CVE-2023-43472
📋 TL;DR
This vulnerability in MLFlow allows remote attackers to access sensitive information through crafted REST API requests. It affects MLFlow deployments with exposed REST APIs, potentially exposing model data, configuration details, or credentials.
💻 Affected Systems
- MLFlow
📦 What is this software?
Mlflow by Lfprojects
⚠️ Risk & Real-World Impact
Worst Case
Complete compromise of MLFlow instance with exposure of all stored models, experiment data, and potentially credentials or API keys.
Likely Case
Unauthorized access to sensitive ML model artifacts, experiment metadata, or configuration files.
If Mitigated
Limited exposure of non-critical information if proper network segmentation and authentication are implemented.
🎯 Exploit Status
Exploitation requires crafting specific REST API requests but does not require authentication.
🛠️ Fix & Mitigation
✅ Official Fix
Patch Version: 2.9.0 or later
Vendor Advisory: https://mlflow.org/docs/latest/releases.html
Restart Required: Yes
Instructions:
1. Upgrade MLFlow to version 2.9.0 or later. 2. Restart all MLFlow services. 3. Verify the upgrade was successful.
🔧 Temporary Workarounds
Network Segmentation
allRestrict access to MLFlow REST API endpoints using firewall rules or network ACLs.
Authentication Enforcement
allImplement authentication for all REST API endpoints if not already enabled.
🧯 If You Can't Patch
- Implement strict network access controls to limit MLFlow API exposure
- Enable and enforce authentication for all MLFlow REST API endpoints
🔍 How to Verify
Check if Vulnerable:
Check MLFlow version using 'mlflow --version' or examine package version in deployment.Check Version:
mlflow --versionVerify Fix Applied:
Confirm MLFlow version is 2.9.0 or later and test REST API endpoints for information disclosure.📡 Detection & Monitoring
Log Indicators:
- Unusual REST API request patterns
- Access to sensitive endpoints without authentication
Network Indicators:
- Unusual traffic to MLFlow REST API endpoints
- Requests attempting to access sensitive endpoints
SIEM Query:
source="mlflow" AND (url_path="/api/*" OR method="GET") AND response_code=200 AND user="anonymous"🔗 References
- https://www.contrastsecurity.com/security-influencers/discovering-mlflow-framework-zero-day-vulnerability-machine-language-model-security-contrast-security
- https://www.contrastsecurity.com/security-influencers/discovering-mlflow-framework-zero-day-vulnerability-machine-language-model-security-contrast-security
