Cross-Domain Machine Learning With MPC for User Privacy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning models trained across multiple sources face challenges in handling private or sensitive data without sharing or leaking it to other parties, particularly in the context of cross-domain data collection and the use of third-party cookies, which compromises user privacy.
Innovation Solution
A privacy-preserving machine learning platform utilizing secure multi-party computation (MPC) trains models using encrypted user profiles and contextual signals, ensuring that no single party can access the complete user data in cleartext, thereby maintaining user privacy while enabling effective selection of digital components based on predicted performance measures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If user data is shared across multiple sources for machine learning training, then model training quality is improved, but user privacy is compromised due to data leakage
Solution Approach 1:
The patent segments user data into multiple independent parts (secret shares) that are distributed to different MPC computers. Each computer holds only a fragment of the original data, making it impossible to reconstruct user profiles from any single computer. This segmentation enables collaborative model training across multiple sources while preserving user privacy through cryptographic secret sharing techniques.
Solution Approach 2:
The patent introduces secure multi-party computation (MPC) as an intermediary mechanism that enables multiple parties to jointly train machine learning models without directly sharing their data. The MPC protocol acts as a trusted intermediary that coordinates computation across distributed computers while ensuring that no single party can access the raw data of others, thus resolving the contradiction between data sharing and privacy protection.
2Productivity
If third-party cookies are used for tracking user behavior, then cross-domain data collection is improved, but system reliability deteriorates due to deprecation of third-party cookies
Solution Approach 1:
The patent replaces the deprecated third-party cookie mechanism with a modern alternative based on encrypted data transmission and secure multi-party computation. Instead of relying on browser-based cookies that are being phased out, the system uses cryptographic protocols to achieve the same cross-domain tracking and data collection objectives, ensuring continued reliability and functionality.
Solution Approach 2:
The patent changes the fundamental parameters of data collection from using browser cookies (which are being deprecated) to using encrypted data structures and cryptographic protocols. This parameter change includes transitioning from client-side storage (cookies) to secure server-side computation with encrypted data transmission, maintaining cross-domain functionality while adapting to modern privacy standards and browser security policies.
3Measurement precision
If all user data is transmitted to the machine learning platform, then model accuracy is improved, but bandwidth usage increases and data transmission efficiency deteriorates
Solution Approach 1:
The patent extracts and transmits only the essential features and aggregated statistics needed for model training, rather than transmitting complete user datasets. By taking out only the necessary information (encrypted feature vectors, aggregated metrics) and leaving the detailed raw data on user devices, the system maintains model accuracy while significantly reducing bandwidth consumption and transmission overhead.
Data Source
AI summary
This document describes a secure machine learning platform. In some aspects, a method includes transmitting by the application to the machine learning platform, a set of data including a user profile, one or more characteristics of a digital component, contextual signals, model identifier, and data indicating a type of event. The application receives a request generated based on the computer-readable instructions to upload a user profile of a user of the client device to a machine learning platform. The computer-readable instructions initiate the request in response to detecting an occurrence of the event with the digital component. In response to the request, the application can obtain the user profile request data element that includes a model identifier for a machine learning model and one or more characteristics of at least one of the digital component or the first content page.