"Power of Words": Stealthy and Adaptive Private Information Elicitation via LLM Communication Strategies
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866912710537510912 |
|---|---|
| author | Zhang, Shuning Bai, Jiaqi Wang, Linzhi Li, Shixuan Yi, Xin Li, Hewu |
| author_facet | Zhang, Shuning Bai, Jiaqi Wang, Linzhi Li, Shixuan Yi, Xin Li, Hewu |
| contents | While communication strategies of Large Language Models (LLMs) are crucial for human-LLM interactions, they can also be weaponized to elicit private information, yet such stealthy attacks remain under-explored. This paper introduces the first adaptive attack framework for stealthy and targeted private information elicitation via communication strategies. Our framework operates in a dynamic closed-loop: it first performs real-time psychological profiling of the users' state, then adaptively selects an optimized communication strategy, and finally maintains stealthiness through prompt-based rewriting. We validated this framework through a user study (N=84), demonstrating its generalizability across 3 distinct LLMs and 3 scenarios. The targeted attacks achieved a 205.4% increase in eliciting specific targeted information compared to stealthy interactions without strategies. Even stealthy interactions without specific strategies successfully elicited private information in 54.8% cases. Notably, users not only failed to detect the manipulation but paradoxically rated the attacking chatbot as more empathetic and trustworthy. Finally, we advocate for mitigations, encouraging developers to integrate adaptive, just-in-time alerts, users to build literacy against specific manipulative tactics, and regulators to define clear ethical boundaries distinguishing benign persuasion from coercion. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_11961 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | "Power of Words": Stealthy and Adaptive Private Information Elicitation via LLM Communication Strategies Zhang, Shuning Bai, Jiaqi Wang, Linzhi Li, Shixuan Yi, Xin Li, Hewu Human-Computer Interaction While communication strategies of Large Language Models (LLMs) are crucial for human-LLM interactions, they can also be weaponized to elicit private information, yet such stealthy attacks remain under-explored. This paper introduces the first adaptive attack framework for stealthy and targeted private information elicitation via communication strategies. Our framework operates in a dynamic closed-loop: it first performs real-time psychological profiling of the users' state, then adaptively selects an optimized communication strategy, and finally maintains stealthiness through prompt-based rewriting. We validated this framework through a user study (N=84), demonstrating its generalizability across 3 distinct LLMs and 3 scenarios. The targeted attacks achieved a 205.4% increase in eliciting specific targeted information compared to stealthy interactions without strategies. Even stealthy interactions without specific strategies successfully elicited private information in 54.8% cases. Notably, users not only failed to detect the manipulation but paradoxically rated the attacking chatbot as more empathetic and trustworthy. Finally, we advocate for mitigations, encouraging developers to integrate adaptive, just-in-time alerts, users to build literacy against specific manipulative tactics, and regulators to define clear ethical boundaries distinguishing benign persuasion from coercion. |
| title | "Power of Words": Stealthy and Adaptive Private Information Elicitation via LLM Communication Strategies |
| topic | Human-Computer Interaction |
| url | https://arxiv.org/abs/2511.11961 |