Skip to content

pyCANON is a Python library and CLI to assess the values of the parameters associated with the most common privacy-preserving techniques.

License

Notifications You must be signed in to change notification settings

IFCA-Advanced-Computing/pycanon

Repository files navigation

pyCANON

License Documentation Status Pipeline Status

pyCANON is a Python library and CLI to assess the values of the parameters associated with the most common privacy-preserving techniques via anonymization.

Authors: Judith Sáinz-Pardo Díaz and Álvaro López García (IFCA - CSIC).

Installation

We recommend to use Python3 with virtualenv:

virtualenv .venv -p python3
source .venv/bin/activate

Then run the following command to install the library and all its requirements:

pip install pycanon

If you also want to install the functionality that allows to generate PDF files for the reports, install as follows

pip install pycanon[PDF]

Documentation

The pyCANON documentation is hosted on Read the Docs.

Getting started

Example using the adult dataset:

import pandas as pd
from pycanon import anonymity, report

FILE_NAME = "adult.csv"
QI = ["age", "education", "occupation", "relationship", "sex", "native-country"]
SA = ["salary-class"]
DATA = pd.read_csv(FILE_NAME)

# Calculate k for k-anonymity:
k = anonymity.k_anonymity(DATA, QI)

# Print the anonymity report:
report.print_report(DATA, QI, SA)

Description

pyCANON allows to check if the following privacy-preserving techniques are verified and the value of the parameters associated with each of them.

Technique pyCANON function Parameters Notes
k-anonymity k_anonymity k: int  
(α, k)-anonymity alpha_k_anonymity α: float k:int  
ℓ-diversity l_diversity : int  
Entropy ℓ-diversity entropy_l_diversity : int  
Recursive (c,ℓ)-diversity recursive_c_l_diversity c: int : int Not calculated if ℓ=1
Basic β-likeness basic_beta_likeness β: float  
Enhanced β-likeness enhanced_beta_likeness β: float  
t-closeness t_closeness t: float For numerical attributes the definition of the EMD (one-dimensional Earth Mover’s Distance) is used. For categorical attributes, the metric "Equal Distance" is used.
δ-disclosure privacy delta_disclosure δ: float  

More information can be found in this paper.

In addition, a report can be obtained including information on the equivalence claases and the usefulness of the data. In particular, for the latter the following three classically used metrics are implemented (as defined in the documentation): average equivalence class size, classification metric and discernability metric.

Citation

If you are using pyCANON you can cite it as follows:

@article{sainzpardo2022pycanon,
   title={A Python library to check the level of anonymity of a dataset},
   author={S{\'a}inz-Pardo D{\'\i}az, Judith and L{\'o}pez Garc{\'\i}a, {\'A}lvaro},
   journal={Scientific Data},
   volume={9},
   number={1},
   pages={785},
   year={2022},
   publisher={Nature Publishing Group UK London}}

Acknowledgments

The authors would like to thank the funding through the European Union - NextGenerationEU (Regulation EU 2020/2094), through CSIC’s Global Health Platform (PTI+ Salud Global) and the support from the project AI4EOSC “Artificial Intelligence for the European Open Science Cloud” that has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement number 101058593.