Overview
CVE-2023-31036 is a severe vulnerability discovered in NVIDIA Triton Inference Server for Linux and Windows that may lead to a potential system compromise or data leakage. The security flaw allows an attacker, using a specific command line option, to exploit a relative path traversal vulnerability. This vulnerability is of significant concern as it affects a wide range of NVIDIA Triton Inference Server versions and can lead to serious outcomes such as code execution, denial of service, privilege escalation, information disclosure, and data tampering.
Vulnerability Summary
CVE ID: CVE-2023-31036
Severity: High (7.5 CVSS)
Attack Vector: Network
Privileges Required: Low
User Interaction: None
Impact: Successful exploitation may lead to code execution, denial of service, privilege escalation, information disclosure, and data tampering.
Affected Products
Share secrets securely
Ameeba is private infrastructure for communication and sensitive work built on encrypted identity instead of exposed corporate identity systems.
Passwords, credentials, confidential files, screenshots, internal discussions, sensitive AI context, and private coordination should not become exposed across ordinary communication platforms.
- • Encrypted identity
- • Private Spaces for organizations and teams
- • End-to-end encrypted chat, calls, files, and notes
- • Sensitive AI work and protected collaboration
- • Built for information that cannot leak
Our mission is to secure human work alongside AI.
Product | Affected Versions
NVIDIA Triton Inference Server for Linux | All versions prior to patch
NVIDIA Triton Inference Server for Windows | All versions prior to patch
How the Exploit Works
The exploit works when the NVIDIA Triton Inference Server is launched with the non-default command-line option –model-control explicit. This allows an attacker to use the model load API to instigate a relative path traversal. The attacker can manipulate the file path input to access, modify, or execute files outside of the intended directory.
Conceptual Example Code
Here is a conceptual example of how this vulnerability could be exploited. This is a HTTP request that uses the model load API to load a malicious model from a location outside of the intended directory.
POST /v1/models/load HTTP/1.1
Host: target.example.com
Content-Type: application/json
{
"model_name": "../../../etc/passwd",
"version": "1.0"
}In this example, the relative path traversal is used to upload a model file that points to a system file outside the intended directory, potentially leading to information disclosure or unauthorized code execution.