Data Sketching for Secure Multi-Party Analytics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Organizations face challenges in combining data from various sources without incurring risks, often resulting in unrealized value due to inadequate risk mitigation in data sharing, which can lead to unintended consequences.
Innovation Solution
A method and system for securely sharing data between data owners using mapping information and data-sketches, generated based on proprietary data, to minimize risks while maintaining data quality, employing deterministic sampling and value obfuscation techniques.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If organizations share proprietary data directly to combine data and derive value, then data quality and insights are improved, but security risks and potential harm increase
Solution Approach 1:
The patent introduces data sketches as an intermediary representation that enables data sharing without exposing raw proprietary data. The sketches capture essential statistical properties and patterns needed for analytics while being computationally derived transformations that do not contain the original sensitive information, thus mediating between data utility and security requirements
Solution Approach 2:
The patent creates simplified copies of the original data in the form of data sketches that preserve key analytical properties. These sketches are computationally derived representations that capture statistical moments and patterns necessary for insights while being fundamentally different from and less sensitive than the original proprietary data
2Reliability
If organizations avoid sharing data to minimize risks, then security is improved, but data value and productivity are reduced
Solution Approach 1:
The patent extracts essential analytical properties from the raw proprietary data to create data sketches. This extraction process separates the valuable statistical information needed for insights from the sensitive raw data, allowing organizations to share only the extracted properties while retaining security controls over the original data
Solution Approach 2:
The patent transforms data from its original form into a different parameter representation through computational sketching. This parameter transformation changes the data from detailed raw records into aggregated statistical representations that maintain analytical value while reducing security risks associated with sharing
3Loss of information
If comprehensive data sharing is implemented without proper risk mitigation, then data insights are improved, but unintended consequences and harm increase
Solution Approach 1:
The patent applies preliminary anti-action by pre-processing data into sketches that inherently limit the potential for harm before sharing occurs. The sketching process itself acts as a protective measure that prevents raw sensitive data from being exposed, thereby preemptively counteracting potential unintended consequences of data sharing
Data Source
AI summary
The present teaching relates to a method and system for securely sharing data between a group of data owners. A data owner generates mapping information in accordance with a model. The data owner generates a first data-sketch corresponding to proprietary data associated with the data owner. The mapping information and the first data-sketch are transmitted by the data owner to other data owners in the group of data owners. The data owner receives, from each of the other data owners, a second data-sketch corresponding to proprietary data associated with the other data owner, wherein the second data-sketch is generated based on the mapping information. The data owner processes the first data-sketch and the second data-sketches to generate combined data.