. These results uncover an overlooked risk: the very interactivity that empowers LLM agents also exposes user privacy, underscoring the urgent need for technical countermeasures alongside regulatory and policy safeguards. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_07176 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Exposing LLM User Privacy via Traffic Fingerprint Analysis: A Study of Privacy Risks in LLM Agent Interactions Zhang, Yixiang Deng, Xinhao Gu, Zhongyi Chen, Yihao Xu, Ke Li, Qi Wu, Jianping Cryptography and Security Large Language Models (LLMs) are increasingly deployed as agents that orchestrate tasks and integrate external tools to execute complex workflows. We demonstrate that these interactive behaviors leave distinctive fingerprints in encrypted traffic exchanged between users and LLM agents. By analyzing traffic patterns associated with agent workflows and tool invocations, adversaries can infer agent activities, distinguish specific agents, and even profile sensitive user attributes. To highlight this risk, we develop AgentPrint, which achieves an F1-score of 0.866 in agent identification and attains 73.9% and 69.1% top-3 accuracy in user attribute inference for simulated- and real-user settings, respectively. These results uncover an overlooked risk: the very interactivity that empowers LLM agents also exposes user privacy, underscoring the urgent need for technical countermeasures alongside regulatory and policy safeguards. |
| title | Exposing LLM User Privacy via Traffic Fingerprint Analysis: A Study of Privacy Risks in LLM Agent Interactions |
| topic | Cryptography and Security |
| url | https://arxiv.org/abs/2510.07176 |