# US 2024/0202360 A1 · Privacy preserving custom embeddings

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

- Number: US 2024/0202360 A1
- Document ID: US20240202360A1
- Status: Published application
- Kind: A1
- Application number: 18/388,763
- Filed: 2023-11-10
- Published: 2024-06-20
- Earliest claimed: 2022-12-16
- Assignees: Google Inc.
- Applicants: Google LLC
- Inventors: Gang Wang; Alexander E. Mayorov
- Web page: https://patentgrep.com/patent/US20240202360A1

## CPC

- G06F 21/6245: {Protecting personal data, e.g. for financial or medical purposes}
- G06F 21/53: by executing in a restricted environment, e.g. sandbox or secure virtual machine
- G06N 3/045: Combinations of networks
- G06N 3/098: Distributed learning, e.g. federated learning
- H04L 9/085: {Secret sharing or secret splitting, e.g. threshold schemes}
- H04L 2209/46: Secure multiparty computation, e.g. millionaire problem
- H04L 2209/60: Digital content management, e.g. content distribution
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M)

## Abstract

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for selecting and distributing digital components to client devices in ways that protect user privacy and confidential data of content platforms and/or digital component providers are described. In one aspect, a method includes receiving, by a secure distribution system and from a client device of a user, a digital component request that includes, for each of multiple content platforms that distribute digital components to users, a corresponding user embedding comprising weights indicative of the relevance of multiple features to the user. The secure distribution system provides each user embedding as input to a respective isolated execution environment for the content platform corresponding to the user embedding, wherein the secure distribution system hosts each isolated execution environment. Digital component selection data generated based on the user embedding is received from each isolated execution environment.

## Claims

### Claim 1

A computer-implemented method comprising:
receiving, by a secure distribution system and from a client device of a user, a digital component request comprising, for each of multiple content platforms that distribute digital components to users, a corresponding user embedding comprising weights indicative of the relevance of multiple features to the user, wherein each user embedding is different from each other user embedding;
providing, by the secure distribution system, each user embedding as input to a respective isolated execution environment for the content platform corresponding to the user embedding, wherein the secure distribution system hosts each isolated execution environment;
receiving, from each isolated execution environment, digital component selection data generated based on the user embedding provided as input to the isolated execution environment;
selecting, by the secure distribution system, a given digital component based on the digital component selection data received from each isolated execution environment; and
sending, by the secure distribution system, the given digital component to the client device for presentation to the user.

### Claim 2 (depends on 1)

The computer-implemented method of claim 1, wherein the digital component selection data received from each isolated execution environment comprises data identifying a selection parameter for a candidate digital component.

### Claim 3 (depends on 2)

The computer-implemented method of claim 2, wherein selecting the given digital component comprises selecting the digital component from a set of candidate digital components based on the selection parameter received from each isolated execution environment.

### Claim 4 (depends on 1)

The computer-implemented method of claim 1, wherein the client device generates each user embedding based on user activity within one or more applications executed by the client device.

### Claim 5 (depends on 4)

The computer-implemented method of claim 4, wherein generating the corresponding user embedding for each content platform comprises:
obtaining, for each event that occurs at the one or more applications, one or more features related to the event;
obtaining, for each feature, a transformed feature using a transformation function of the content platform; and
generating, using an aggregation function of the content platform, the user embedding based on the one or more features for each of multiple events.

### Claim 6 (depends on 5)

The computer-implemented method of claim 5, wherein the transformation function is configured to transform generic event-based features within a taxonomy of features to content platform specific features of a custom taxonomy of the content platform.

### Claim 7 (depends on 5)

The computer-implemented method of claim 5, wherein obtaining the one or more features related to the event comprises providing information related to the event as input to a machine learning model executed by the client device and receiving the one or more features as an output of the machine learning model.

### Claim 8 (depends on 5)

The computer-implemented method of claim 5, wherein the transformation function comprises a machine learning model.

### Claim 9 (depends on 5)

The computer-implemented method of claim 5, wherein obtaining the transformed feature for each feature comprises sending each feature to one or more retrieval servers and receiving each transformed feature from the one or more retrieval servers.

### Claim 10 (depends on 9)

The computer-implemented method of claim 9, wherein the one or more retrieval servers comprises a set of multi-party computation servers and wherein sending each transformed feature comprises sending a respective secret share of each transformed feature to each multi-party computation server.

### Claim 11

A system comprising:
one or more processors; and
one or more storage devices storing instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:
receiving, by a secure distribution system and from a client device of a user, a digital component request comprising, for each of multiple content platforms that distribute digital components to users, a corresponding user embedding comprising weights indicative of the relevance of multiple features to the user, wherein each user embedding is different from each other user embedding;
providing, by the secure distribution system, each user embedding as input to a respective isolated execution environment for the content platform corresponding to the user embedding, wherein the secure distribution system hosts each isolated execution environment;
receiving, from each isolated execution environment, digital component selection data generated based on the user embedding provided as input to the isolated execution environment;
selecting, by the secure distribution system, a given digital component based on the digital component selection data received from each isolated execution environment; and
sending, by the secure distribution system, the given digital component to the client device for presentation to the user.

### Claim 12 (depends on 11)

The system of claim 11, wherein the digital component selection data received from each isolated execution environment comprises data identifying a selection parameter for a candidate digital component.

### Claim 13 (depends on 12)

The system of claim 12, wherein selecting the given digital component comprises selecting the digital component from a set of candidate digital components based on the selection parameter received from each isolated execution environment.

### Claim 14 (depends on 11)

The system of claim 11, wherein the client device generates each user embedding based on user activity within one or more applications executed by the client device.

### Claim 15 (depends on 14)

The system of claim 14, wherein generating the corresponding user embedding for each content platform comprises:
obtaining, for each event that occurs at the one or more applications, one or more features related to the event;
obtaining, for each feature, a transformed feature using a transformation function of the content platform; and
generating, using an aggregation function of the content platform, the user embedding based on the one or more features for each of multiple events.

### Claim 16 (depends on 15)

The system of claim 15, wherein the transformation function is configured to transform generic event-based features within a taxonomy of features to content platform specific features of a custom taxonomy of the content platform.

### Claim 17 (depends on 15)

The system of claim 15, wherein obtaining the one or more features related to the event comprises providing information related to the event as input to a machine learning model executed by the client device and receiving the one or more features as an output of the machine learning model.

### Claim 18 (depends on 15)

The system of claim 15, wherein the transformation function comprises a machine learning model.

### Claim 19 (depends on 15)

The system of claim 15, wherein obtaining the transformed feature for each feature comprises sending each feature to one or more retrieval servers and receiving each transformed feature from the one or more retrieval servers.

### Claim 20

A non-transitory computer readable medium carrying instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:
receiving, by a secure distribution system and from a client device of a user, a digital component request comprising, for each of multiple content platforms that distribute digital components to users, a corresponding user embedding comprising weights indicative of the relevance of multiple features to the user, wherein each user embedding is different from each other user embedding;
providing, by the secure distribution system, each user embedding as input to a respective isolated execution environment for the content platform corresponding to the user embedding, wherein the secure distribution system hosts each isolated execution environment;
receiving, from each isolated execution environment, digital component selection data generated based on the user embedding provided as input to the isolated execution environment;
selecting, by the secure distribution system, a given digital component based on the digital component selection data received from each isolated execution environment; and
sending, by the secure distribution system, the given digital component to the client device for presentation to the user.
