NVIDIA Triton Inference Server Integer Underflow Vulnerability Leading to Denial-of-Service

A vulnerability exists in NVIDIA Triton Inference Server for Windows and Linux, as well as the TensorRT backend, allowing an attacker to cause an integer underflow by using a specific model configuration and input. Exploitation of this vulnerability could result in a denial-of-service condition.

Impact

Exploitation of this vulnerability could lead to a denial-of-service condition, causing the server to become unresponsive or unavailable.

Remediation

Users are advised to update to version 25.05 or later. Instructions for updating can be found on the NVIDIA Triton Inference Server Releases page on GitHub.