Secure MPC User Categorization With Nearest-Neighbor Privacy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning models trained on data from multiple sources risk exposing private or sensitive user data when third-party cookies are blocked, necessitating a privacy-preserving and secure method for user categorization.
Innovation Solution
Implementing secure multi-party computation (MPC) techniques to train machine learning models on encrypted user profiles at client devices, using centroid and k-NN models to categorize users into groups without revealing cleartext data, allowing for accurate user group expansion and content distribution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning models are trained on data from multiple sources to improve categorization accuracy, then user categorization accuracy is improved, but user privacy and data security deteriorate due to exposure of sensitive data
Solution Approach 1:
The patent segments user data into multiple encrypted shares distributed across different MPC systems. Each system holds only a portion of the data in encrypted form, making it impossible to reconstruct sensitive information without collaboration. This allows training on distributed data while preserving privacy, resolving the contradiction between accuracy improvement and privacy protection.
Solution Approach 2:
The patent introduces secure multi-party computation (MPC) as an intermediary mechanism that enables collaborative model training without direct data sharing. The MPC protocol acts as a trusted mediator that allows multiple parties to jointly compute machine learning models on their respective data while keeping the data encrypted and private throughout the process.
2Loss of information
If third-party cookies are used to collect user data across websites, then user profile data availability is improved, but data security and user privacy control worsen
Solution Approach 1:
The patent inverts the traditional approach by instead of collecting data and then protecting it, the system protects data first through encryption and secret sharing, then enables controlled access for training. This inversion allows data to remain secure while still being available for legitimate purposes through the MPC framework.
3Object-affected harmful factors
If encrypted data is used for model training to preserve privacy, then user privacy is improved, but computational complexity and processing time worsen
Solution Approach 1:
The patent performs preliminary data preparation by encrypting and segmenting user data into shares before the training process begins. This preliminary action ensures privacy is established upfront, and the subsequent training operations work with pre-processed encrypted shares, reducing the computational burden during the actual training phase.
Data Source
AI summary
A first multi-party computation (MPC) system of an MPC cluster can receive, from an application on a client device, an inference request comprising a first share of a given user profile for a user of the application and a performance threshold. A set of nearest neighbors to the user profile can be identified by performing a secure MPC process using a trained machine learning model in collaboration with one or more second MPC systems. One or more nearest neighbors having a performance measure that satisfies the performance threshold can be selected from the set of nearest neighbors. The first MPC system can transmit data derived from the one or more nearest neighbors to the application.