PatchSiren

PatchSiren cyber security CVE debrief

CVE-2026-24213 NVIDIA CVE debrief

CVE-2026-24213 is a high-severity vulnerability in NVIDIA Triton Inference Server’s DALI backend. According to the official records, an attacker could trigger an out-of-bounds read, and successful exploitation might result in code execution, data tampering, denial of service, or information disclosure. The NVD record marks Triton Inference Server versions before 26.03 as affected.

Vendor
NVIDIA
Product
Triton Inference Server
CVSS
HIGH 8
CISA KEV
Not listed in stored evidence
Original CVE published
2026-05-20
Original CVE updated
2026-05-20
Advisory published
2026-05-20
Advisory updated
2026-05-20

Who should care

Operators and security teams running NVIDIA Triton Inference Server, especially environments that use the DALI backend or expose model-serving endpoints to untrusted users. This is most relevant where inference services handle externally supplied inputs or are reachable from less-trusted networks.

Technical summary

The official vulnerability data identifies CWE-125 (out-of-bounds read) in the DALI backend of NVIDIA Triton Inference Server. The NVD CVSS vector is AV:N/AC:L/PR:L/UI:R/S:U/C:H/I:H/A:H, indicating network reachability, low attack complexity, and potential high impact. The affected CPE in the supplied record is nvidia:triton_inference_server with vulnerable versions ending before 26.03.

Defensive priority

High. The issue is network-reachable and rated CVSS 8.0 (HIGH) in the supplied record, with potential for broad impact across confidentiality, integrity, and availability. Prioritize inventorying Triton deployments and moving to a fixed version or vendor-recommended mitigation as soon as possible.

Recommended defensive actions

  • Inventory all NVIDIA Triton Inference Server deployments and determine whether the DALI backend is enabled or exposed.
  • Compare installed Triton versions against the affected range in the official record; versions before 26.03 are marked vulnerable in the supplied NVD data.
  • Apply the vendor-recommended update or mitigation referenced in NVIDIA’s advisory.
  • Limit network exposure of inference endpoints to trusted clients and enforce authentication and access controls where possible.
  • Monitor for anomalous crashes, unexpected output corruption, or unusual data access patterns in Triton logs and surrounding telemetry.

Evidence notes

This debrief is based on the supplied official vulnerability records only: the NVD entry for CVE-2026-24213, the CVE.org record, and NVIDIA’s referenced vendor advisory link. The supplied NVD metadata identifies the vulnerability as CWE-125, lists the CVSS vector AV:N/AC:L/PR:L/UI:R/S:U/C:H/I:H/A:H, and marks Triton Inference Server versions before 26.03 as affected. No exploit details or additional mitigation text were provided in the source corpus.

Official resources

Publicly disclosed on 2026-05-20 in the official CVE/NVD records supplied here, with the record modified later the same day. No KEV listing was present in the provided enrichment data.