Skip to content

Latest commit

 

History

History
43 lines (29 loc) · 2 KB

File metadata and controls

43 lines (29 loc) · 2 KB

Contributing

Thank you for contributing. This is a volunteer project, so we keep the process lightweight.

Where to contribute

What you're adding Where to put it
Privacy method code privacy_methods/d1_structured_data/<method>/ or d2_unstructured_data/<method>/
Method documentation docs/d1_structured_data/<method>/ or d2_unstructured_data/<method>/
Paper reviews / research research/d1_structured_data/<method>/ or d2_unstructured_data/<method>/
Planning docs research/planning/
Dataset metadata datasets/<dataset_folder>/

How to contribute

  1. Create or switch to your group branch
  2. Make your changes
  3. Open a Pull Request into main
  4. A maintainer reviews and merges

Folder conventions

All three core folders — privacy_methods/, docs/, and research/ — use the same D1/D2 hierarchy:

  • d1_structured_data/ — baseline_deidentification, differential_privacy, federated_learning, synthetic_data
  • d2_unstructured_data/ — deidentification, llm_privacy_controls, privacy_attacks

Minimum requirements for new method folders

  • A README.md with concrete content: purpose, specific methods in scope, defined benchmark tasks (inputs/outputs/evaluation), and named metrics — not template prompts
  • A runner script (run.py or run.R) that writes output to results/ (local only); the scaffold already provides CLI structure, so new contributions should add real benchmark logic

Moving beyond scaffolds

The initial repository structure uses scaffold runners and template READMEs to bootstrap the folder layout. Contributions that replace scaffold content with working implementations are the highest priority. See next_steps.md for the current roadmap and ownership assignments.

Data safety

  • Never commit raw patient-level data
  • Never commit credentials or secrets
  • Store local data in data/ (git-ignored)
  • Store experiment outputs in results/ or experiments/ (git-ignored)