CVE-2026-119344

Cocos AI attestation can authorize a different trusted platform

Ultraviolet · Cocos AI attested TLS

First documented disclosure · Review checkpoint

Summary

Valid attestation from another platform can satisfy a verifier that does not check the intended machine’s identity.

Severity and exploitation evidence

Critical · 9.1/10 · CVSS 4.0. Severity source.

Exploitation not established: The maintainer cites formal analysis and demonstration. No malicious exploitation or affected deployment is documented.

Exploitation evidence source

Technical severity is not a measure of observed attacks or of risk to your deployment. Attempts, demonstrations, and successful compromise retain the qualifications in the source.

How the weakness works

The verifier accepts matching firmware and workload measurements without enforcing platform ownership, unique identity or revocation.

Conditions required

The attacker controls or compromises another platform accepted by the same verifier and can substitute its valid evidence.

Potential impact

Trust can be diverted to an unintended platform despite correct workload measurements.

Affected versions and configurations

0.8.2 and earlier, according to the CNA.

Fix and mitigation status

The CNA lists 0.9.0, but the maintainer explicitly says source-code changes alone cannot fix this identity problem.

Practical next steps

Use the supported attestation implementation and explicitly authorize hardware identities or ownership evidence; review revocation and the applicable provider guarantees.

Operating lesson

Matching software measurements does not identify which machine is entitled to receive a secret.

Evidence and limits

September 25 is the reviewed Cocos-specific disclosure; general attested-TLS research predates it. The March 27 release documents a separate relay CVE. Its existence is not proof that all platform-identity risks are eliminated.

Sources