Memory Overflow Vulnerability in NVIDIA Triton Inference Server for Windows and Linux
CVE-2025-23320

7.5HIGH

Key Information:

Vendor

Nvidia
Status
Triton Inference Server
Vendor
CVE Published:
6 August 2025

Badges

đź“° News Worthy

What is CVE-2025-23320?

A memory overflow vulnerability exists in the Python backend of NVIDIA Triton Inference Server for both Windows and Linux platforms. This issue allows attackers to exploit the server by sending excessively large requests, potentially causing the shared memory limits to be exceeded. If successfully exploited, the vulnerability could lead to sensitive information being disclosed, compromising the security and integrity of the data processed by the server.

Affected Version(s)

Triton Inference Server Windows All versions prior to 25.07

News Articles

Nvidia Patches Critical RCE Vulnerability Chain

The flaws in the company's Triton Inference Server enables model theft, data leaks, and response manipulation.

2 days ago

References

https://nvd.nist.gov/vuln/detail/CVE-2025-23320
https://www.cve.org/CVERecord?id=CVE-2025-23320
https://nvidia.custhelp.com/app/answers/detail/a_id/5687

CVSS V3.1

Score:
7.5
Severity:
HIGH
Confidentiality:
High
Integrity:
None
Availability:
High
Attack Vector:
Network
Attack Complexity:
Low
Privileges Required:
None
User Interaction:
None
Scope:
Unchanged

Timeline

  • Vulnerability published

  • đź“°

    First article discovered

  • Vulnerability Reserved

.
CVE-2025-23320 : Memory Overflow Vulnerability in NVIDIA Triton Inference Server for Windows and Linux