Privacy-aware query management system
View Patent ↗A privacy-aware query management system receives a query to aggregate data from a number of datasets, wherein at least one of the datasets contains private information for one or more people. The privacy-aware query management system retrieves data from the datasets. Each of the plurality of datasets has associated privacy parameters that determine a level of noise to be applied to query results obtained from that dataset, and the levels of noise are used to determine an order of operations for the query to retrieve the data from the datasets. The privacy-aware query management system applies the levels of noise to each query result obtained from each of the datasets in accordance with the privacy parameters associated with that dataset, computes aggregated data from the query results with the levels of noise applied, and responds to the query with the aggregated data.
1. A method of query optimization, the method being implemented by one or more processors and comprising:
receiving a query to aggregate data from a plurality of datasets, wherein at least one of the plurality of datasets contains private information for one or more people;
retrieving data from the plurality of datasets, each of the plurality of datasets having associated privacy parameters that determine a level of noise to be applied to query results obtained from that dataseta;
applying the levels of noise to each query result obtained from each of the plurality of datasets in accordance with the privacy parameters associated with that dataset, the query results being obtained in accordance with an order of database operations determined by the levels of noise;
computing aggregated data from the query results with the levels of noise applied; and
responding to the query with the aggregated data.
2. The method of claim 1 , wherein the order of database operations comprises a query plan constructed to minimize an overall level of noise applied to the aggregated data from the query results.
3. The method of claim 2 , further comprising:
receiving a privacy directive along with the query; and
constructing the query plan to optimize the overall level of noise and satisfy the privacy directive.
4. The method of claim 1 , wherein the level of noise to be applied for each dataset increases based on each operation performed on that dataset.
5. The method of claim 1 , wherein the privacy parameters associated with each dataset include a separate set of privacy parameters for each field in that dataset, and the level of noise to be applied increases based on the separate set of privacy parameters for each field inspected by each operation.
6. The method of claim 1 , wherein the privacy parameters include a base level of noise and a privacy budget.
7. The method of claim 6 , wherein the base levels of noise and privacy budgets for each dataset are predetermined values chosen to protect privacy for the one or more people.
8. A query optimization system comprising:
a memory resource to store instructions;
one or more processors using the instructions stored in the memory resource to:
receive a query to aggregate data from a plurality of datasets, wherein at least one of the plurality of datasets contains private information for one or more people;
construct a query plan to retrieve data from the plurality of datasets, each of the plurality of datasets having associated privacy parameters that determine a level of noise to be applied to query results obtained from that dataset;
in response to executing the query plan on the plurality of datasets, apply the levels of noise to each query result obtained from each of the plurality of datasets in accordance with the privacy parameters associated with that dataset, the query results being obtained in accordance with an order of database operations determined by the levels of noise;
compute aggregated data from the query results with the levels of noise applied; and
respond to the query with the aggregated data.
9. The system of claim 8 , wherein constructing the query plan to optimize application of the levels of noise includes minimizing an overall level of noise applied to the aggregated data from the query results.
10. The system of claim 8 , further comprising instructions used by the one or more processors to:
receive a privacy directive along with the query; and
construct the query plan to optimize the level of noise and satisfy the privacy directive.
11. The system of claim 8 , wherein the level of noise to be applied for each dataset increases based on each operation performed on that dataset.
12. The system of claim 8 , wherein the privacy parameters associated with each dataset include a separate set of privacy parameters for each field in that dataset, and the level of noise to be applied increases based on the separate set of privacy parameters for each field inspected by each operation.
13. The system of claim 8 , wherein the privacy parameters include a base level of noise and a privacy budget.
14. The system of claim 13 , wherein the base levels of noise and privacy budgets for each dataset are predetermined values chosen to protect privacy for the one or more people.
15. A non-transitory computer-readable medium that stores instructions, executable by one or more processors, to cause the one or more processors to:
receive a query to aggregate data from a plurality of datasets, wherein at least one of the plurality of datasets contains private information for one or more people;
construct a query plan to retrieve data from the plurality of datasets, each of the plurality of datasets having associated privacy parameters that determine a level of noise to be applied to query results obtained from that dataset;
in response to executing the query plan on the plurality of datasets, apply the levels of noise to each query result obtained from each of the plurality of datasets in accordance with the privacy parameters associated with that dataset, the query results being obtained in accordance with an order of database operations determined by the levels of noise;
compute aggregated data from the query results with the levels of noise applied; and
respond to the query with the aggregated data.
16. The non-transitory computer-readable medium of claim 15 , wherein constructing the query plan to optimize application of the levels of noise includes minimizing an overall level of noise applied to the aggregated data from the query results.
17. The non-transitory computer-readable medium of claim 15 , further comprising instructions used by the one or more processors to:
receive a privacy directive along with the query; and
construct the query plan to optimize the level of noise and satisfy the privacy directive.
18. The non-transitory computer-readable medium of claim 15 , wherein the level of noise to be applied for each dataset increases based on each operation performed on that dataset.
19. The non-transitory computer-readable medium of claim 15 , wherein the privacy parameters associated with each dataset include a separate set of privacy parameters for each field in that dataset, and the level of noise to be applied increases based on the separate set of privacy parameters for each field inspected by each operation.
20. The non-transitory computer-readable medium of claim 15 , wherein the privacy parameters include a base level of noise and a privacy budget.