Synthetic Data Generation for Privacy-Preserving Application Training

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

The existing technologies for mobile devices fail to effectively address user privacy concerns by allowing applications to access sensitive user data, which can lead to hacking and data leaks, as they often require real user information to function properly.

Innovation Solution

A system comprising a secure data module and a synthetic data generating module that creates a synthetic dataset based on user data, allowing applications to access this synthetic data instead of real sensitive information, with the synthetic data generating module being stored in a trusted or untrusted zone of the device and periodically updated using transfer learning.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If applications access real sensitive user data to function properly, then application functionality is improved, but user privacy security deteriorates

Engineering Contradiction:

Improveapplication functionalityVSAvoiduser privacy security

Core Design Contradiction:

Ease of operationVSObject-affected harmful factors

Solution Approach 1:

The patent creates synthetic copies of real user data that preserve the statistical properties and patterns needed for application functionality while containing no actual sensitive information. The synthetic data generation module produces artificial datasets that mimic the structure and relationships of real user data, allowing applications to operate normally without accessing genuine personal information.

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The synthetic data generation module acts as an intermediary between real user data and applications. It receives real sensitive data, transforms it into synthetic equivalents, and provides these to applications instead of direct access to real data. This intermediary layer preserves application functionality while blocking exposure of actual sensitive information.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Object-affected harmful factors

If synthetic data is used instead of real sensitive data, then user privacy security is improved, but data accuracy deteriorates

Engineering Contradiction:

Improveuser privacy securityVSAvoiddata accuracy

Core Design Contradiction:

Object-affected harmful factorsVSMeasurement precision

Solution Approach 1:

The patent transforms data parameters by changing the representation of user information while preserving statistical properties. The synthetic data generation module modifies data parameters such as creating artificial variations in user profiles, preferences, and behaviors that maintain the underlying distributions and relationships, ensuring applications receive data with comparable analytical value to real data.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS11755771B2System, method, and computer-accessible medium for training models on mixed sensitivity datasets

Publication Date: 2023.09.12 CAPITAL ONE SERVICES LLC

AI summary

A system can include, for example, a secure data module(s) configured to store sensitive data regarding the user(s), a synthetic dataset generating module(s) configured to generate the synthetic dataset based on the sensitive data, and a control module configured to receive a request from an application for a dataset related to the user(s), provide the request to the synthetic dataset generating module(s), receive the synthetic dataset from the synthetic dataset generating module(s), and provide the synthetic dataset to the application. The synthetic dataset generating module(s) can be configured to generate the synthetic dataset based on the dataset.