Generation of representative data to preserve membership privacy

A method for generating representative data, the method comprising:

• creating, by a generator using a learning model, one or more generated records based on a plurality of training records obtained from a sensitive database;
• identifying, by a discriminator, the generated records as being generated based on the training records;
• and
• identifying, by a privacy adversary, a training sample as being more similar to a distribution of the generated records than a distribution of reference records.

A non-transitory computer readable medium comprising computer executable instructions which when executed by a computer cause the computer to perform the method of:

• implementing a generator, a discriminator, and a privacy adversary;
• creating, by the generator using a learning model, one or more generated records based on a plurality of training records obtained from a sensitive database;
• identifying, by the discriminator, the generated records as being generated based on the training records;
• and
• identifying, by the privacy adversary, a training sample as being more similar to a distribution of the generated records than a distribution of reference records.