Abstract

A partition of a dataset is obtained by querying the dataset using a grouping key. Each data point of the partition is stored in a set of bound queues 125. A representative parameter for each bound queue of the set of bound queues 125 is determined. The representative parameter can include a target percentile value and index, neighboring distinct values and indices, and a count of distinct entries within a respective bound queue. A sensitivity value is calculated using the representative parameter for each bound queue of the set of bound queues 125. The sensitivity value calibrates the amount of noise added to the partition to minimize the impact of any single datapoint of the partition to a result of SQL aggregations.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.

Share

COinS