Synthetic Data Valuation for Private Dataset Transfer
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge of securely and privately evaluating and transferring datasets in a data marketplace is hindered by the risk of intellectual property and privacy violations due to unauthorized access and replication of digital assets, making sellers and buyers hesitant to share their datasets.
Innovation Solution
A method involving the generation of synthetic data samples that represent original data samples, using a machine learning model trained on a second entity's dataset, allows for secure and private data valuation without disclosing the original data, by optimizing utility and security objectives through a common model architecture and training algorithm, ensuring that the synthetic data is perceptually unintelligible and ineffective for statistical analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If datasets are transferred between buyers and sellers for valuation, then data utility and valuation accuracy are improved, but intellectual property and privacy security deteriorate due to unauthorized access and replication risks
Solution Approach 1:
The patent applies copying by creating synthetic data samples that replicate the statistical properties and utility of original datasets without copying the actual data. The synthetic data is generated through a process that mimics the distribution and characteristics of the source data, enabling valuation accuracy while preventing IP and privacy violations since the synthetic copies cannot be reverse-engineered to reveal the original sensitive information
Solution Approach 2:
The patent introduces an intermediary mechanism - a trusted facilitator or secure computation platform - that mediates the valuation process. This intermediary enables the computation of utility metrics and valuation without direct data transfer between buyers and sellers, thus maintaining valuation accuracy while eliminating the security risks associated with direct data sharing
2Loss of information
If original data is shared for valuation computation, then data utility is improved, but privacy protection deteriorates due to exposure of sensitive information
Solution Approach 1:
Instead of sharing original sensitive data, the patent creates synthetic data copies that preserve the statistical properties and utility needed for valuation. These synthetic copies maintain the same distribution characteristics, correlations, and patterns as the original data, ensuring that utility computation remains accurate while the sensitive original information never leaves the owner's secure environment
Solution Approach 2:
The patent applies local quality by ensuring that different parts of the data ecosystem maintain different security properties. The original sensitive data remains in the seller's secure environment with full privacy protection, while the synthetic data is shared with buyers for valuation purposes. Each entity operates with the appropriate level of data access needed for their function, optimizing both utility and privacy locally
3Object-affected harmful factors
If synthetic data is generated to protect privacy, then security is improved, but data quality may deteriorate if synthetic data does not accurately represent original data
Solution Approach 1:
The patent applies parameter changes by carefully controlling the generation parameters of synthetic data to ensure they match the statistical properties of the original data. The synthetic data generation process adjusts parameters such as distribution characteristics, correlations, and feature relationships to maintain high fidelity representation of the original data's utility while preserving privacy through the synthetic nature of the generated samples
Data Source
AI summary
Systems, methods and computer-readable media for generating, by a first entity, a set of synthetic data samples that represent a corresponding set of original data samples; sending, by the first entity, the set of synthetic data samples for use by a second entity to generate a set of second entity predictions for the set of synthetic data samples using a machine learning (ML) model that has been trained using a second entity dataset; sending, by the first entity, for a third entity, a set of trusted labels corresponding to the set of original data samples; and receiving, by the first entity, from the third entity, valuation information for the second entity dataset that is based on a comparison by the third entity of the set of trusted labels and the set of second entity predictions.