NVIDIA Display Driver EUVD-2026-90167

Exploitation requires the attacker to already possess privileged local access (PR:H - administrator/root or equivalent) on a system running a vulnerable NVIDIA GPU Display Driver, and to interact with the driver's secure microcontroller component to abuse an incorrect permission assignment (CWE-732) against protected memory. High attack complexity (AC:H) further limits exploitation, implying a race, specific timing, or non-trivial state that must be met rather than a deterministic one-shot trigger. No user interaction is required (UI:N). Affected are GeForce (only Maxwell, Volta, and Pascal architecture GPUs), RTX/Quadro/NVS, and Tesla driver branches below the fixed versions listed in NVIDIA bulletin 5861 (e.g., GeForce < 582.78/610.60/615.71.09/616.56 depending on branch; Tesla and RTX/Quadro/NVS various branch-specific fixes). The privileged-access and high-complexity prerequisites are the primary factors limiting realistic risk.

This is a genuine but lower-tier priority. The CVSS 3.1 vector (AV:L/AC:H/PR:H/UI:N/S:U/C:H/I:H/A:L) describes a locally-exploitable flaw that requires an attacker to already hold HIGH privileges (PR:H) on the host AND overcome HIGH attack complexity (AC:H) - two significant gates that sharply limit real-world exploitability despite the C:H/I:H code-execution impact. The 6.0 base score is accurate and notably lower than the PR:L NVIDIA memory-corruption siblings (7.8) because of the PR:H + AC:H combination, and its A:L (rather than the A:H seen across most sibling CVEs) further caps it. EPSS is not provided, there is NO CISA KEV listing, and NO public exploit is identified at time of analysis, so there is no external signal of active or imminent exploitation. Signals are consistent - there is no network-vs-local conflict here; the vector and description both agree this is a privileged-local issue. Note the input 'RCE' tag overstates reach: with AV:L this is local code execution, not remote. Bottom line: a real vulnerability warranting patching in scheduled maintenance, but not an emergency given the privileged-access and high-complexity prerequisites.

An attacker who has already gained privileged (administrator/root) local access to a workstation or GPU compute host performs a difficult-to-win sequence against the NVIDIA GPU driver's secure microcontroller component, abusing an incorrect permission assignment to write to protected memory that should be off-limits. Winning the race or condition allows them to execute code within the secured microcontroller context and escalate privileges further into the driver/firmware trust boundary. No public proof-of-concept exists, and the high attack complexity means reliable exploitation is non-trivial.