metaprivBIDS documentation

metaprivBIDS is a local disclosure-risk workbench for tabular data. It offers the same reusable privacy and anonymisation functions through a Python API, command-line interface, and cross-platform browser interface.

Risk assessment includes:

• k-anonymity and l-diversity;

• K-global and K-combined contribution analysis;

• Personal Information Factor (PIF), cell information gain (CIG), and row information gain (RIG); and

• SUDA2 through the R sdcMicro package.

Mitigation includes direct-identifier pseudonymisation with row shuffling, equal-width binning, rounding, decimal removal, Laplacian or Gaussian noise, categorical generalisation, and per-column reversion.

Contents

• Getting started
  • Cross-platform installation
  • Tests and notebooks
  • Local browser interface
  • Command line
  • Documentation screenshots
  • Optional activation
  • Python API
• Understand Metrics
  • Worked example: Table 2
  • k-anonymity and sample-unique rows
  • l-diversity
  • K-global: contribution of one variable
  • K-combined: contribution of several variables
  • SUDA2: minimal combinations that single someone out
  • CIG, RIG, and PIF: information revealed by the other variables
  • Reading the results together
• Command-line interface
  • Inspect data and measure privacy
  • Transform values
  • Replace direct identifiers
  • Compute CIG, RIG, and PIF
  • Compute SUDA2
  • Inspect JSON metadata
• Browser workflow
  • 1. Load and describe the data
  • 2. Calculate the privacy summary
  • 3. Transform the working copy
  • 4. Inspect PIF and information gain
  • 5. Run SUDA2
  • Reproduce the screenshots
• Functionality reference
  • Data workspace
  • Risk analysis
  • PIF / information gain
  • SUDA2
  • Anonymisation
  • Interaction and safety
• metaprivBIDS
  • metaprivBIDS package

License

metaprivBIDS is licensed under the MIT License.

Indices

• Index

• Module Index

• Search Page