NVIDIA Triton Inference Server Denial-of-Service Vulnerability via Large Compressed Payload

A denial-of-service vulnerability has been identified in NVIDIA Triton Inference Server, present in all versions prior to 26.01. The issue arises in the HTTP endpoint, where an attacker can cause a denial of service by sending a large compressed payload. Exploiting this vulnerability may lead to a significant disruption of service.

Impact

Exploitation of this vulnerability can cause a denial-of-service condition, leading to increased resource consumption and potential service unavailability.

Remediation

Users are advised to update to NVIDIA Triton Inference Server version 26.01 or later. The updated version can be downloaded from the NVIDIA GitHub repository.