Distributed Data Collaboration Using Privacy-Preserving Synthetic Datasets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data collaboration systems face challenges in sharing high-quality data without compromising data security and privacy, leading to inaccurate and insufficient insights due to the exclusion of personally identifiable information, and they are prone to data breaches in centralized setups.
Innovation Solution
A distributed secure data collaboration framework that generates synthetic datasets using local generators and a central model, maintaining privacy by keeping sensitive information isolated at local nodes, and utilizing transformers and conditional vector sampling to ensure statistical representativeness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If data is shared between organizations to improve data insights, then analytics and decision making strategies are enhanced, but data security and privacy are compromised
Solution Approach 1:
The system generates synthetic datasets that copy the statistical properties and distribution patterns of real datasets without containing actual sensitive information. The synthetic data maintains the joint distribution of multiple input datasets while being completely fabricated, allowing analysis without exposing real personal information.
Solution Approach 2:
The system introduces an intermediary processing layer that transforms real data into synthetic data through a generator model. This intermediary layer acts as a buffer between the real data and the analysis process, enabling data sharing while preventing direct exposure of sensitive information.
2Object-affected harmful factors
If personally identifiable information is excluded from shared data, then data security is improved, but data quality and representativeness deteriorate
Solution Approach 1:
The system changes the fundamental parameters of the data from real to synthetic while preserving statistical properties. The generator model transforms the data representation to maintain distribution characteristics, correlations, and patterns without retaining identifiable information, thus changing data nature while preserving analytical value.
Solution Approach 2:
The synthetic data copies the statistical structure and distribution patterns of real data without copying actual records or identifiable information. This allows the data to represent real-world patterns while being completely anonymized.
3Ease of operation
If centralized data processing is used to generate synthetic data, then data collaboration is simplified, but data breaches become more vulnerable
Solution Approach 1:
The system segments the data processing into distributed components where each organization's data remains local. The generator model processes data in a distributed manner, allowing each organization to generate synthetic data from their own local datasets without centralizing sensitive information, thus maintaining simplicity while improving security.
4Adaptability or versatility
If raw data is exposed to multiple computing devices for analysis, then data collaboration flexibility is improved, but data security is compromised
Solution Approach 1:
Instead of sharing raw data, the system shares synthetic data copies that replicate the statistical properties needed for analysis. This allows flexible data collaboration where organizations can process and analyze data on their own devices without exposing raw sensitive information to multiple computing devices.
Data Source
AI summary
The present disclosure relates to systems, methods, and non-transitory computer-readable media that implements a secure distributed data collaboration architecture for generating synthetic datasets. For example, the disclosed system sends a request to perform a data collaboration with a first dataset of a first local node and a second dataset of a second local node. The disclosed system receives intermediate feature maps from the local nodes that correspond with the datasets and generates a combined feature map. Further, the disclosed system generates a synthetic dataset from the combined feature map by utilizing a central generative model. Moreover, the synthetic dataset generated by the disclosed system is statistically representative of the first dataset and the second dataset.