| Version | Supported |
|---|---|
| 0.1.x | ✅ |
If you discover a security vulnerability, please do not open a public issue.
Instead, report it privately:
- Email the maintainers directly
- Include a clear description, steps to reproduce, and impact assessment
- Allow up to 14 days for acknowledgment and an initial fix
We take security seriously and will respond as quickly as possible.
This package is a research library for Information Bottleneck methods. Vulnerabilities to be aware of:
- Pickle/deserialization:
torch.loadwith untrusted checkpoints can execute arbitrary code. Always verify checkpoint provenance. - YAML parsing: Use
yaml.safe_load(notyaml.load) for untrusted config files. Our codebase usessafe_loadby default. - Dependency supply chain: Third-party packages (PyTorch, transformers, etc.) may contain vulnerabilities. Keep dependencies updated.
- 0–48 hours: Initial acknowledgment
- 48 hours–7 days: Investigation and fix development
- 7–14 days: Coordinated disclosure