# US 2023/0274028 A1 · Building data-secure group profiles

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2023/0274028 A1
- Document ID: US20230274028A1
- Status: Published application
- Kind: A1
- Application number: 17/683,269
- Filed: 2022-02-28
- Published: 2023-08-31
- Earliest claimed: 2022-02-28
- Assignees: Microsoft Technology Licensing, LLC
- Applicants: Microsoft Technology Licensing, LLC
- Inventors: Ryan M. Rogers; Mingyang Hu; Adrian Rivera Cardoso; Raymond E. Ortigas; Humberto A. Gonzalez Cabezas
- Web page: https://patentgrep.com/patent/US20230274028A1

## CPC

- G06F 21/6254: {by anonymising data, e.g. decorrelating personal data from the owner's identification}
- G06F 21/6227: {where protection concerns the structure of data, e.g. records, types, queries}
- G06F 21/6245: {Protecting personal data, e.g. for financial or medical purposes}
- G06F 21/6263: {during internet communication, e.g. revealing personal data from cookies}
- H04L 67/306: User profiles
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M)

## Abstract

Technologies for digital content distribution include creating a data set that includes a plurality of entity data records each comprising a plurality of attribute values. By applying a first data security technique to the data set, a first query term and a matching subset of the plurality of entity data records are determined. A second data security technique is applied to the matching subset. When output of the second data security technique satisfies a first noisy threshold, the first query term is added to a set of query terms. A third data security technique is applied to the matching subset. When output of the third data security technique satisfies a second noisy threshold, the set of query terms is expanded to include a second query term. The set of query terms is used to perform the digital content distribution.

## Claims

### Claim 1

A method for creating a group profile with data security, the method comprising:
creating a data set that includes a plurality of entity data records each comprising a plurality of attribute values;
ordering a list of query terms of size m in descending order of value;
the value is based on conversions associated with a query term;
selecting a first query term from the ordered list of query terms;
(a) splitting the data set into a matching subset and a non-matching subset by applying the first query term to the data set;
the matching subset includes data records that each include an attribute value that matches the first query term;
the non-matching subset includes data records that do not include an attribute value that matches the first query term;
(b) determining function output for each of the matching subset and the non-matching subset by applying a function to both the matching subset and the non-matching subset; and
(c) when the function output for the matching subset and the function output for the non-matching subset are both greater than or equal to a first threshold k and the function output for the matching subset plus first noise Zi is greater than or equal to a second threshold T plus second noise Z, (i) adding the first query term to the group profile, (ii) storing a mapping of the first query term to the matching subset, (iii) generating a signal that the data set can be split into a first subgroup that has a first group profile that includes the query term and a second subgroup that has a second group profile that does not include the query term, and (iv) when at least an amount t*noisy size ñ of entity data records in the matching subset contain a second query term different from the first query term, expanding the group profile to include the second query term.

### Claim 2 (depends on 1)

The method of claim 1, further comprising (d) when the function output for the matching subset is less than the first threshold k or the function output for the matching subset is greater than or equal to the first threshold k and the function output for the matching subset plus the first noise Zi is less than the second threshold T plus the second noise Z, (i) generating a signal that the matched subset cannot be split for the first query term, and (ii) skipping the first query term.

### Claim 3 (depends on 2)

The method of claim 2, further comprising:
(e) selecting a next query term from the ordered list of query terms; and
repeating (a), (b), (c), (d), and (e) for the next query term until the number of repetitions is equal to the size m or the function output for the matching subset is less than the first threshold k.

### Claim 4 (depends on 1)

The method of claim 1, further comprising:
receiving a request from a third party application;
matching the group profile to the request;
selecting a content distribution based on the matched group profile; and
providing the content distribution to a user system that is associated with the group profile.

### Claim 5 (depends on 1)

The method of claim 1, further comprising:
generating, by the function, as the function output, a first true count of entity data records in the matching subset and a second true count of entity data records in the non-matching subset; and
determining whether both the first true count and the second true count are greater than or equal to the first threshold k.

### Claim 6 (depends on 1)

The method of claim 1, further comprising determining a data security requirement for the group profile and setting the first threshold k based on the data security requirement.

