Trusted Execution Environment for Privacy-Sensitive ML Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning models face challenges in utilizing privacy-sensitive data for training due to privacy restrictions, which prevent human users from accessing or viewing the data, and impose time limits on its usage, hindering effective model training.
Innovation Solution
A privacy-sensitive computing system is implemented, utilizing a code authorization unit and data aggregation unit to create a protected environment. The code authorization unit verifies that the code does not violate privacy restrictions by generating a hash, ensuring only authorized code interacts with the data, while the data aggregation unit aggregates and provides limited access to the data within the protected environment, enforcing privacy restrictions such as preventing data access after a predetermined time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If privacy restrictions are enforced to prevent human users from viewing sensitive data, then privacy protection is improved, but the ability to use the data for training machine learning models deteriorates
Solution Approach 1:
The patent introduces a trusted execution environment (TEE) as an intermediary between the privacy-sensitive data and the machine learning training process. The TEE creates a protected computing space where code can access and process unseeable data without human users being able to view the raw data, thus enabling model training while maintaining privacy restrictions.
Solution Approach 2:
The system segments the computing environment into a protected trusted execution environment and an untrusted external environment. By dividing the data access process into segments (data aggregation, code verification, controlled processing), the patent enables selective access where only authorized code within the TEE can interact with the sensitive data.
2Reliability
If code verification through hash generation is implemented to ensure code integrity, then security against code tampering is improved, but the complexity of the authorization process increases
Solution Approach 1:
The system performs preliminary code verification by generating hashes of the training code before the code is executed within the trusted execution environment. This preliminary action ensures code integrity is checked in advance, preventing tampering while the code runs in the protected environment.
3Reliability
If data aggregation is performed to provide limited access to privacy-sensitive data, then privacy control is improved, but the accessibility of data for processing deteriorates
Solution Approach 1:
The trusted execution environment acts as an intermediary that aggregates privacy-sensitive data and provides controlled access to authorized code. The data aggregation unit within the TEE collects and prepares the data, then makes it accessible only to verified code through secure interfaces, balancing privacy control with processing accessibility.
Data Source
AI summary
The present disclosure describes systems and methods for a privacy sensitive computing system. One or more embodiments provide a protected computing environment, a code authorization unit, and a data aggregation unit. For example, some embodiments of the privacy sensitive computing system may train unsupervised or self-supervised ML models on user-generated assets subject to privacy considerations that mandate those assets are not viewed directly by human eyes.