NVIDIA Triton Inference Server Model Loading API Integer Overflow Vulnerability Leading to Denial-of-Service

A vulnerability exists in the model loading API of NVIDIA Triton Inference Server, where a user could induce an integer overflow or wraparound error by uploading a model with an excessively large file size. This oversized file can overflow an internal variable, potentially leading to a denial-of-service condition.

Impact

Exploitation of this vulnerability can cause a denial-of-service, disrupting the availability of the Triton Inference Server.

Remediation

Users can upgrade to NVIDIA Triton Inference Server version 24.12 to address this vulnerability. The latest release can be downloaded from the Triton Inference Server Releases page on GitHub.