An unauthenticated attacker can cause a denial of service in the NVIDIA Triton Inference Server by triggering a memory allocation error with an invalid request.

The software fails to properly validate the size of memory allocations when processing requests, which is classified as CWE-789. An unauthenticated attacker can exploit this by sending a crafted, invalid request to the server, resulting in a segmentation fault and subsequent service disruption.

Successful exploitation of this vulnerability leads to a denial of service, which can cause significant operational disruption for environments relying on the Triton Inference Server for model deployment. With a CVSS score of 7.5, this high severity flaw is particularly concerning because the attack vector is network-based and requires no authentication, making it easily automatable by remote adversaries.

Immediate Action: Update the NVIDIA Triton Inference Server to version 25.06 or later to incorporate the necessary memory allocation validation fixes.

Proactive Monitoring: Monitor server logs for frequent segmentation faults or abnormal memory consumption patterns that may indicate attempts to trigger this memory allocation flaw.

Compensating Controls: Deploy a Web Application Firewall (WAF) or an API gateway to inspect incoming traffic and block malformed or unusually large requests before they reach the inference server.

Public Exploit Available: Unknown.

Given the high severity and the potential for remote, unauthenticated disruption of critical inference services, immediate patching is required. Organizations should prioritize updating their Triton Inference Server instances to the latest version to prevent service availability risks.