CVE-2025-23335: 
Triton Inference Server 취약성 분석 및 완화

NVIDIA Triton Inference Server for Windows and Linux and the Tensor RT backend contain a vulnerability identified as CVE-2025-23335, discovered in August 2025. The vulnerability affects all versions prior to 25.05 of the Triton Inference Server. This security issue involves an underflow condition that can be triggered by a specific model configuration and input (NVIDIA Advisory).

The vulnerability is classified as a CWE-191 (Integer Underflow) type issue. It has been assigned a CVSS v3.1 base score of 4.4 (Medium) with the vector string AV:N/AC:H/PR:H/UI:N/S:U/C:N/I:N/A:H, indicating that while the vulnerability is network accessible, it requires both high attack complexity and high privileges to exploit (NVIDIA Advisory, NVD).

A successful exploitation of this vulnerability can lead to denial of service conditions in the affected systems. The vulnerability specifically affects the Tensor RT backend of the Triton Inference Server, potentially disrupting service availability (NVIDIA Advisory).

NVIDIA has released version 25.05 of the Triton Inference Server to address this vulnerability. Users are advised to upgrade to this version or later. The fix is available for both Windows and Linux platforms. Users deploying NVIDIA Triton Inference Server in production settings should follow the Secure Deployment Considerations Guide (NVIDIA Advisory).

Security researchers and organizations have acknowledged the vulnerability's disclosure, with ASEC including it in their security advisories. The medium severity rating and specific exploitation requirements have resulted in measured industry response (ASEC).