Trusted Execution Environment Model Training via Shared Memory
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current model training methods in AI applications face efficiency issues due to large training data volumes requiring significant trusted memory space and complex data transmission processes, which hinder the activation and operation efficiency of Trusted Execution Environments (TEEs).
Innovation Solution
A model training method employing a master-workers architecture, where training data is divided into sub-data pieces, encrypted, and stored in a shared memory, allowing each node to decrypt and train the model collaboratively, thereby reducing data transmission overhead and improving efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If large training data is processed in a single trusted execution environment, then data privacy and security are maintained, but the activation and operation efficiency of the trusted execution environment deteriorates due to significant memory space requirements
Solution Approach 1:
The patent divides the large training data into multiple smaller sub-training data pieces and distributes them across multiple worker nodes in addition to the master node. Each node processes only its assigned subset, significantly reducing the memory footprint and processing load on each individual trusted execution environment while maintaining the overall security framework through distributed processing.
2Reliability
If training data is transmitted to trusted execution environment for processing, then data security is ensured, but the data transmission process becomes complex and overhead increases
Solution Approach 1:
The host process performs data encryption before storing the training data in shared memory. The training nodes then directly decrypt the data from shared memory without complex transmission protocols. This preliminary encryption and direct access approach simplifies the overall transmission process while maintaining data security throughout the workflow.
3Reliability
If encrypted training data is copied to trusted execution environment, then data privacy is protected, but copying overhead and ecall overhead increase processing time
Solution Approach 1:
The patent introduces shared memory as an intermediary between the host process and the training nodes. The host process stores encrypted data in shared memory, and training nodes directly access and decrypt from this shared memory space. This eliminates the need for complex data copying and multiple ecall transitions, significantly reducing processing time while maintaining encryption-based privacy protection.
Data Source
AI summary
The present disclosure provides a model training method; and the model training method includes: by a host process, acquiring training data, and dividing the training data for a training node cluster employing a master-workers architecture to obtain multiple pieces of sub-training data; wherein the training node cluster comprises a master node and multiple worker nodes; wherein the host process runs in a non-trusted execution environment, and the training node cluster runs in a trusted execution environment; by the host process, encrypting each piece of sub-training data, and storing the encrypted sub-training data in a shared memory of the host process; and controlling the master node and each of the worker nodes to acquire corresponding encrypted sub-training data from the shared memory in accordance with corresponding data storage addresses, respectively, and train a preset model by using corresponding respective decrypted sub-training data, respectively, to obtain a trained model.