metaprivBIDS

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

Methods

Assesing Privacy risk:

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

Mitigating Privacy Risk

• Noise addition
• Field generalisation
• Rounded Approximation

Input data format

Input can be in either CSV or TSV format. For meta information an option of load of json file is possible.

Software installation

Option 1

The metaprivBIDS software runs on multiple platforms (e.g. Linux, MacOS, Windows) that have a Python 3 installation. It is recommended (but not required) to first create a virtual environment.

In the event of permission issues for system dependent files, you might want to set the pkgs_dirs option in Conda's configuration to use a directory that is writable by you.

conda config --add pkgs_dirs ~/conda_pkgs

Creates the enviroment.

conda create --name venv python=3.x.x

Activates the environment.

conda activate venv 

Graphviz requires system level dependencies as well as rpy2 and might need to be installed with

conda create --name venv -c conda-forge "python>=3.7" graphviz pygraphviz r-base r-sdcMicro rpy2

if not avaliable.

You can then install metaprivBIDS by furst cloning the git respository.

git clone https://github.com/CPernet/metaprivBIDS.git

cd into the MetaprivBIDS folder

cd MetaprivBIDS

and then run

pip install -e . 

Option 2

python -m venv venv
source venv/bin/activate

You can then install metaprivBIDS by cloning the git respository.

git clone https://github.com/CPernet/metaprivBIDS.git

Dependencies

To execute the program make sure all dependencies from pyproject.toml is avalible in a python 3.7 enviroment as stated in the software installation. This can be done by first cd into the MetaprivBIDS directiory and then running

pip install -e . 

Usage

To execute the program run from command line

metaprivBIDS

prompting the program to start.

Command-Line Execution

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

e.g.

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

# Load the data
data_info = metapriv.load_data('metaprivBIDS/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"]

Related tools

Footnotes

1. Sweeney, L. (2002). k-Anonymity: A Model for Protecting Privacy. International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems, 10(05), 557-570. ↩

2. Machanavajjhala, A., Kifer, D., Gehrke, J., & Venkitasubramaniam, M. (2007). ℓ-Diversity: Privacy Beyond k-Anonymity. ACM Transactions on Knowledge Discovery from Data (TKDD), 1(1), 3-es. ↩

3. Elliott, M. J., & Skinner, C. J. (2000). Identifying population uniques using limited information. Proceedings of the Annual Meeting of the American Statistical Association. ↩

4. Information Governance ANZ. (2019). Privacy Impact Assessment eReport. Link ↩