tapas.threat_models.mia

Threat models for Membership Inference Attacks (MIA).

Membership inference attacks aim at detecting the presence of a specific record in the training dataset from the synthetic dataset observed.

Classes

MIALabeller(attacker_knowledge, target_records)	Randomly add a given target to the datasets sampled from auxiliary data.
TargetedMIA(attacker_knowledge_data, ...[, ...])	This threat model implements a MIA with arbitrary attacker knowledge on data and generator.