Formal Security Analysis of Agent Protocol Composition

Shenghan Zheng, Qifan Zhang, Zheng Zhang, Haonan Li, Christophe Hauser · 2026

Read Formal Security Analysis of Agent Protocol Composition by Shenghan Zheng, Qifan Zhang, Zheng Zhang, Haonan Li, Christophe Hauser (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-28690.txt.

Description

Shenghan Zheng, Qifan Zhang, Zheng Zhang, Haonan Li, and Christophe Hauser audit the protocols that let artificial intelligence agents use tools and hand work to one another. These protocols state their security requirements in part and enforce them unevenly, and nobody had traced a requirement from the specification text all the way into running software.

Their AgentThread framework links sources end to end. It turns protocol-derived checks into logical invariants, compiles the specs into model-checkable form, and replays counterexamples against real development kits through adapters. Across five emerging protocols it logged thirty-five spec-level findings, backed by eighty code tests. It also found thirty more failures that appear only when protocols are composed. Only one protocol enforced a security-relevant control in practice, and none took charge of cross-protocol behavior, a responsibility gap the authors call the core problem.

Catalog record

CREATORShenghan Zheng, Qifan Zhang, Zheng Zhang, Haonan Li, Christophe Hauser
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE56 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.28690

Browse the full library · About the Tower of Babel