obronchain / leakage_certification_revisited Goto Github PK

date:       June 2019
author:     Olivier Bronchain
contact:    [email protected]
Institute:  UCLouvain Crypto Group
eprint:     https://eprint.iacr.org/2019/132 

REQUIREMENTS:
 This code is written in Python3 and is mainly based on Numpy1.14 and Scipy1.1
 running on a Linux distribution. The required packages are:
    - numpy
    - scipy
    - matplotlib
    - tqdm
 These can be installed by running:
    # sudo pip install numpy matplotlib tqdm scipy

USAGE:
 The code can be use to generate the simulations used in the paper "Leakage
 Certification Revisited: Bounding Model Errors in Side-Channel Security Evaluations"
 published at Crypto 2019. The parameters can be changed in main.py and have the
 following usage:
    -D      - number of shares
    -var    - if a matrix, covariance matrix. if scalar, independent noise with variance var
    -Nk     - Size of the secret space (i.e. 256 for 8 bits secrets)
    -serial - Leakage of the shares in different dimensions

 Once the desired parameters are set, one can start the script with:
    python3 main.py


FILES DESCRIPTION
 Leakage_generation.py  - Generates leakage samples
 MI_computations.py     - Actual MI computation
 Models.py              - PDFs estimators
 main.py                - Main script setting the parameters