CVE-2026-6727

CVE-2026-6727 exposes a timing side-channel in the TPM 2.0 RSA Endorsement Key (EK) implementation, specifically in the OAEP decryption path. The vulnerability allows a privileged local attacker with chosen-ciphertext access to potentially recover the private key material through timing analysis of decryption operations. This is not a specification flaw—the TPM 2.0 spec is sound—but rather vendors implementing OAEP without constant-time arithmetic, choosing performance over timing safety in cryptographic code that anchors entire trust models. The CVSS 5.9 rating is dangerously misleading. The EK is the foundational identity substrate for TPM-based remote attestation; compromising it potentially invalidates every attestation issued by that TPM with no retrospective audit capability. The 'privileged local attacker' constraint assumes an attacker who has already reached that privilege level is uninteresting—but in cloud tenants, MDM-managed endpoints, and container escape scenarios, this is precisely the starting point for sophisticated attacks, not the goal. You cannot detect whether this timing channel was exploited against your EK; no forensic artifacts survive such attacks, and the 'certain conditions' language in the CVE masks how common those conditions are in modern infrastructure. For immediate action: inventory all systems relying on TPM attestation and identify their TPM vendor and firmware version. Contact your hardware vendors (Dell, HP, Lenovo for enterprise systems) to confirm whether your specific SKUs are affected and what firmware update timeline applies. The gap between 'patch available' and 'patched in production' for TPM firmware is measured in years for many embedded and OEM systems—this is not a vulnerability you can sprint to remediate. The harder truth: if you have systems deployed with vulnerable TPM firmware, you must assume the attestation chain for those systems may have been observed or forged during the window before you can patch. There is no interim software mitigation that preserves TPM semantics. Options are limited to accepting weakened trust (bind to software measurements), abandoning attestation entirely, or accelerating hardware replacement. The architectural lesson is that we built security models assuming perfect cryptographic implementation in trust anchors, but the pattern of RSA OAEP timing vulnerabilities across security-critical infrastructure demonstrates that assumption is repeatedly invalidated. Design your trust models with defensive depth that doesn't collapse when the hardware root of trust fails.

Reviewed through automated stages and approved by a human before publication.