| CPC G06F 21/6254 (2013.01) | 19 Claims |
|
1. A computer-implemented method comprising:
receiving an input distribution that is an average of a number of individual distributions, each individual distribution representing data from a respective different entity, wherein the input distribution is a two-dimensional distribution;
using a multi-scale spatial transform, transforming the input distribution into a multi-scale representation that includes two or more representations of the input distribution at respective different spatial scales;
adding noise to the multi-scale representation to generate an anonymized multi-scale representation, wherein adding noise to the multi-scale representation to generate the anonymized multi-scale representation comprises adding a different amount of noise to each of the two or more representations of the input distribution at respective different spatial scales; and
based on the anonymized multi-scale representation, reconstructing an output distribution such that the output distribution, when transformed using the multi-scale spatial transform, results in an approximation of the anonymized multi-scale representation to generate a heatmap ensuring the output distribution preserves the privacy of the input distribution.
|