### Claim 7 (depends on 1)

The method of claim 1, further comprising computing the second threshold T as a function of the first threshold k, at least one first differential privacy parameter E, a second differential privacy parameter 4, a third differential privacy parameter δ, and the size m.

### Claim 8 (depends on 1)

The method of claim 1, further comprising setting the second threshold T to a value that is greater than the first threshold k.

### Claim 9 (depends on 1)

The method of claim 1, further comprising computing the first noise Z, by sampling a first Laplace distribution created using at least one first differential privacy parameter and computing the second noise Z by sampling a second Laplace distribution created using at least one second differential privacy parameter.

### Claim 10 (depends on 1)

The method of claim 1, wherein a scale or standard deviation of the second noise Z is greater than a scale or standard deviation of the first noise Zi.

### Claim 11 (depends on 1)

The method of claim 1, further comprising changing at least one of the first threshold k, the second threshold T, and the first noise Zi for the next query term in the ordered list of query terms.

### Claim 12 (depends on 1)

The method of claim 1, further comprising keeping the value of the second noise Z the same for each next query term in the ordered list of query terms until the signal indicates that a split can occur.

### Claim 13 (depends on 1)

The method of claim 1, further comprising determining the amount t based on a frequency of occurrence of the second query term in the matching subset.

### Claim 14 (depends on 1)

The method of claim 1, further comprising computing the third noise ñ by sampling a Gaussian distribution or by sampling a Laplace distribution.

### Claim 15

A method comprising:
creating a data set that includes a plurality of entity data records each comprising a plurality of attribute values;
splitting the data set into a matching subset and a non-matching subset;
the matching subset includes data records that each include an attribute value that matches a first query term of a plurality of different query terms;
the non-matching subset includes data records that do not include any attribute values that match the first query term;
applying a function to each of the subsets to produce function output for each of the subsets;
when the function output for each of the subsets is greater than or equal to a first threshold and the function output for the matching subset plus first noise is greater than or equal to a second threshold plus second noise, generating a signal that the data set can be split into a first subgroup that has a first group profile that includes the first query term and a second subgroup that has a second group profile that does not include the query term; and
when the function output for the matching subset is less than or equal to the first threshold or the function output for the matching subset is greater than or equal to the first threshold and the function output for the matching subset plus the first noise is less than the second threshold plus the second noise, generating a signal that the data set cannot be split into the first subgroup and the second subgroup.

### Claim 16 (depends on 15)

The method of claim 15, further comprising matching the first subgroup of the data set to a query that comprises the first query term.

### Claim 17 (depends on 15)

The method of claim 15, further comprising, when the function output for each of the subsets is greater than or equal to a first threshold and the function output for the matching subset plus first noise is greater than or equal to a second threshold plus second noise, adding the first query term to a group profile.

### Claim 18 (depends on 15)

The method of claim 15, further comprising, when the function output for the matching subset is less than the first threshold or the function output for the matching subset is greater than or equal to the first threshold and the function output for the matching subset plus the first noise is less than the second threshold plus the second noise, (i) skipping the first query term and (ii) selecting a next query term from the plurality of different query terms.

### Claim 19 (depends on 15)

The method of claim 15, further comprising, when the function output for each of the subsets is greater than or equal to the first threshold and the function output for the matching subset plus first noise is greater than or equal to the second threshold plus the second noise and at least a third threshold plus third noise of entity data records in the matching subset contain a second query term different from the first query term, creating a group profile that includes the first query term and the second query term.

### Claim 20

A method for digital content distribution comprising:
creating a data set that includes a plurality of entity data records each comprising a plurality of attribute values;
by applying a first data security technique to the data set, determining a first query term and a matching subset of the plurality of entity data records;
applying a second data security technique different than the first data security technique to the matching subset;
when output of the second data security technique satisfies a first noisy threshold, adding the first query term to a set of query terms;
applying a third data security technique different than the first and second data security techniques to the matching subset;
when output of the third data security technique satisfies a second noisy threshold, expanding the set of query terms to include a second query term; and
using the set of query terms to perform the digital content distribution.
