Collaborative AI Privacy Mapping for Overlapping Transaction Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Collaborative learning between parties to build AI models using shared datasets often compromises privacy due to the need for data sharing, which is constrained by complex privacy regulations and the challenge of managing overlapping data across interacting entities.
Innovation Solution
A method that assesses data intersection between parties, evaluates AI application instructions, and computes a mapping of data to operands with associated privacy metrics, treating overlapping data differently to improve privacy, while executing the AI application and outputting parameters for a trained version.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If data is shared between parties for collaborative AI learning, then model accuracy is improved, but privacy is compromised
Solution Approach 1:
The patent segments data into overlapping and non-overlapping portions, applying different processing strategies to each. Non-overlapping data is processed first with higher privacy protection, while overlapping data is processed later with reduced privacy constraints, enabling collaborative learning while preserving privacy for unique data points.
Solution Approach 2:
The patent performs preliminary processing of non-overlapping data points before processing overlapping data. By extracting features and creating embeddings for non-overlapping data first, the system establishes a privacy-protected foundation that prevents subsequent privacy leakage when overlapping data is shared for collaborative modeling.
2Ease of operation
If traditional collaborative learning is used, then model training is simplified, but privacy protection is insufficient
Solution Approach 1:
The patent introduces a privacy processor as an intermediary component that sits between data sources and the collaborative learning system. This processor automatically assesses data intersections, determines privacy levels, and applies appropriate protection mechanisms, maintaining ease of operation while significantly improving privacy protection without requiring complex manual configuration.
3Ease of operation
If all data points are processed equally, then processing is straightforward, but privacy leakage increases for non-shared data
Solution Approach 1:
The patent applies local quality by treating different data points with different privacy levels based on their overlap characteristics. Non-overlapping data points receive higher privacy protection and are processed with additional safeguards, while overlapping data points are processed with standard protocols, optimizing the balance between privacy protection and processing efficiency for each specific data point.
Data Source
AI summary
A data intersection is assessed of data to be used between at least two parties. The data is to be used in an artificial intelligence (AI) application. Evaluation is performed of set of instructions required for the AI application, where the evaluation creates a modified set of instructions where operands are symbolically associated with corresponding privacy levels. Using the assessed data intersection and the modified set of instructions, a mapping is created from the data to operands with associated privacy metrics. The mapping treats overlapping data from the assessed data intersection differently from data that is not overlapping to improve privacy relative to without the mapping. The AI application is executed using the data to produce at least one parameter of the AI application. The at least one parameter is output for use for a trained version of the AI application. Apparatus, methods, and computer program products are described.