Secure and private hyper-personalization system and method

Patent metadata

Field	Value
Publication number	US-12353595-B2
Application number	US-202318160744-A
Country	US
Kind code	B2
Filing date	Jan 27, 2023
Priority date	Aug 23, 2019
Publication date	Jul 8, 2025
Grant date	Jul 8, 2025

Abstract

A secured virtual container is enabled to securely store personal data corresponding to a user, where such data is inaccessible to processes running outside the secured virtual container. The secured virtual container may also include an execution environment for a machine learning model where the model is securely stored and inaccessible. Personal data may be feature engineered and provided to the machine learning model for training purposes and/or to generate inference values corresponding to the user data. Inference values may thereafter be relayed by a broker application from the secured virtual container to applications external to the container. Applications may perform hyper-personalization operations based at least in part on received inference values. The broker application may enable external applications to subscribe to notifications regarding availability of inference values. The broker may also provide inference values in response to a query.

First claim

What is claimed is:

1. A method comprising:
selecting a set of features from feature data stored in a secured virtual environment of a computing device, the secured virtual environment comprising a secured data processor executing in isolation from an operating system of the computing device, the feature data comprising image data of at least one of a user or an environment of the user and audio data of at least one of the user or the environment of the user;
providing the feature data to a machine learning (ML) model, the ML model being implemented within the secured virtual environment by the secured data processor and performing data obfuscation operations for data specific to the user and being trained using the data specific to the user;
generating, by the ML model, an inference value for an inference category based on the image data and the audio data, the inference value representing an accuracy of a proposition about the user; and
causing the inference value to be provided to a process executing in the operating system implemented outside of the secured virtual environment, wherein the process is configured to personalize at least one of content or an interface of the computing device for the user based on the inference value.

2. The method of claim 1 , wherein the secured virtual environment is a virtual machine and the virtual machine and the operating system of the computing device execute in parallel via a shared hypervisor.

3. The method of claim 1 , wherein the feature data further comprises user data that includes at least one of:
a risk profile;
a financial profile; or
an application usage profile.

4. The method of claim 3 , wherein the user data further includes at least one of:
habit information for the user;
relationship information for the user; or
demographic information for the user.

5. The method of claim 1 , wherein the feature data is stored in a graph comprising a plurality of logical layers including at least two of:
a policy layer comprising at least one of policies, rules, or values for a personalization system;
a knowledge graph comprising user data for the user; or
a transient layer comprising a rolling window of signals captured by the computing device.

6. The method of claim 1 , wherein:
the ML model performs obfuscation operations on the feature data, the obfuscation operations preventing the feature data from being exposed outside of the secured virtual environment; and
the inference value is received based on the obfuscation operations.

7. The method of claim 1 , wherein causing the inference value to be provided to the process comprises:
providing the inference value to a broker external to the secured virtual environment; and
providing, by the broker, the inference value to the process.

8. The method of claim 7 , wherein the process has subscribed to receive new and changed inference values for inference categories associated with the user, the inference categories including the inference category.

9. The method of claim 1 , wherein the inference category corresponds to at least one of:
a location of the user;
a display characteristic of the computing device; or
a configuration preference of the user.

10. The method of claim 1 , wherein the current usage further indicates at least one of:
policy violations on the computing device; or
atmospheric data of the computing device.

11. The method of claim 1 , wherein the current usage further indicates:
whether additional users are present with the user.

12. The method of claim 1 , wherein the proposition is related to at least one of:
a location of the user; or
a persona of the user.

13. A system comprising:
a processing system; and
memory coupled to the processing system, the memory comprising computer executable instructions that, when executed, perform operations comprising:
selecting, by a personalization data processor implemented in a secured virtual environment of a computing device, a set of features from feature data stored in the secured virtual environment, at least a portion of the secured virtual environment executing within a context of an operating system of the computing device, the feature data comprising policy data for usage of the computing device, the policy data being based on a risk profile for a user of the computing device and indicating a set of actions the user is permitted to perform while operating the computing device;
providing the feature data to a machine learning (ML) model implemented within the secured virtual environment by the personalization data processor and performing data obfuscation operations for data specific to the user, the ML model being trained using the data specific to the user;
generating, by the ML model, an inference value for an inference category based on at least one of the risk profile of the user or the set of actions, the inference value representing an accuracy of a proposition about the user; and
causing the inference value to be provided to a process executing in the operating system and external to the secured virtual environment, wherein the process is configured to personalize an interface of the computing device for the user based on the inference value.

14. The system of claim 13 , wherein the portion of the secured virtual environment is a virtual sandbox that is secured to prevent the operating system from directly accessing the feature data.

15. The system of claim 13 , wherein a personalization data processor of the secured virtual environment is implemented as an interface between the feature data and processes external to the secured virtual environment.

16. The system of claim 15 , wherein the personalization data processor:
maintains a plurality of inference categories associated with the user, the plurality of inference categories including the inference category; and
publishes availability of the plurality of inference categories to entities external to the secured virtual environment.

17. The system of claim 13 , wherein the inference category corresponds to at least one of:
a time of day;
an amount of ambient lighting; or
a habit of the user.

18. The system of claim 13 , wherein personalizing the interface of the computing device comprises altering display characteristics of the interface.

19. The system of claim 13 , wherein the feature data is stored in a graph database system of the secured virtual environment.

20. A device comprising:
a personalization data processor implemented in a secure personalization environment of the device, the personalization data processor performing operations comprising:
selecting features from feature data stored in the secure personalization environment, the secured personalization environment executing in isolation from an operating system of the device, the feature data comprising at least one of image data or audio data of an environment of a user;
providing the feature data to a machine learning (ML) model implemented within the secure personalization environment by the personalization data processor and performing data obfuscation operations for data specific to the user, the ML model being trained using the data specific to the user;
generating, by the ML model, an inference value for an inference category based on the at least one of the image data and the audio data, the inference value representing an accuracy of a location of the user; and
causing the inference value to be provided to a process executing in the operating system, wherein the proces

Assignees

• Microsoft Technology Licensing Llc

Inventors

• Mukhopadhyay Debasish
• Nitta Bryston
• Liang Chen

Classifications

• G06N5/04

  Inference or reasoning models · CPC title

• G06F2009/45587

  Isolation or security of virtual machine instances · CPC title

• G06F9/45558

  Hypervisor-specific management and integration aspects · CPC title

• G06N20/00

  Machine learning · CPC title

• G06N3/084

  Backpropagation, e.g. using gradient descent · CPC title