NVIDIA Display Driver CVE-2026-47548

Requires local access to a system with the NVIDIA GPU Display Driver (or Virtual GPU Manager) installed and the attacker already holding low-level privileges (PR:L) - e.g., an authenticated local user, a compromised low-privileged process, or code running inside a sandbox with driver access. No user interaction is needed (UI:N) and the flaw is reachable against default driver installations across the affected GeForce, RTX/Quadro/NVS, Tesla, and vGPU product lines on both Windows and Linux. The principal limiting factor is that exploitation is not remote: the attacker must first obtain code execution on the target host, and successful escalation requires developing a reliable kernel memory-corruption exploit; it cannot be triggered over the network or by a purely remote unauthenticated actor.

This is a locally-exploitable kernel-mode out-of-bounds write (CWE-787) in the NVIDIA GPU Display Driver, with a vendor/NVD CVSS 3.1 vector of AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H (7.8, High). The signals are internally consistent: local attack vector, low complexity, and only low privileges required (an existing local user or sandboxed process), with no user interaction - a textbook local privilege-escalation primitive where a memory-corruption bug in a kernel driver yields SYSTEM/root. No EPSS score, CISA KEV listing, or public POC is present in the supplied data, so there is no public exploit identified at time of analysis and no evidence of active exploitation; exploitation would currently require an attacker to develop a working kernel exploit, which is non-trivial. Real-world priority is moderate-to-high rather than emergency: the local-only vector means it is not remotely wormable and is not an initial-access bug, but on the enormous installed base of NVIDIA drivers it is a valuable second-stage escalation and sandbox-escape tool, so it should be patched on a standard prioritized cadence. No key signals conflict - the CVSS local vector matches the 'kernel mode layer' description; the main gap is the absence of exploitability telemetry (EPSS/KEV/POC all missing), so risk is inferred from the impact profile rather than observed exploitation.

An attacker who already has a low-privileged foothold on a Windows or Linux host with NVIDIA drivers installed (e.g., a standard user, a compromised service account, or code escaped into a restricted sandbox) issues crafted requests/ioctls to the vulnerable kernel-mode driver interface to trigger the out-of-bounds write. By shaping adjacent kernel memory, they corrupt kernel structures to execute code at the kernel privilege level, escalating to SYSTEM/root or crashing the system for denial of service. No public exploit code is identified in the available data, so weaponization would require custom kernel-exploit development.