TEE-GPU Confidential Computing With Ciphertext Scrambling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in securely combining a Trusted Execution Environment (TEE) with a Graphics Processing Unit (GPU) for confidential computing, as there are risks of data security breaches during data transmission and processing, which compromises the security boundary of the TEE.
Innovation Solution
The method involves scrambling encrypted data using a ciphertext scrambling function in the TEE and processing it with a target white-box decryption combined function on the GPU, ensuring that only processed data are obtained, and returning the result to the TEE for final plaintext computation, thereby enhancing data protection through multiple layers of encryption and obfuscation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If encrypted data are directly sent to GPU for processing, then computing efficiency is improved, but data security is compromised due to exposure during transmission and processing
Solution Approach 1:
The patent segments the data processing into distinct phases: encryption/scrambling in the TEE, processing in the GPU, and decryption/scrambling reversal in the TEE. The data remain encrypted or scrambled throughout transmission and processing, with only the necessary processing operations performed on the protected data, never exposing plaintext to external systems.
Solution Approach 2:
The patent introduces an intermediary mechanism where the TEE acts as a secure intermediary between data owners and the GPU. The TEE performs encryption and scrambling before data leaves its control, and only provides decrypted data to the GPU after verification, ensuring that the GPU processes only encrypted or scrambled data without access to plaintext.
2Reliability
If the security boundary of TEE is expanded to GPU, then data protection coverage is improved, but system complexity increases due to additional encryption and obfuscation layers
Solution Approach 1:
The patent merges multiple security functions into a unified approach: encryption, scrambling, and decryption operations are combined into a single integrated mechanism that operates consistently across all data flows between TEE and GPU. The white-box decryption combined function integrates multiple decryption and scrambling reversal operations into one unified function that the GPU can execute.
Solution Approach 2:
The patent changes the state parameters of the data through systematic transformation: data transition from plaintext to encrypted, then to scrambled form during transmission, and back to decrypted state for processing. The scrambling function applies parameter transformation to the encrypted data, changing its representation without revealing the underlying plaintext, thereby maintaining security while enabling processing.
Data Source
AI summary
This specification provides computer-implemented methods, systems, electronic devices, and storage media for implementing confidential computing. In an example method, encrypted data is scrambled based on a ciphertext scrambling function in a trusted execution environment, and sent to at least one processor, so that the at least one processor processes the scrambled encrypted data by using a target white-box decryption combined function, and obtains, through computing, a corresponding computing result by using obtained processed data. The target white-box decryption combined function is used to perform ciphertext descrambling processing, decryption processing, and first data protection processing on the scrambled encrypted data. The computing result or a derived result of the computing result is obtained. Processing is performed in the trusted execution environment to obtain a plaintext computing result.