Evaluating Linkability of Biometric Templates

If you want to evaluate the linkability of your biometric templates using the Maximal Linkability, you can use the provided function in maximal_linakbility.py and as follows:

import numpy

def maximal_linakbility_metric(mated_scores, nonmated_scores, n_bin=30):
    """
    Parameters:
        (array) mated_scores:    mated scores
        (array) nonmated_scores: nonmated scores
        (int)   n_bin:         the number of histogram bins 

    Return:
        (float) linkability of templates in [0,1] interval. 
                higher value indicates higher link between templates.
    """
    
    # find histogram bin edges
    min_interval_scores = min(mated_scores.min(), nonmated_scores.min())
    max_interval_scores = max(mated_scores.max(), nonmated_scores.max())
    bin_edges = min_interval_scores + numpy.arange(n_bin+1)* (max_interval_scores - min_interval_scores)/n_bin

    # calculate histograms
    y1 = numpy.histogram(mated_scores, bins = bin_edges, density = True)[0]
    y2 = numpy.histogram(nonmated_scores, bins = bin_edges, density = True)[0]

    # calclulate maximal leakage
    MaxLeakage_score = 0 
    for i in range(n_bin):
        MaxLeakage_score += max(y1[i], y2[i])

    return numpy.log2( MaxLeakage_score * ( bin_edges[1]-bin_edges[0] ) )

NOTE: This code only requires numpy package to be installed.

You can also use the following command to evaluate the linkability of biometric templates:

python maximal_linakbility.py --mated_scores <path_to_mated_scores> --nonmated_scores <path_to_nonmated_scores>

Reference

If you use this metric, please cite the following paper, which is published in the IEEE Transactions on Information Forensics and Security. The PDF version of the paper is available as open access on the IEEE-Xplore. The complete source code for reproducing all experiments in the paper is also publicly available in the official repository.

@article{linkability_maxleakage,
  title={Measuring Linkability of Protected Biometric Templates using Maximal Leakage},
  author={Otroshi Shahreza, Hatef and Shkel, Yanina Y. and Marcel, S{\'e}bastien},
  journal={IEEE Transactions on Information Forensics and Security},
  year={2023},
  publisher={IEEE}
}