US 12,229,314 B2
Enhanced data privacy through structure-preserving autoencoder with latent space augmentation
Christopher Allan Ralph, Toronto (CA); and Gerald Fahner, Austin, TX (US)
Assigned to FAIR ISAAC CORPORATION, Minneapolis, MN (US)
Filed by FAIR ISAAC CORPORATION, Roseville, MN (US)
Filed on May 7, 2022, as Appl. No. 17/739,107.
Prior Publication US 2023/0359767 A1, Nov. 9, 2023
Int. Cl. G06F 21/62 (2013.01)
CPC G06F 21/6254 (2013.01) 18 Claims
OG exemplary drawing
 
1. A computer implemented method, comprising:
receiving, using at least one processor, one or more source data from one or more data sources;
generating, using the at least one processor, one or more encoded source data from the one or more source data;
generating a synthetic data by decoding the one or more encoded source data;
selecting one or more variables in the synthetic data;
associating one or more predetermined identifiability values with the one or more variables in the synthetic data, a predetermined identifiability value determined by a latent space augmentation process as a first threshold that indicates a level of identifiability between the synthetic data and corresponding source data according to on one or more augmented vectors defined based on the one or more variables in the synthetic data and a distance between the one or more augmented vectors and one or more variables in the one or more source data;
associating one or more predetermined anonymity values with the one or more variables in the synthetic data, a predetermined anonymity value determined by the latent space augmentation process as a second threshold that indicates a level of anonymity between the synthetic data and the corresponding source data;
decoding the generated synthetic data including the selected variables using the associated one or more predetermined identifiability values and one or more predetermined anonymity values; and
outputting the decoded synthetic data.