CVE-2025-32434
📋 TL;DR
A critical Remote Command Execution vulnerability exists in PyTorch when loading models with torch.load(weights_only=True). Attackers can craft malicious model files to execute arbitrary code on systems running vulnerable versions. This affects all users who load untrusted PyTorch model files.
💻 Affected Systems
- PyTorch
📦 What is this software?
Pytorch by Linuxfoundation
⚠️ Risk & Real-World Impact
Worst Case
Full system compromise allowing attackers to execute arbitrary commands, steal data, install malware, or pivot to other systems.
Likely Case
Data theft, cryptocurrency mining, or ransomware deployment through malicious PyTorch model files.
If Mitigated
Limited impact if only trusted model files are loaded and proper network segmentation is in place.
🎯 Exploit Status
Exploitation requires loading a malicious model file. Public proof-of-concept exists in the advisory.
🛠️ Fix & Mitigation
✅ Official Fix
Patch Version: 2.6.0
Vendor Advisory: https://github.com/pytorch/pytorch/security/advisories/GHSA-53q9-r3pm-6pq6
Restart Required: No
Instructions:
1. Upgrade PyTorch to version 2.6.0 or later using pip: pip install --upgrade torch==2.6.0
2. Verify installation with: python -c "import torch; print(torch.__version__)"
3. Test that torch.load with weights_only=True works with your models.
🔧 Temporary Workarounds
Avoid loading untrusted models
allOnly load PyTorch model files from trusted sources. Implement strict validation of model files before loading.
Use pickle.load with restrictions
allIf you must load untrusted files, use pickle.load with restricted unpicklers (though this is complex and error-prone).
🧯 If You Can't Patch
- Implement strict file validation: only allow loading of model files from trusted, verified sources.
- Isolate PyTorch processes in containers or sandboxes with minimal privileges and network access.
🔍 How to Verify
Check if Vulnerable:
Check PyTorch version: python -c "import torch; print(torch.__version__)". If version is 2.5.1 or earlier, you are vulnerable.Check Version:
python -c "import torch; print(torch.__version__)"Verify Fix Applied:
After upgrading, verify version is 2.6.0 or later: python -c "import torch; print(torch.__version__)"📡 Detection & Monitoring
Log Indicators:
- Unexpected process execution from Python/PyTorch processes
- Loading of model files from unusual locations or network sources
Network Indicators:
- Unexpected outbound connections from PyTorch processes
- Downloads of PyTorch model files from untrusted sources
SIEM Query:
Process execution where parent process contains 'python' AND (command_line contains 'torch.load' OR image_path contains 'python')