Privacy-Preserving ID Intersection for Neural Network Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing secure multi-party computing protocols expose customer IDs during the calculation of data set intersections, compromising ID privacy and security.
Innovation Solution
Implement a method where computing participants use secret sharing and OPPRF to determine intersection IDs and feature values, using random numbers for non-intersection IDs, ensuring only intersection data is used for AI model training, with threshold-based determination to optimize data usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If PSI technology is used to calculate intersection of IDs, then data set intersection can be calculated without exposing data beyond intersection, but ID plaintext in the intersection is exposed leading to poor security
Solution Approach 1:
The patent segments the ID into multiple shares using secret sharing scheme, where no single party can reconstruct the original ID alone. This divides the confidential information into independent parts that only become meaningful when combined, thus protecting ID privacy while enabling intersection calculation.
Solution Approach 2:
The patent introduces an intermediary mechanism (secret sharing protocol) between the parties to facilitate intersection calculation without direct exposure of sensitive data. The intermediary structure ensures that even during the computation process, the actual ID values remain hidden from all participants.
2Productivity
If all IDs are processed for AI model training, then comprehensive data is available, but unnecessary computations and costs increase for non-intersection data
Solution Approach 1:
The patent extracts only the intersection IDs from the complete ID sets before proceeding to AI model training. By removing non-intersection data beforehand, the system avoids performing unnecessary computations on irrelevant data, thus improving training efficiency and reducing computational costs.
Solution Approach 2:
The patent performs preliminary filtering to identify and isolate intersection IDs before the main AI model training process. This preliminary action ensures that only relevant data enters the training pipeline, preventing waste of computational resources on non-intersection data.
Data Source
AI summary
A data processing method includes obtaining first ciphertext data. The second computing participant obtains, from the first computing participant, a share of a first feature value corresponding to a first identifier (ID). The first computing participant obtains second data, where if the first ID is an intersection ID, the second data is a share of a second feature value that is in the second computing participant and that corresponds to the first ID. If the first ciphertext data indicates that the first ID is an intersection ID, the first computing participant uses the second data and a share of the first feature value that is held by the first computing participant as training data of a neural network.