What is the CVSS score of CVE-2026-24189?

CVE-2026-24189 has a CVSS 3.1 base score of 8.2 (High). CVSS vector: CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:N/A:H. EPSS exploitation probability: 0.0%.

Which versions of NVIDIA CUDA-Q are affected by CVE-2026-24189?

NVIDIA CUDA-Q contains a critical remote denial-of-service and memory disclosure vulnerability (CVE-2026-24189, CVSS 8.2) affecting internet-facing deployments without authentication requirements.

How to fix or mitigate CVE-2026-24189?

Within 24 hours: Identify all internet-facing CUDA-Q endpoints using network scanning and asset inventory tools; restrict network access via firewall rules to trusted IP ranges only. Within 7 days: Implement network segmentation to isolate CUDA-Q services from untrusted networks; establish monitoring for malformed requests to affected endpoints. Within 30 days: Contact NVIDIA support for patch timeline; evaluate alternative GPU compute frameworks if remediation timeline is unacceptable; conduct memory forensics on previously exposed systems.

NVIDIA CUDA-Q CVE-2026-24189

EUVD-2026-24149 HIGH

Out-of-bounds Read (CWE-125)

2026-04-21 nvidia

GHSA-gw6p-c344-pv6p

Denial Of Service Information Disclosure Nvidia Buffer Overflow

8.2

CVSS 3.1 · NVD

Severity by source

NVD PRIMARY

8.2 HIGH

AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:N/A:H

Primary rating from NVD · only source for this CVE.

CVSS VectorNVD

CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:N/A:H

Attack Vector

Network

Attack Complexity

Low

Privileges Required

None

User Interaction

None

Scope

Unchanged

Confidentiality

Low

Integrity

None

Availability

High

Lifecycle Timeline

Re-analysis Queued

Apr 21, 2026 - 17:22 vuln.today

cvss_changed

Analysis Generated

Apr 21, 2026 - 17:00 vuln.today

EUVD ID Assigned

Apr 21, 2026 - 16:30 euvd

EUVD-2026-24149

Analysis Generated

Apr 21, 2026 - 16:30 vuln.today

CVE Published

Apr 21, 2026 - 16:17 nvd

HIGH 8.2

DescriptionCVE.org

NVIDIA CUDA-Q contains a vulnerability in an endpoint, where an unauthenticated attacker could cause an out-of-bounds read by sending a maliciously crafted request. A successful exploit of this vulnerability might lead to denial of service and information disclosure.

AnalysisAI

Out-of-bounds read in NVIDIA CUDA-Q endpoint allows remote unauthenticated attackers to crash services and disclose sensitive memory contents via malformed network requests. The vulnerability affects an exposed network endpoint with no authentication barrier (CVSS AV:N/AC:L/PR:N/UI:N), enabling trivial exploitation against internet-facing deployments. EPSS data not available; no CISA KEV listing or public POC identified at time of analysis, suggesting exploitation remains theoretical or limited to targeted scenarios.

Technical ContextAI

CUDA-Q is NVIDIA's platform for quantum-classical hybrid computing workloads. The vulnerability stems from a CWE-125 out-of-bounds read in a network-facing endpoint's request parsing logic. When processing input parameters, the code reads memory beyond allocated buffer boundaries without proper bounds validation. This classic memory corruption issue allows attackers to leak heap or stack contents through error responses or timing channels, while triggering crashes via access violations. The CPE string (cpe:2.3:a:nvidia:cuda-q:*:*:*:*:*:*:*:*) indicates all versions may be affected pending vendor clarification. Out-of-bounds reads typically occur in C/C++ code handling serialized data formats (JSON, Protocol Buffers, custom binary protocols) common in quantum computing frameworks' REST APIs or RPC interfaces.

RemediationAI

Apply the vendor-issued security update per NVIDIA Security Bulletin 5820 available at https://nvidia.custhelp.com/app/answers/detail/a_id/5820, which should specify patched CUDA-Q versions (exact version not provided in available data). Until patching, implement network-level access controls: restrict CUDA-Q endpoint access to trusted IP ranges via firewall rules (blocks remote exploitation but prevents legitimate external access to quantum computing services); deploy a reverse proxy with request size limits and input validation (adds latency and complexity, may break legitimate large payloads); or disable the vulnerable endpoint entirely if not required for operations (eliminates attack surface but may impact quantum workload submission workflows). For air-gapped quantum computing environments, risk is minimal and patching can follow normal maintenance windows. Organizations exposing CUDA-Q to research collaborators should prioritize immediate patching given the unauthenticated attack vector.