Trusted Zone Model Training for Privacy Protection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing model training methods for service providers face challenges in improving model accuracy while protecting user privacy, as direct access to sensitive user data leads to insufficient training samples and potential data leaks.
Innovation Solution
Implementing a model training method that utilizes a trusted zone in a service device to decrypt and process encrypted user data, ensuring that only specified operations, such as decryption and training, occur within this isolated environment, preventing data exposure and leaks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If service provider directly accesses user data for model training, then model accuracy is improved, but user privacy security deteriorates
Solution Approach 1:
The system segments the data processing workflow into two distinct parts: data remains encrypted during transmission and storage, and decryption occurs only within the isolated trusted execution environment. This segmentation allows the service provider to access decrypted data for training while preventing unauthorized access to the actual user data, thus resolving the contradiction between model accuracy and privacy protection.
Solution Approach 2:
The trusted execution environment acts as an intermediary between the encrypted user data and the model training process. It receives encrypted data, decrypts it within its secure boundaries, performs training operations, and then discards the decrypted data. This intermediary mechanism enables accurate model training while preventing direct access to user privacy information.
2Object-affected harmful factors
If service provider uses encrypted data for model training, then user privacy security is improved, but model accuracy deteriorates
Solution Approach 1:
The system performs preliminary decryption within the trusted execution environment before model training begins. This preliminary action converts encrypted data into usable format for training operations, ensuring that the model receives high-quality input data while the decryption process itself is protected by the secure environment boundaries.
Solution Approach 2:
The trusted execution environment serves as an intermediary that transforms encrypted data into training-ready format without exposing the actual data to external systems. It enables the model training process to access decrypted data temporarily while maintaining privacy protection through its isolated architecture.
3Measurement precision
If comprehensive user data is collected for training, then model accuracy is improved, but data security risks increase
Solution Approach 1:
The system segments data handling into encrypted transmission/storage phase and decrypted processing phase within trusted environment. This allows comprehensive user data to be collected for improved model accuracy while maintaining security through the segmentation of sensitive operations into the protected trusted execution environment.
Solution Approach 2:
The trusted execution environment creates an inert or isolated atmosphere for data processing. Within this environment, comprehensive user data can be safely decrypted and processed without external access risks. The environment's isolation properties ensure that even if comprehensive data is present, security risks are minimized due to the controlled, secure context.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Encrypted data sent by user equipment is received by a service device. The encrypted data is decrypted in the trusted zone to obtain user data. A model is trained by using the user data in the trusted zone.