Please do not open a public GitHub issue for security-sensitive reports.
Use GitHub's private vulnerability reporting form:
If that form is unavailable, email jfengy04@foxmail.com with the subject
[steel-defect-detection security]. Include the affected file or workflow,
steps to reproduce, expected impact, and safe proof-of-concept details.
You should receive an acknowledgement within seven days. Fix timing depends on severity, reproducibility, and whether an upstream dependency is involved.
This project is a research and learning-oriented computer vision project. Security reports are most relevant for:
- unsafe file handling in scripts or the GUI
- dependency vulnerabilities that affect normal installation
- unsafe model or dataset loading behavior
- CI or repository configuration issues
Only load model weights from sources you trust. PyTorch checkpoint formats may
contain serialized Python objects, so a malicious .pt or .pth file can be a
security risk even when its filename looks normal.
Please do not include real production secrets, private datasets, or proprietary model weights in reports.