Metraipolis › archive › dc_2beb81bfc720
Inference-Engine Fingerprinting Attacks are Practical: Exploring Model-Driven Environmental Discovery, Exploitation, and Escape — arXiv:2609.20614
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.20614
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-10T03:24:22.058Z
- SHA-256 of the archived text
b803b3de10fc19b17dc6a3f4126b270ad61fd921e66282c7d704d3e723009d32- Stable URL of this page
https://metraipolis.com/documento/dc_2beb81bfc720
How to cite
Inference-Engine Fingerprinting Attacks are Practical: Exploring Model-Driven Environmental Discovery, Exploitation, and Escape — arXiv:2609.20614 — arXiv (en), https://arxiv.org/abs/2609.20614 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2beb81bfc720 (Metraipolis archive dc_2beb81bfc720, sha256 b803b3de10fc19b17dc6a3f4126b270ad61fd921e66282c7d704d3e723009d32)
BibTeX
@misc{dc2beb81bfc720,
title = {Inference-Engine Fingerprinting Attacks are Practical: Exploring Model-Driven Environmental Discovery, Exploitation, and Escape — arXiv:2609.20614},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.20614},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2beb81bfc720; sha256 b803b3de10fc19b17dc6a3f4126b270ad61fd921e66282c7d704d3e723009d32},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.20614. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Frontier AI models are rapidly gaining the ability to exploit vulnerabilities in complex pieces of software. The risk is not theoretical, as evidenced by recent sandbox escapes performed by frontier models at OpenAI and Anthropic. Discussions of how to sandbox inference stack components often focus on components other than the inference engine itself (e.g., network proxies or code execution environments). However, the inference engine is an attractive target for a misaligned model. For example, if a model can trigger exploits in that engine merely by generating specially-crafted output tokens, the model can initiate a multi-step, to-the-bare-metal exploit chain in the engine, without relying on vulnerabilities in other components of the inference stack, and without assistance from externally-provided, maliciously-crafted input tokens. In this paper, we show that a misaligned model can perform inference engine fingerprinting to determine the specific engine (e.g., vLLM, SGLang) which executes the model. Once the engine has been fingerprinted, the model can leverage engine-specific exploits to take control of the engine using only carefully-selected output tokens. We provide concrete examples of model fingerprints in five popular engines, and demonstrate how realistic agentic harnesses allow a model to leverage those fingerprints to identify the local engine. We also describe a proof-of-concept, to-the-bare-metal exploit chain that originates from a fingerprinted (and subsequently compromised) inference engine. We conclude by discussing several ways that inference engines could be changed to make fingerprinting attacks more difficult. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-18 in cs.AI). Categorie arXiv: Cryptography and Security (cs.CR) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2609.20614 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-10T03:04:13.974Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-18?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).