NVIDIA Triton Inference Server Remote Code Execution Vulnerability in Python Backend

A remote code execution vulnerability has been identified in NVIDIA Triton Inference Server for Windows and Linux. This issue arises in the Python backend, where an attacker can manipulate the model name parameter in the model control APIs to execute arbitrary code. Exploitation of this vulnerability could also result in a denial of service, unauthorized information disclosure, and data tampering.

Impact

Successful exploitation allows for remote code execution, with additional risks of causing a denial of service, disclosing sensitive information, and tampering with data.

Remediation

Users are advised to update to version 25.08, available on the Triton Inference Server Releases page on GitHub. For those in production environments, consult the Secure Deployment Considerations Guide.