opendp 0.14.2-dev.20260401.2

A library of differential privacy algorithms for the statistical analysis of sensitive private data.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111

use cdp_delta::test::cdp_epsilon;

use crate::{
    combinators::make_approximate, domains::AtomDomain, measurements::make_gaussian,
    metrics::AbsoluteDistance,
};

use super::*;

#[test]
fn test_zCDP_to_approxDP_nontrivial() -> Fallible<()> {
    let d_in = 1.0;
    let scale = 4.0;
    let profile = make_zCDP_to_approxDP(make_gaussian(
        AtomDomain::<f64>::new_non_nan(),
        AbsoluteDistance::<f64>::default(),
        scale,
        None,
    )?)?
    .map(&d_in)?;
    let rho = (d_in / scale).powi(2) / 2.0;

    assert_eq!(profile.epsilon(0.)?, f64::INFINITY);

    // using reverse map to check correctness
    // implementation of reverse map is slightly looser by 1 ulp due to numerical imprecision
    assert_eq!(cdp_epsilon(rho, 1e-3)?, 0.6880024554878086);
    assert_eq!(profile.epsilon(1e-3)?, 0.6880024554878085);
    assert_eq!(profile.epsilon(1.0)?, 0.);

    // using reverse map to check correctness
    assert_eq!(cdp_epsilon(rho, 0.1508457845622862)?, 0.0);
    assert_eq!(profile.delta(0.)?, 0.1508457845622862);
    assert_eq!(profile.delta(0.6880024554878085)?, 1e-3);
    Ok(())
}

#[test]
fn test_zCDP_to_approxDP_insensitive() -> Fallible<()> {
    let profile = make_zCDP_to_approxDP(make_gaussian::<_, _, ZeroConcentratedDivergence>(
        AtomDomain::<f64>::new_non_nan(),
        AbsoluteDistance::<f64>::default(),
        4.,
        None,
    )?)?
    .map(&0.0)?;

    assert_eq!(profile.epsilon(0.0)?, 0.0);

    assert!(profile.epsilon(-0.0).is_err());
    assert!(profile.delta(-0.0).is_err());
    Ok(())
}

#[test]
fn test_zCDP_to_approxDP_nonprivate() -> Fallible<()> {
    let profile = make_zCDP_to_approxDP(make_gaussian(
        AtomDomain::<f64>::new_non_nan(),
        AbsoluteDistance::<f64>::default(),
        0.,
        None,
    )?)?
    .map(&1.0)?;

    assert_eq!(profile.epsilon(0.0)?, f64::INFINITY);
    assert_eq!(profile.epsilon(0.1)?, f64::INFINITY);
    assert_eq!(profile.delta(0.0)?, 1.0);
    assert_eq!(profile.delta(0.1)?, 1.0);
    Ok(())
}

#[test]
fn test_zCDP_to_approxDP_insensitive_nonprivate() -> Fallible<()> {
    let profile = make_zCDP_to_approxDP(make_gaussian::<_, _, ZeroConcentratedDivergence>(
        AtomDomain::<f64>::new_non_nan(),
        AbsoluteDistance::<f64>::default(),
        0.,
        None,
    )?)?
    .map(&0.0)?;

    assert_eq!(profile.epsilon(0.0)?, 0.0);
    assert_eq!(profile.epsilon(0.1)?, 0.0);
    assert_eq!(profile.delta(0.0)?, 0.0);
    assert_eq!(profile.delta(0.1)?, 0.0);
    Ok(())
}

#[test]
fn test_approx_zCDP_to_approx_approxDP() -> Fallible<()> {
    let m_zcdp = make_gaussian(
        AtomDomain::<f64>::new_non_nan(),
        AbsoluteDistance::<f64>::default(),
        1.,
        None,
    )?;

    let m_azcdp = make_approximate(m_zcdp)?;
    let m_adp = make_zCDP_to_approxDP(m_azcdp)?;

    let (curve, delta) = m_adp.map(&1.0)?;
    assert_eq!(delta, 0.0);

    let epsilon = curve.epsilon(1e-7)?;

    // when scale is 1 and sensitivity is 1, then rho = (d_in / scale)^2 / 2 = 0.5
    let expected_epsilon = cdp_epsilon(0.5, 1e-7)?;
    assert_eq!(epsilon, expected_epsilon);

    Ok(())
}