The vulnerability is a stack-based buffer overflow (CWE-121) that allows an attacker to trigger a denial of service in NVIDIA Triton Inference Server for Linux. A successful exploit would overwrite return addresses on the stack, causing the process to crash and rendering the inference service unavailable. The impact is loss of availability, affecting any applications that rely on continuous model inference.

NVIDIA Triton Inference Server for Linux is the affected product. Vendor and product information is provided, but the data does not list specific versions; therefore all deployed versions may be impacted until a patched release is available.

The CVSS score of 7.5 indicates a high severity potential. The EPSS score of less than 1% suggests that, as of now, exploitation probability is very low, and the vulnerability is not listed in the CISA KEV catalog. The likely attack vector is remote, where an attacker sends specially crafted requests to the inference endpoints to trigger the buffer overflow. If exploited, the denial of service could affect service continuity and potentially cascade to dependent systems in a deployment network.