BibTeX for a paper by David Kotz at Dartmouth College.
For more information about this paper, visit this web page:
https://www.cs.dartmouth.edu/~kotz/research/abdollahi-agentee/index.html

@InProceedings{abdollahi:agentee,
  author =        {Sina Abdollahi and Mohammad M Maheri and Javad Forough and Amir {Al Sadi} and Josh Millar and David Kotz and Marios Kogias and Hamed Haddadi},
  title =         {{AgenTEE: Confidential LLM Agent Execution on Edge Devices}},
  booktitle =     {{Proceedings of the Workshop on Machine Learning and Systems (EuroMLSys)}},
  year =          2026,
  month =         {April},
  pages =         {473--480},
  publisher =     {ACM},
  copyright =     {ACM},
  DOI =           {10.1145/3805621.3807660},
  URL =           {https://www.cs.dartmouth.edu/~kotz/research/abdollahi-agentee/index.html},
  abstract =      {Large Language Model (LLM) agents provide powerful automation capabilities, but they also create a substantially broader attack surface than traditional applications due to their tight integration with non-deterministic models and third-party services. While current deployments primarily rely on cloud-hosted services, emerging designs increasingly execute agents directly on edge devices to reduce latency and enhance user privacy. However, securely hosting such complex agent pipelines on edge devices remains challenging. These deployments must protect proprietary assets (e.g., system prompts and model weights) and sensitive runtime state on heterogeneous platforms that are vulnerable to software attacks and potentially controlled by malicious users.  \par  To address these challenges, we present \emph{AgenTEE}, a system for deploying confidential agent pipelines on edge devices. \emph{AgenTEE} places the agent runtime, inference engine, and third-party applications into independently attested confidential virtual machines (cVMs) and mediates their interaction through explicit, verifiable communication channels. Built on Arm Confidential Compute Architecture (CCA), a recent extension to Arm platforms, \emph{AgenTEE} enforces strong system-level isolation of sensitive assets and runtime state. Our evaluation shows that such multi-cVMs system is practical, achieving near-native performance with less than 5.15\% runtime overhead compared to commodity OS multi-process deployments.  },
}