opendp 0.16.0

A library of differential privacy algorithms for the statistical analysis of sensitive private data.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17

# type: ignore
def sample_discrete_gaussian(scale: RBig) -> int:
    if scale == 0:
        return 0

    t = floor(scale) + 1  # |\label{line:t}|
    sigma2 = scale**2

    while True:
        candidate = sample_discrete_laplace(t)  # |\label{line:candidate}|

        # prepare rejection probability: "bias"
        x = abs(candidate) - sigma2 / t
        bias = x**2 / (2 * sigma2)  # |\label{line:bias}|

        if sample_bernoulli_exp(bias):  # |\label{line:bern}|
            return candidate