NVIDIA Triton Inference Server Double Free Vulnerability Leading to Denial-of-Service

A double free vulnerability has been identified in NVIDIA Triton Inference Server for Windows and Linux, affecting all versions prior to 25.06. This vulnerability arises when multiple requests cause a stream to be cancelled before processing, potentially leading to a denial-of-service condition.

Impact

Exploitation of this vulnerability can cause a denial-of-service condition, disrupting the normal functioning of the server.

Remediation

Users can upgrade to version 25.06 or later to address this vulnerability. For those deploying Triton Inference Server in production, it is recommended to follow the Secure Deployment Considerations Guide available on the Triton Inference Server GitHub repository.