Intel TDX Guest CVE-2026-20765

Exploitation requires: (1) local access to the target system running Intel TDX Guest software before version 0.3.1; (2) the attacker must already possess high system-level privileges (PR:H per CVSS 4.0 vector) - described as a 'system software adversary with a privileged user'; (3) no user interaction is required (UI:N); (4) no special attack timing or environmental conditions are needed (AT:N, AC:L). What limits exploitation: the requirement for pre-existing high privileges is the critical barrier - this is not a remote or unauthenticated attack path. The vulnerability is scoped to the TDX guest environment and does not affect the host, hypervisor, or subsequent systems (SC:N/SI:N/SA:N). Systems not deploying Intel TDX Guest software or running version 0.3.1 and above are not affected.

Real-world risk is moderate-to-low despite the privilege escalation label. The CVSS 4.0 vector (AV:L/AC:L/AT:N/PR:H) demands local access and high prior privileges, meaning an attacker must already be a privileged user on the guest system - significantly narrowing the attack surface. The CVSS 4.0 score of 4.6 reflects this constrained impact (VC:L/VI:L/VA:L). No EPSS data was provided, and there is no CISA KEV listing or known public exploit, further reducing urgency. The attack complexity is low (AC:L) and no special attack timing or conditions (AT:N) are required, meaning exploitation is straightforward once the attacker holds the required privilege level - but reaching that level is itself a prerequisite barrier. The scope is strictly contained to the vulnerable guest system (SC:N/SI:N/SA:N), making lateral movement or hypervisor escape scenarios unsupported by this CVE alone. This is a real but low-priority vulnerability: environments running TDX workloads with strict privilege separation should still patch, as privilege escalation within a confidential compute guest can undermine the isolation guarantees TDX is designed to provide.

A local user who already holds high system-level privileges within an Intel TDX guest environment triggers the incorrect comparison in the Ring 3 TDX Guest software, causing the flawed logic to elevate their effective privilege level further within the guest. No public proof-of-concept has been identified, but the low attack complexity (AC:L) means exploitation requires no special timing or conditions beyond the initial privileged foothold. The impact is confined to the guest system, with no evidence this chain can be used to escape the TDX boundary into the hypervisor or host.