Face Authentication via TEE-REE Segmentation

The patent divides the face recognition process into two segments: security-critical operations (biometric sample acquisition, verification) are executed in the Trusted Execution Environment, while computationally intensive operations (classification using neural networks) are executed in the Rich Execution Environment. This segmentation allows the system to leverage the security of TEE while utilizing the processing power of REE to avoid unacceptable recognition times.

The patent introduces an intermediary mechanism where biometric samples are transmitted from TEE to REE for classification, and intermediary outputs are transmitted back from REE to TEE for verification. This intermediary approach allows the system to benefit from both environments: the security of TEE and the processing power of REE, resolving the contradiction between security and speed.

The patent segments the computationally intensive 3D face recognition tasks (neural network classification) from security-critical tasks, executing the former in REE and the latter in TEE. This allows 3D face recognition to be implemented with high reliability while managing device complexity by distributing computational loads appropriately across different execution environments.

The patent enables the mobile device to utilize multiple processing resources (CPU cores, GPU, neural network modules) in the REE for 3D face recognition while maintaining security through TEE verification. This multi-functional approach allows the system to handle complex 3D recognition tasks without requiring all high-performance components to be exclusively available to TEE.