NVIDIA Triton Inference Server Python Backend Shared Memory Limit Vulnerability Allowing Information Disclosure

A vulnerability exists in the Python backend of NVIDIA Triton Inference Server for Windows and Linux. This issue allows an attacker to exceed the shared memory limit by sending an excessively large request. Exploiting this vulnerability could lead to unauthorized information disclosure.

Impact

Successful exploitation may result in the unintentional release of sensitive information.

Remediation

Users can update to NVIDIA Triton Inference Server version 25.07 or later. For guidance on secure deployment, refer to the NVIDIA Triton Inference Server Secure Deployment Considerations Guide.