The vulnerability in ultravioletrs cocos affects the attested TLS (aTLS) AMD SEV-SNP verification path used in confidential computing for AI workloads. Specifically, when the expected reportData value is nil, empty, or omitted, the SEV-SNP policy ReportData is unset, causing the verifier to accept stale or unrelated attestation evidence not bound to the current connection. This flaw allows a relying party that uses this verification path without an expected reportData to be induced to trust an unintended attestation context. If a non-empty reportData is supplied, validation occurs correctly. The flaw is addressed in version 0.9.0.

An attacker can cause a relying party to trust stale or unrelated attestation evidence, potentially leading to unauthorized trust or authorization decisions. This compromises the integrity of the attestation process in confidential computing environments, risking unauthorized access or misuse of AI workloads running inside trusted execution environments. The vulnerability has a CVSS 3.1 score of 9.1, indicating critical impact on confidentiality and integrity without affecting availability.

Upgrade to ultravioletrs cocos version 0.9.0 or later, where this vulnerability is fixed. The vendor advisory confirms the issue is resolved in version 0.9.0. No other mitigation steps are specified or required.