LLM-Assisted Automatic Security Proofs for Cryptographic Protocols: How Far Are We?
LLMs produce useful cryptographic-protocol lemmas but average only 38.82% proof-tree coverage.
A cs.CR paper reports the first systematic evaluation of state-of-the-art LLMs on cryptographic symbolic protocol verification. The authors introduce CRoST, Coverage Rate of Solve Tree, which scores generated lemmas against the verifier's proof skeleton, and justify it theoretically and empirically. Models average 38.82% coverage, with 14.4% of lemmas above 80%, but complex multi-phase protocols still fail, naive scaling shows diminishing returns, and verification overhead is substantial.
- CRoST compares generated lemmas with reference proof lemmas.
- State-of-the-art models average 38.82% coverage.
- 14.4% of generated lemmas exceed 80% coverage.
- Complex multi-phase protocols still produce non-trivial failures.
From the original article
1 min read at arxiv.orgLarge language models (LLMs) have shown strong potential for assisting software and security analysis tasks, yet their effectiveness in cryptographic symbolic protocol verification remains insufficiently understood. In this paper, we conduct the first systematic evaluation of the capability of state-of-the-art LLMs in cryptographic symbolic protocol verification.
By Tianjian Liu · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.