NVIDIA Triton Inference Server < 25.07 - Information Disclosure via Python Backend Shared Memory Exhaustion
Title source: llmExploitation Summary
EIP tracks 1 public exploit for CVE-2025-23320. PoCs published by There-was-a-bird.
AI-analyzed exploit summary This repository demonstrates CVE-2025-23320, an information leakage vulnerability in NVIDIA Triton Inference Server (CWE-209). The exploit leverages error messages containing sensitive shared memory keys to trigger a race condition, leading to an out-of-bounds write via Triton's unregister/register API.
Description
NVIDIA Triton Inference Server for Windows and Linux contains a vulnerability in the Python backend, where an attacker could cause the shared memory limit to be exceeded by sending a very large request. A successful exploit of this vulnerability might lead to information disclosure.
Exploits (1)
This repository demonstrates CVE-2025-23320, an information leakage vulnerability in NVIDIA Triton Inference Server (CWE-209). The exploit leverages error messages containing sensitive shared memory keys to trigger a race condition, leading to an out-of-bounds write via Triton's unregister/register API.
References (3)
Scores
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:N