NVIDIA Display Driver EUVD-2026-90145

Requires the attacker to already have local access with a low-privilege authenticated account (PR:L) on a system running the NVIDIA GPU Display Driver kernel-mode component on Windows or Linux; no user interaction is needed. Exploitation is limited to local vectors only (AV:L) - there is no remote or network path - so the attacker must first obtain code execution or an interactive session on the GPU host. Impact is bounded to denial of service and information disclosure (no integrity/write primitive), so this cannot by itself yield privilege escalation or code execution. Multi-user, VDI, and cloud-GPU deployments where untrusted users share a GPU host are the most exposed configurations.

The CVSS 3.1 vector (AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:H, 7.1) describes a local, low-complexity attack requiring only low-privilege (authenticated) access with no user interaction, yielding high confidentiality and availability impact but no integrity impact - consistent with an untrusted-pointer dereference that can crash the driver (DoS) or leak kernel memory (info disclosure) without granting write/code-execution primitives. There is no EPSS score, no CISA KEV listing, and no public POC in the provided data, so exploitation status is 'no public exploit identified at time of analysis.' This is NVIDIA-reported (bulletin 5861) with a patch available, indicating an internally/coordinated-discovered defect rather than opportunistic exploitation. Realistically this is a MEDIUM-priority local privilege-adjacent bug: the local-only vector (AV:L) sharply limits mass exploitability, and the absence of integrity impact plus no code-execution claim (unlike sibling UAF/type-confusion CVEs in the same bulletin) makes it lower-value than the 7.8 RCE-adjacent NVIDIA siblings. No signals conflict; the vector and description are mutually consistent (DoS↔A:H, info disclosure↔C:H, no writeable primitive↔I:N).

A local low-privileged user on a Windows or Linux workstation or a multi-tenant/VDI/cloud-GPU host issues crafted calls to the NVIDIA kernel-mode GPU driver that cause it to dereference an attacker-influenced untrusted pointer. This either crashes the driver/system (denial of service) or reads out-of-bounds kernel memory that is disclosed back to the attacker. No public exploit code is identified, and the local attack vector requires the attacker to already have code execution on the host.