Computing confidential data insight histograms and combining with smoothed posterior distribution based histograms

摘要

In an example embodiment, a number of buckets is determined for an empirical histogram of confidential data values submitted to a computer system. The empirical histogram for the confidential data values is computed. Posterior distribution endpoints corresponding to the computed empirical histogram endpoints of the empirical histogram are computed. Then an interval between the posterior distribution endpoints is divided into the determined number of buckets, producing a smoothed histogram based on the posterior distribution. A weight factor is determined based on a ratio between a number of valid entries in a cohort of the confidential data values and a threshold used to determine whether smoothing needs to be performed. Linear interpolation of bucket endpoints is performed for the empirical histogram and the smoothed histogram, using the weight factor to weight the empirical histogram.