AgentMark: Utility-Preserving Behavioral Watermarking for Agents
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913058737094656 |
|---|---|
| author | Huang, Kaibo Tan, Jin Wei, Yukun Li, Wanling Zhang, Zipei Tian, Hui Yang, Zhongliang Zhou, Linna |
| author_facet | Huang, Kaibo Tan, Jin Wei, Yukun Li, Wanling Zhang, Zipei Tian, Hui Yang, Zhongliang Zhou, Linna |
| contents | LLM-based agents are increasingly deployed to autonomously solve complex tasks, raising urgent needs for IP protection and regulatory provenance. While content watermarking effectively attributes LLM-generated outputs, it fails to directly identify the high-level planning behaviors (e.g., tool and subgoal choices) that govern multi-step execution. Critically, watermarking at the planning-behavior layer faces unique challenges: minor distributional deviations in decision-making can compound during long-term agent operation, degrading utility, and many agents operate as black boxes that are difficult to intervene in directly. To bridge this gap, we propose AgentMark, a behavioral watermarking framework that embeds multi-bit identifiers into planning decisions while preserving utility. It operates by eliciting an explicit behavior distribution from the agent and applying distribution-preserving conditional sampling, enabling deployment under black-box APIs while remaining compatible with action-layer content watermarking. Experiments across embodied, tool-use, and social environments demonstrate practical multi-bit capacity, robust recovery from partial logs, and utility preservation. The code is available at https://github.com/Tooooa/AgentMark. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_03294 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | AgentMark: Utility-Preserving Behavioral Watermarking for Agents Huang, Kaibo Tan, Jin Wei, Yukun Li, Wanling Zhang, Zipei Tian, Hui Yang, Zhongliang Zhou, Linna Cryptography and Security Artificial Intelligence LLM-based agents are increasingly deployed to autonomously solve complex tasks, raising urgent needs for IP protection and regulatory provenance. While content watermarking effectively attributes LLM-generated outputs, it fails to directly identify the high-level planning behaviors (e.g., tool and subgoal choices) that govern multi-step execution. Critically, watermarking at the planning-behavior layer faces unique challenges: minor distributional deviations in decision-making can compound during long-term agent operation, degrading utility, and many agents operate as black boxes that are difficult to intervene in directly. To bridge this gap, we propose AgentMark, a behavioral watermarking framework that embeds multi-bit identifiers into planning decisions while preserving utility. It operates by eliciting an explicit behavior distribution from the agent and applying distribution-preserving conditional sampling, enabling deployment under black-box APIs while remaining compatible with action-layer content watermarking. Experiments across embodied, tool-use, and social environments demonstrate practical multi-bit capacity, robust recovery from partial logs, and utility preservation. The code is available at https://github.com/Tooooa/AgentMark. |
| title | AgentMark: Utility-Preserving Behavioral Watermarking for Agents |
| topic | Cryptography and Security Artificial Intelligence |
| url | https://arxiv.org/abs/2601.03294 |