EUVIMEDEUVIMED entdecken
Studienprotokoll

BiasAgent: Exploiting Agent Bias for Preference Manipulation Attacks on Model Context Protocol

Ji Guo, Zhijing Wang, Wenbo Jiang, Rui Zhang, Jian Xiong, Qiyang Song, Hao Wu, Yijing Liu

IEEE Transactions on Cognitive Communications and Networking · 2026 · 12 · 10358-10371

DOI: 10.1109/tccn.2026.3714063 ↗

Kurz gefasst

Studienprotokoll in der Zeitschrift IEEE Transactions on Cognitive Communications and Networking, 2026. Verfasst von Ji Guo und 7 weiteren Autor:innen. Die Arbeit nennt 38 Literaturangaben. Gefördert von: National Key Research and Development Program of China, National Natural Science Foundation of China, Sichuan Science and Technology Program.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Model Context Protocol (MCP) has recently emerged as a unified framework for connecting Large Language Model (LLM) agents with external tools and applications, enabling more flexible interactions with their environments. However, recent studies have shown that MCP is also vulnerable to preference manipulation attacks, where an attacker deploys a customized MCP server to manipulate an LLM agent, causing it to favor that server over competing alternatives. Prior approaches rely on injecting manipulative keywords or phrases into tool names or descriptions, making them easily detectable. In this paper, we explore leveraging agent bias to directly manipulate tool selection without modifying tool descriptions or user inputs. Specifically, we construct a biased dataset that favors a designated tool and train a biased agent using Decoupled Clip and Dynamic Sampling Policy Optimization (DAPO). The resulting agent selectively prefers the attacker-specified tool for relevant tasks while maintaining normal functionality on other tasks. Experimental results demonstrate that our approach achieves high effectiveness and strong stealthiness across multiple LLM agents and task settings.

Förderung

  • National Key Research and Development Program of China
  • National Natural Science Foundation of China
  • Sichuan Science and Technology Program
  • Fundamental Research Funds for Chinese Central Universities

Zitieren

APA

Guo, J., Wang, Z., Jiang, W., Zhang, R., Xiong, J., Song, Q., Wu, H., & Liu, Y. (2026). BiasAgent: Exploiting Agent Bias for Preference Manipulation Attacks on Model Context Protocol. IEEE Transactions on Cognitive Communications and Networking, 12, 10358-10371. https://doi.org/10.1109/tccn.2026.3714063

Vancouver

Guo J, Wang Z, Jiang W, Zhang R, Xiong J, Song Q, et al. BiasAgent: Exploiting Agent Bias for Preference Manipulation Attacks on Model Context Protocol. IEEE Transactions on Cognitive Communications and Networking. 2026;12:10358-10371. doi:10.1109/tccn.2026.3714063

BibTeX
@article{Guo2026,
  author = {Guo, Ji and Wang, Zhijing and Jiang, Wenbo and Zhang, Rui and Xiong, Jian and Song, Qiyang and Wu, Hao and Liu, Yijing},
  title = {BiasAgent: Exploiting Agent Bias for Preference Manipulation Attacks on Model Context Protocol},
  journal = {IEEE Transactions on Cognitive Communications and Networking},
  year = {2026},
  volume = {12},
  pages = {10358--10371},
  doi = {10.1109/tccn.2026.3714063}
}
Kostenlos starten →