The vulnerability lies in the use of a hardcoded nonce for TPM quote attestations in Keylime's push model. An attacker with root access on a monitored machine could exploit this to perform replay attacks, effectively hiding malicious activity.
The analysis of the patch commit 5e1ac3f4d70e55cd72673ead3dcf88259798c33a reveals the exact location of the vulnerability. The commit is titled "verifier: Fix hardcoded attestation challenge nonce (CVE-2026-6420)".
The file keylime/models/verifier/evidence.py contains the vulnerable code. Specifically, the generate_challenge method within the CertificationParameters class was using a static, hardcoded byte string for the challenge: self.challenge = bytes.fromhex("49beed365aac777dae23564f5ad0ec").
The patch replaces this hardcoded value with a call to Nonce.generate(bit_length), which generates a cryptographically random nonce, thus mitigating the replay attack vector.
The function EvidenceItem.generate_challenge is the caller of the vulnerable function and is also part of the vulnerable code path. During exploitation, both EvidenceItem.generate_challenge and CertificationParameters.generate_challenge would appear in a runtime profile. Therefore, both are identified as vulnerable functions.