Dynamic calibration of noise parameters for data security
View Patent ↗Data security for a dataset in interactive query or operation from users regarding data stored in the dataset is provided. A method for providing data security for the dataset in secure data computation and communication includes generating a query result corresponding to the user query for the dataset, determining a magnitude range of the query result, and generating an amount of random noise data based on the magnitude range. The amount of random noise data is calibrated by adjusting at least one of a first tunable parameter within a first range and a second tunable parameter within a second range of a differential privacy (DP) configuration. A noise-laden query result is generated by applying the noise data to the query result to satisfy an error tolerance level.
1 . A method for providing data security for a dataset in secure data computation and communication, the method comprising:
generating a query result corresponding to a user query for the dataset;
determining a magnitude range of the query result;
adjusting at least one of a first tunable parameter within a first range and a second tunable parameter within a second range of a differential privacy (DP) configuration;
calibrating an amount of random noise data based on the magnitude range and at least one of the first tunable parameter and the second tunable parameter;
generating a noise-laden query result by applying the amount of random noise data to the query result to satisfy an error tolerance level, thereby perturbing the query result before releasing the query result, and thereby providing data security for the dataset; and
caching the query result, the magnitude range of the query result, and the error tolerance level in an expandable search grid data structure.
2 . The method of claim 1 , wherein calibrating the amount of random noise data further comprises generating a plurality of random sample values within the magnitude range of the query result.
3 . The method of claim 2 , further comprising generating a random sample noise for each of the plurality of random sample values in iterations by varying the first tunable parameter within the first range or varying the second tunable parameter within the second range.
4 . The method of claim 3 , further comprising generating noise-laden sample values by applying the random sample noise to corresponding ones of the plurality of random sample values.
5 . The method of claim 4 , further comprising determining a value for at least one of the first tunable parameter and the second tunable parameter from the iterations to allow corresponding noise-laden sample values to satisfy the error tolerance level.
6 . The method of claim 1 , further comprising:
receiving a second user query for the dataset;
generating a second query result for the second user query;
determining a second magnitude range of the second query result; and
determining a second value for at least one of the first tunable parameter and the second tunable parameter by traversing through the expandable search grid data structure based on the second magnitude range of the second query result and a second error tolerance level.
7 . The method of claim 1 , wherein the amount of random noise is generated according to a noise distribution.
8 . The method of claim 7 , wherein the noise distribution comprises at least one of a Laplace distribution, an exponential distribution, or a Gaussian distribution.
9 . The method of claim 1 , further comprising receiving a user input indicative of the error tolerance level.
10 . The method of claim 1 , wherein the user query comprises an aggregate query.
11 . A system for providing data security for a dataset in secure data computation and communication, the system comprising:
a memory to store the dataset; and
a processor to:
generate a query result corresponding to a user query for the dataset;
determine a magnitude range of the query result;
adjust at least one of a first tunable parameter within a first range and a second tunable parameter within a second range of a differential privacy (DP) configuration;
calibrate an amount of random noise data based on the magnitude range and at least one of the first tunable parameter and the second tunable parameter;
generate a noise-laden query result by applying the amount of random noise data to the query result to satisfy an error tolerance level, thereby perturbing the query result before releasing the query result, and thereby providing data security for the dataset; and
cache the query result, the magnitude range of the query result, and the error tolerance level in an expandable search grid data structure.
12 . The system of claim 11 , wherein the processor is further configured to generate a plurality of random sample values within the magnitude range of the query result.
13 . The system of claim 12 , wherein the processor is further configured to:
generate a random sample noise for each of the plurality of random sample values in iterations by varying the first tunable parameter within the first range or varying the second tunable parameter within the second range; and
generate noise-laden sample values by applying the random sample noise to corresponding ones of the plurality of random sample values.
14 . The system of claim 13 , wherein the processor is further configured to determine a value for at least one of the first tunable parameter and the second tunable parameter from the iterations to allow corresponding noise-laden sample values to satisfy the error tolerance level.
15 . A non-transitory computer-readable medium having computer-executable instructions stored thereon that, upon execution, cause one or more processors to perform operations comprising:
generating a query result corresponding to a user query for a dataset;
determining a magnitude range of the query result;
adjusting at least one of a first tunable parameter within a first range and a second tunable parameter within a second range of a differential privacy (DP) configuration;
calibrating an amount of random noise data based on the magnitude range and at least one of the first tunable parameter and the second tunable parameter;
generating a noise-laden query result by applying the amount of random noise data to the query result to satisfy an error tolerance level, thereby perturbing the query result before releasing the query result, and thereby providing data security for the dataset; and
caching the query result, the magnitude range of the query result, and the error tolerance level in an expandable search grid data structure.
16 . The computer-readable medium of claim 15 , wherein the operations further comprise:
generating a plurality of random sample values within the magnitude range of the query result;
generating a random sample noise for each of the plurality of random sample values in iterations by varying the first tunable parameter within the first range or varying the second tunable parameter within the second range;
generating noise-laden sample values by applying the random sample noise to corresponding ones of the plurality of random sample values; and
determining a value for at least one of the first tunable parameter and the second tunable parameter from the iterations to allow corresponding noise-laden sample values to satisfy the error tolerance level.
17 . The computer-readable medium of claim 15 , wherein the operations further comprise:
receiving a second user query for the dataset;
generating a second query result for the second user query;
determining a second magnitude range of the second query result; and
determining a second value for at least one of the first tunable parameter and the second tunable parameter by traversing through the expandable search grid data structure based on the second magnitude range of the second query result and a second error tolerance level.