MembershipAuditScenario
Statistics scenarios · Implementation: mlsysim/engine/scenario_definitions/statistics.py.
Construct this scenario with keyword inputs. The fields below show the types and defaults declared by this class.
A membership-inference audit on the ROC plane against the ceiling a pure epsilon-DP release allows; the audit curve FPR + a FPR (1 - FPR) is illustrative.
Notes
Equation: The DP ceiling delegates to the privacy helper; the audit curve adds a quadratic uplift to FPR.
Assumptions: False-positive rates lie in [0,1] and the audit curve is constructed by the supplied shape parameter.
Excludes: Empirical attack calibration and approximate-DP delta accounting.
References
Derivation: The DP ceiling delegates to the privacy helper; the audit curve adds a quadratic uplift to FPR. Individual property contracts identify the exact accounting expression or reusable calculation. Registry entries supply the scenario inputs; this class does not turn them into measured outcomes.
Examples
Evaluate a named scenario using its declared inputs.
>>> from mlsysim import Scenarios
>>> scenario = Scenarios.MembershipAuditAtEpsilonOne
>>> round(scenario.ceiling(0.1), 4)
0.2718Attributes
| Name | Description |
|---|---|
| audit_shape | |
| epsilon |
Methods
| Name | Description |
|---|---|
| audit | Illustrative audit TPR at fpr: a quadratic uplift controlled by audit_shape; not |
| ceiling | Pure-epsilon-DP TPR ceiling at the dimensionless fpr, from |
audit
engine.scenario_types.MembershipAuditScenario.audit(fpr)Illustrative audit TPR at fpr: a quadratic uplift controlled by audit_shape; not a measured attack curve.
Equation: fpr + audit_shape * fpr * (1 - fpr).
ceiling
engine.scenario_types.MembershipAuditScenario.ceiling(fpr)Pure-epsilon-DP TPR ceiling at the dimensionless fpr, from calc_membership_tpr_ceiling.
Equation: calc_membership_tpr_ceiling(fpr, epsilon).
epsilon
Type: float.
Declared default: required.
audit_shape
Type: float.
Declared default: required.