TEE Key Management for Secure Model and Data Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Large language models face risks of data and model leakage during fine-tuning and deployment due to untrustworthy cloud or cluster environments, and existing security measures like federated learning and secure computing are inefficient or impractical for large-scale scenarios.
Innovation Solution
Implementing data and model encryption within a trusted execution environment (TEE) using session keys and key management services to ensure secure data processing, including model fine-tuning and inference tasks, with remote proof mechanisms to verify secure hardware environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If data and models are processed in untrustworthy cloud or cluster environments, then processing efficiency and resource utilization are improved, but data and model security deteriorate due to potential leakage
Solution Approach 1:
The system segments the trust relationship by introducing a TEE-based key management service that separates key storage from data processing. The TEE environment creates an isolated secure segment that protects cryptographic keys while allowing untrusted cloud infrastructure to handle data processing, thus resolving the contradiction between utilizing cloud resources and maintaining security.
Solution Approach 2:
The system creates a composite security architecture combining TEE hardware security with software-based encryption protocols. This composite approach integrates the security properties of trusted execution environments with the flexibility of cryptographic key management, enabling secure data processing in untrustworthy cloud environments without sacrificing processing efficiency.
2Reliability
If traditional encryption methods are used for data protection, then data security is improved, but processing efficiency deteriorates due to decryption overhead
Solution Approach 1:
The system performs preliminary encryption of data before it enters the processing pipeline, and maintains encrypted state throughout processing where possible. The TEE-based key management service pre-establishes secure key distribution channels and manages cryptographic operations in advance, reducing the need for frequent decryption operations during processing and thus minimizing efficiency overhead.
3Reliability
If federated learning or secure computing methods are applied, then data security is improved, but system complexity and implementation difficulty worsen for large-scale scenarios
Solution Approach 1:
The TEE-based key management service acts as an intermediary between the untrusted cloud infrastructure and the data processing applications. It abstracts away the complex security protocols and cryptographic operations, providing a simplified interface that maintains security without requiring applications to implement complex federated learning or secure computing mechanisms.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
The present disclosure relates to a data processing method and apparatus, and a readable medium and an electronic device, so as to ensure the security of data and a model in a data processing process. The data processing method comprises: receiving a data processing task; determining target data that corresponds to the data processing task and a target first key that corresponds to the target data, wherein the target data is obtained by using a first key to encrypt data to be processed, and the target first key is obtained by using a first session key to encrypt the first key; by means of a target computing node, decrypting the target first key on the basis of the first session key, so as to obtain the first key, and decrypting the target data on the basis of the first key, so as to obtain the data to be processed, wherein the target computing node is run in a trusted execution environment; and determining a data processing result on the basis of a target model and the data to be processed, wherein the target model is obtained by means of decrypting an encrypted target model on the basis of a model key, the model key is stored in a key management service, and the key management service is run in the trusted execution environment.