NVIDIA Display Driver CVE-2026-47500

Requires the attacker to already have local code execution on a machine running the NVIDIA GPU Display Driver (kernel mode layer) - i.e., an authenticated low-privileged local user, sandboxed process, or, per the 'Guest Driver'/'Virtual GPU Manager' tags, code inside a vGPU guest VM. The trigger is reaching a driver error path where reference counts are cleaned up improperly, producing the use-after-free; no user interaction and no elevated privileges are needed (AV:L/AC:L/PR:L/UI:N). Limiting factors: exploitation is NOT remote (local vector only), it needs a vulnerable NVIDIA driver version installed, and reliable UAF exploitation of kernel memory typically requires heap-grooming skill even though CVSS rates complexity low. Affected families are GeForce, RTX/Quadro/NVS, and Tesla drivers on Windows and Linux, including vGPU deployments.

This is a genuine, high-value local privilege-escalation issue, though not an emergency. The CVSS 3.1 vector (AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H, 7.8 HIGH) describes a use-after-free (CWE-416) in the NVIDIA kernel-mode driver reachable by an already-authenticated local user with low privileges, requiring no user interaction and low attack complexity, yielding full C/I/A compromise of the host - the classic profile of a kernel driver LPE. No EPSS score, no CISA KEV listing, and no public POC were provided in the input, so exploitation status is 'no public exploit identified at time of analysis'; however, absence of these signals is not evidence of low real-world risk, because kernel UAFs in the ubiquitous NVIDIA driver are historically attractive sandbox-escape/EoP primitives and NVIDIA (the reporter) rates it 7.8. No SSVC decision framework data was supplied. There are no conflicting signals in the core vector, but note the description also enumerates the vGPU path ('Guest Driver', 'Virtual Gpu Manager' tags): in a virtual-GPU deployment a malicious guest could reach the host hypervisor layer, which the S:U (unchanged scope) base vector does not fully capture - verify the vGPU-specific vector against NVIDIA's advisory. Net: a real patching priority for workstation/gaming/datacenter fleets, but driven by exposure and impact rather than by confirmed in-the-wild activity.

A local attacker with an unprivileged account (or code running inside a sandbox or a vGPU guest) issues a crafted sequence of GPU driver calls that forces an error path; improper reference-count cleanup leaves a dangling kernel object, and the attacker reclaims and reuses that freed memory to corrupt kernel state. This yields code execution in the kernel and escalation to SYSTEM/root. No public exploit code was identified at time of analysis, but the low complexity and no-interaction requirement make weaponization plausible for a skilled attacker.