Granted Patent B2

US 12,321,478 · App. 17/610,795 · Granted Jun 3, 2025

Inventors: Mengyuan Zhang (Hong Kong SAR, CN); Yosr Jarraya (Montreal, CA); Makan Pourzandi (Montreal, CA); Meisam Mohammady (Montreal, CA); Shangyu Xie (Chicago, IL); Yuan Hong (Chicago, IL); Lingyu Wang (Montreal, CA); Mourad Debbabi (Dollard des Ormeaux, CA)

Assignee: Telefonaktiebolaget LM Ericsson (Publ)

G06F21/6227

Abstract

Claims (39)

1. A differential privacy, DP, node, comprising:

processing circuitry configured to:

receive, from a data analyst, a query request;

receive a first input corresponding to a utility parameter, the utility parameter being associated with the data analyst and being a constraint on data privacy;

receive a second input corresponding to a privacy parameter, the privacy parameter being associated with a data owner;

select a baseline DP mechanism type from a plurality of available baseline DP mechanism types based at least on a query request type of the query request, the first input and the second input, the baseline DP mechanism type including at least a noise parameter;

generate a noise distribution based on the baseline DP mechanism type using a first value of the noise parameter; and

determine a DP query result based on applying the noise distribution to the query request applied on a data set.

2. The DP node of claim 1 , wherein the processing circuitry is further configured to:

determine a noise parameter distribution for the noise parameter; and

select the first value of the noise parameter from the noise parameter distribution, the first value of the noise parameter having a greater utility than the remaining plurality of values of the noise parameter distribution.

3. The DP node of claim 2 , wherein the noise parameter distribution for the noise parameter meets requirements for the first and second inputs.

4. The DP node of claim 2 , wherein the noise parameter distribution corresponds to a plurality of values of the noise parameter including the first value; and

the selection of the first value of the plurality of values of the noise parameter being based at least on the first value being greater than or equal to a query sensitivity associated with the query request.

5. The DP node of claim 1 , wherein the DP query result corresponds to an aggregation of a query result to the query request and the noise distribution.

6. The DP node of claim 1 , wherein the baseline DP mechanism type includes one of a Laplace mechanism type, exponential mechanism type and Gaussian mechanism type.

7. The DP node of claim 1 , wherein the processing circuitry is further configured to determine a query sensitivity associated with the query request.

8. The DP node of claim 7 , wherein the selection of the baseline DP mechanism type is based at least on the query sensitivity.

9. The DP node of claim 1 , wherein the privacy parameter corresponds to a privacy constraint associated with a probability of a predefined loss of privacy.

10. The DP node of claim 1 , wherein the utility parameter corresponds to an error constraint.

11. A method implemented by a differential privacy, DP, node, the method comprising:

receiving, from a data analyst, a query request;

receiving a first input corresponding to a utility parameter, the utility parameter being associated with the data analyst and being a constraint on data privacy;

receiving a second input corresponding to a privacy parameter, the privacy parameter being associated with a data owner;

selecting a baseline DP mechanism type from a plurality of available baseline DP mechanism types based at least on a query request type of the query request, the first input and the second input, the baseline DP mechanism type including at least a noise parameter;

generating a noise distribution based on the baseline DP mechanism type using a first value of the noise parameter; and

determining a DP query result based on applying the noise distribution to the query request applied on a data set.

12. The method of claim 11 , further comprising:

determining a noise parameter distribution for the noise parameter; and

selecting the first value of the noise parameter from the noise parameter distribution, the first value of the noise parameter having a greater utility than the remaining plurality of values of the noise parameter distribution.

13. The method of claim 12 , wherein the noise parameter distribution for the noise parameter meets requirements for the first and second inputs.

14. The method of claim 12 , wherein the noise parameter distribution corresponds to a plurality of values of the noise parameter including the first value; and

the selection of the first value of the plurality of values of the noise parameter being based at least on the first value being greater than or equal to a query sensitivity associated with the query request.

15. The method of claim 11 , wherein the DP query result corresponds to an aggregation of a query result to the query request and the noise distribution.

16. The method of claim 11 , wherein the baseline DP mechanism type includes one of a Laplace mechanism type, exponential mechanism type and Gaussian mechanism type.

17. The method of claim 11 , further comprising determining a query sensitivity associated with the query request.

18. The method of claim 17 , wherein the selection of the baseline DP mechanism type is based at least on the query sensitivity.

19. The method of claim 11 , wherein the privacy parameter corresponds to a privacy constraint associated with a probability of a predefined loss of privacy.

20. The method of claim 11 , wherein the utility parameter corresponds to an error constraint.

Assignments (1)

Continuity (2)

References Cited (20)