System, Method, and Computer Program Product For Deep Learning With Plausible Deniability

A method, comprising:

• (i) obtaining, with at least one processor, a dataset including a plurality of batches of data samples;
• (ii) computing, with the at least one processor, a plurality of gradients of the plurality of batches of data samples;
• (iii) selecting, with the at least one processor, a gradient of the plurality of gradients;
• (iv) adding, with the at least one processor, noise to the selected gradient to generate a noised gradient;
• (v) determining, with the at least one processor, based on the noised gradient and the plurality of gradients, a number of gradients of the plurality of gradients that satisfy a privacy function;
• (vi) in response to the number of gradients of the plurality of gradients that satisfy the privacy function satisfying a threshold number, training, with the at least one processor, using the noised gradient, a machine learning model;
• and
• (vii) in response to the number of gradients of the plurality of gradients that satisfy the privacy function failing to satisfy the threshold number, reshuffling, with the at least one processor, the dataset to generate an updated plurality of batches of data samples and returning to step (ii).

A computer program product, comprising:

• at least one non-transitory computer-readable medium including program instructions that, when executed by at least one processor, cause the at least one processor to:
• (i) obtain a dataset including a plurality of batches of data samples;
• (ii) compute a plurality of gradients of the plurality of batches of data samples;
• (iii) select a gradient of the plurality of gradients;
• (iv) add noise to the selected gradient to generate a noised gradient;
• (v) determine, based on the noised gradient and the plurality of gradients, a number of gradients of the plurality of gradients that satisfy a privacy function;
• (vi) in response to the number of gradients of the plurality of gradients that satisfy the privacy function satisfying a threshold number, train, using the noised gradient, a machine learning model;
• and
• (vii) in response to the number of gradients of the plurality of gradients that satisfy the privacy function failing to satisfy the threshold number, reshuffle the dataset to generate an updated plurality of batches of data samples and returning to step (ii).