Privattacks documentation

Privattacks is a Python package for evaluating privacy risks in tabular datasets. It provides tools to quantify re-identification and attribute inference vulnerabilities based on combinations of quasi-identifiers an adversary knows about a target. The package supports both high-level and granular analysis, including per-record vulnerability distributions, multiprocessing for efficiency, and flexible input handling via pandas DataFrames and other sources.

Available tools

• Prior vulnerability for re-identification and attribute inference.

• Posterior vulnerability for a given combination of qids and/or sensitive attribute.

• Posterior vulnerability for a subset of all possible combinations of QIDs.

• Parellel code.

• Generate the histogram of vulnerabilities (i.e., vulnerability per record).

Installation

You can install via PyPI:

pip install privattacks

or manuallly by copying this repository to your local machine and running:

pip install path/to/privattacks

To verify if the package was install corretly, you can run tests:

cd path/to/privattacks
python -m unittest discover tests

License

This project is licensed under the MIT License.

Contents:

• Attack definitions
  • Definitions
  • Example dataset
  • Re-identification
  • Attribute Inference
• Getting Started
  • Data Preparation
  • Defining QIDs and Sensitive Attribute
  • Evaluating Prior and Posterior Vulnerabilities
  • Using the Optimized Evaluation Method
  • Analyzing Individual Vulnerabilities
  • Optimized Method with Distributions
  • Evaluating Multiple Combinations of QIDs
• Privattacks modules
  • privattacks.attacks
  • privattacks.data
  • privattacks.util