This Python build tool enables a given user to calculate a variety of different data privacy metrics on tabular data from a user interface.

• K-anonymity ¹
• ℓ-diversity ²
• Sample Unique Detection Algorithm (SUDA) ³
• Privacy Information Factor (PIF) ⁴

• Noise addition
• Field generalisation
• Rounded Approximation

Input can be in either CSV or TSV format. For meta information, an option to load a JSON file is available.

• Python 3.7+ (tested with Python 3.13)
• Conda environment (recommended)

⚠️ Windows users must install R and Rtools before proceeding:

1. Install R: Download and install R from CRAN
2. Install Rtools: Download and install Rtools from CRAN Rtools
   • Rtools provides the necessary build tools (make, gcc) required for compiling R packages
   • Make sure to add Rtools to your system PATH during installation
3. Install required R packages: Open R or RStudio and run:

   install.packages("sdcMicro")

4. Set R_HOME environment variable (if needed):

   set R_HOME=C:\Program Files\R\R-4.5.1

Why this is needed: The metaprivBIDS package depends on rpy2 which requires R and build tools to compile properly on Windows. Additionally, the sdcMicro R package is required for privacy analysis functionality. Without Rtools, you'll get "make: command not found" errors.

First, activate your conda environment:

conda activate your-env-name  # or source your-venv/bin/activate

Navigate to the MetaprivBIDS directory and run the interactive installer:

cd metaprivBIDS
python install.py

The installer will:

• ✅ Fix pkg_resources deprecation warning (automatically pins setuptools<81 if needed)
• ✅ Install Qt dependencies (attempts to resolve GUI compatibility issues)
• ❓ Ask about pygraphviz (optional package for advanced graph visualization)

On some systems, the Qt GUI may hang due to missing X11/XCB libraries or platform compatibility issues. The installer provides several launch options:

Use the intelligent launcher that tests Qt compatibility and provides fallbacks:

python run_metaprivBIDS_safe.py

This launcher will:

1. Test CLI functionality first
2. Ask if you want to try the GUI
3. Test Qt compatibility with timeouts
4. Provide fallback to CLI mode if GUI fails

If you only need the core functionality without GUI:

python test_cli.py

For direct GUI access (use with caution):

python run_metaprivBIDS.py

If this hangs, use Ctrl+C to interrupt and try the safe launcher.

If the GUI hangs, try these Qt platform alternatives:

# Minimal platform (no visual output but functional)
QT_QPA_PLATFORM=minimal python run_metaprivBIDS.py

# Offscreen platform (for headless servers)
QT_QPA_PLATFORM=offscreen python run_metaprivBIDS.py

After following the installation guide, the metrics within the MetaprivBIDS tool can be called through an import statement without making use of the GUI.

from metaprivBIDS.corelogic.metapriv_corelogic import metaprivBIDS_core_logic
metapriv = metaprivBIDS_core_logic()

# Load the data
data_info = metapriv.load_data('Use_Case_Data/adult_mini.csv')

# Inspect {column, unique value count, column type}
data = data_info["data"]
print("Column Types:", '\n')
print(data_info["column_types"], '\n')

# Select Quasi-Identifiers
selected_columns = ["age", "education", "marital-status", "occupation", "relationship", "sex", "salary-class"]
results_k_global = metapriv.find_lowest_unique_columns(data, selected_columns)
print('Find Influential Columns:', '\n')
print(results_k_global)

# Compute Personal Information Factor 
pif_value, cig_df = metapriv.compute_cig(data, selected_columns)
print("PIF Value:", pif_value)
print("CIG DataFrame:")
print(cig_df)

# Run SUDA2 computation
results_suda = metapriv.compute_suda2(data, selected_columns, sample_fraction=0.3, missing_value=-999)

# Access results
data_with_scores = results_suda["data_with_scores"]
attribute_contributions = results_suda["attribute_contributions"]
attribute_level_contributions = results_suda["attribute_level_contributions"]

To run tests, navigate to the tests folder and activate your environment:

cd tests
conda activate your-env-name  # or source your-venv/bin/activate
python test_metaprivBIDS_core_logic.py

Note: Install pytest if needed: pip install pytest

UserWarning: pkg_resources is deprecated as an API

Solution: The installer automatically fixes this by pinning setuptools<81. If you see this warning, run:

pip install "setuptools<81"

qt.qpa.plugin: Could not load the Qt platform plugin "xcb"
This plugin does not support propagateSizeHints()

Solutions:

1. Use the safe launcher: python run_metaprivBIDS_safe.py
2. Use CLI-only mode: python test_cli.py
3. Try platform fallbacks:
   • QT_QPA_PLATFORM=minimal python run_metaprivBIDS.py
   • QT_QPA_PLATFORM=offscreen python run_metaprivBIDS.py

If the GUI hangs, press Ctrl+C to interrupt and use:

• Safe launcher: python run_metaprivBIDS_safe.py
• CLI mode: python test_cli.py

If you get module import errors, ensure:

1. Your conda environment is activated
2. The package is installed: python install.py
3. You're in the correct directory

For running tests:

pip install pytest
cd tests
python test_metaprivBIDS_core_logic.py

If you cannot install system packages (sudo access), the CLI mode will work without additional system dependencies.

Error: make: command not found or R was not built as a library

Cause: Missing R installation or Rtools build tools on Windows.

Solutions:

1. Install R: Download from CRAN
2. Install Rtools: Download from CRAN Rtools
3. Add Rtools to PATH: During Rtools installation, check "Add to PATH"
4. Set R_HOME environment variable:

   set R_HOME=C:\Program Files\R\R-4.5.1

5. Alternative: Use pre-compiled rpy2:

   pip uninstall rpy2
   pip install --only-binary=all rpy2

6. Alternative: Use conda for rpy2:

   conda install -c conda-forge rpy2

Note: If rpy2 is causing issues and isn't critical for your use case, you may be able to skip R-related functionality.

Error: The R package "sdcMicro" is not installed

Cause: Required R package for privacy analysis functionality is missing.

Solutions:

1. Install via R console:

   install.packages("sdcMicro")

2. Install via RStudio: Open RStudio and run the same command
3. Install from R command line:

   R -e "install.packages('sdcMicro')"

4. If installation fails, try installing dependencies first:

   install.packages(c("VIM", "robustbase", "cluster"))
   install.packages("sdcMicro")

Note: The sdcMicro package is essential for k-anonymity, l-diversity, and other privacy metrics. The application may not function properly without it.

For additional support:

• Use CLI mode for core functionality: python test_cli.py
• Check the debug script: python debug_test.py
• The core logic is fully functional without GUI dependencies