What is the CVSS score of CVE-2026-24210?

CVE-2026-24210 has a CVSS 3.1 base score of 7.5 (High). CVSS vector: CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H. EPSS exploitation probability: 0.0%.

Which versions of NVIDIA Triton Inference Server are affected by CVE-2026-24210?

NVIDIA Triton Inference Server contains a remotely exploitable denial-of-service vulnerability requiring no authentication, creating availability risk for organizations running inference workloads.

How to fix or mitigate CVE-2026-24210?

Within 24 hours: Identify and inventory all NVIDIA Triton deployments, including version numbers and network exposure. Within 7 days: Implement network segmentation to restrict Triton access to trusted networks and disable internet-facing endpoints where possible. Within 30 days: Establish continuous monitoring for vendor advisories; evaluate rate limiting and access control configurations; prepare update procedures for rapid patching when vendor releases a fix.

NVIDIA Triton Inference Server CVE-2026-24210

EUVD-2026-31050 HIGH

Integer Overflow or Wraparound (CWE-190)

2026-05-20 nvidia

GHSA-m589-x4wc-g6m3

Denial Of Service Integer Overflow Triton Inference Server Nvidia

7.5

CVSS 3.1 · Vendor: nvidia

Severity by source

Vendor (nvidia) PRIMARY

7.5 HIGH

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

Primary rating from Vendor (nvidia) · only source for this CVE.

CVSS VectorVendor: nvidia

Attack Vector

Network

Attack Complexity

Low

Privileges Required

None

User Interaction

None

Scope

Unchanged

Confidentiality

None

Integrity

None

Availability

High

Lifecycle Timeline

Analysis Generated

May 20, 2026 - 04:02 vuln.today

DescriptionCVE.org

NVIDIA Triton Inference Server contains a vulnerability where an attacker could cause an integer overflow. A successful exploit of this vulnerability might lead to denial of service.

AnalysisAI

Denial of service in NVIDIA Triton Inference Server can be triggered remotely by unauthenticated attackers via an integer overflow condition (CWE-190). The CVSS 7.5 score reflects high availability impact with no confidentiality or integrity loss, and no public exploit has been identified at time of analysis. Defenders running Triton in network-exposed inference deployments should prioritize patching since exploitation requires no privileges, no user interaction, and low attack complexity.

Technical ContextAI

NVIDIA Triton Inference Server is an open-source serving platform used to deploy and scale AI/ML models (TensorRT, PyTorch, ONNX, TensorFlow, Python backends) across GPU and CPU infrastructure, typically exposed via HTTP/REST and gRPC inference endpoints. The root cause is CWE-190 (Integer Overflow or Wraparound), where arithmetic on attacker-influenced integer values exceeds the bounds of the target data type, producing a wrapped value that downstream code treats as valid - commonly leading to undersized allocations, malformed length checks, or out-of-bounds memory operations. In this instance the overflow manifests as a denial of service rather than memory corruption with code execution, suggesting the wrapped value drives Triton into an abort, crash, or resource-exhaustion path rather than a controllable write primitive. The affected CPE is cpe:2.3:a:nvidia:triton_inference_server with the version field wildcarded, indicating NVD has not yet narrowed down the precise vulnerable range.

RemediationAI

Patch available per vendor advisory - consult the NVIDIA security bulletin at https://nvidia.custhelp.com/app/answers/detail/a_id/5828 for the exact fixed Triton Inference Server version and upgrade accordingly; exact fix version was not provided in the input data and should not be inferred. Until patched, restrict network exposure of the Triton HTTP and gRPC inference endpoints (default ports 8000, 8001, 8002) to trusted clients only via firewall rules, service mesh policies, or Kubernetes NetworkPolicies, accepting that this breaks any cross-tenant or external inference consumers. Place an authenticating reverse proxy or API gateway in front of Triton to drop anomalous or oversized inference requests, and monitor the server for unexpected restarts that may indicate exploitation attempts; rate-limit inbound inference requests to slow brute-force crash attempts, with the trade-off of throttling legitimate high-volume workloads.