Trusted Execution Environment Face Authentication via Rich Environment Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Executing face recognition software in the Trusted Execution Environment of mobile devices is inefficient due to high computing and memory requirements, leading to unacceptable recognition times, especially for 3D face recognition which demands powerful resources like additional CPU cores and a graphic processor not available in the TEE.
Innovation Solution
A method that acquires biometric face samples in the Trusted Execution Environment, classifies them in the Rich Execution Environment, and verifies the results in the TEE, utilizing the Rich Execution Environment's computing power while ensuring security by detecting hacks through test biometric samples and intermediary outputs verification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If face recognition software is executed in the Trusted Execution Environment, then security is improved, but recognition time becomes unacceptable due to limited computing power
Solution Approach 1:
The patent divides the face recognition system into two segments: the TEE handles security-critical functions (biometric sample acquisition, test sample insertion, result verification) while the REE handles computationally intensive classification. This segmentation allows each environment to operate in its optimal performance zone, resolving the contradiction between security and speed.
Solution Approach 2:
The patent introduces test biometric face samples as an intermediary mechanism to bridge the TEE and REE. These test samples are inserted by the TEE into the biometric sequence, processed by the REE classifier, and used to verify the integrity of the classification process, enabling secure outsourced computation.
2Measurement precision
If 3D face recognition is implemented with neural networks and additional sensors, then recognition accuracy is improved, but computing and memory requirements increase beyond TEE capabilities
Solution Approach 1:
The patent extracts the computationally intensive neural network classification function from the TEE and places it in the REE. The TEE retains only the essential security functions, while the complex 3D face recognition processing with multiple sensors and neural networks is outsourced to the more powerful REE environment.
3Productivity
If the Rich Execution Environment is used for face classification, then processing speed is improved, but security risks increase due to potential hacks
Solution Approach 1:
The patent applies preliminary action by inserting test biometric face samples into the biometric sequence before classification. These test samples serve as预先 prepared security checks that will reveal any hacking attempts during the classification process, allowing the TEE to verify REE integrity beforehand.
Solution Approach 2:
The patent implements a feedback mechanism where the TEE verifies the classification results by processing test samples and comparing outcomes. This feedback loop provides continuous security monitoring of the REE classification process, detecting and preventing potential hacks while maintaining high processing speed.
Data Source
AI summary
A method used in a mobile equipment for authenticating or identifying a user based on his face includes acquiring a sequence of biometric face sample corresponding to successive image frames of the user in a Trusted Execution Environment of the mobile equipment, and, in the Trusted Execution Environment, modifying or replacing some of the biometric face samples in the sequence, so as to generate a modified sequence with test biometric face samples. The method includes transmitting the modified sequence to a Rich Execution Environment of the mobile equipment, and classifying at least one portion of the biometric face samples using a classifier in the Rich Execution Environment, so as to generate for each classified biometric face sample a set of intermediary outputs and a classification result. The method also includes, for each classified biometric face sample, sending at least some of the intermediary outputs or classification result, and/or a digest of the intermediary outputs, to the Trusted Execution Environment. The method further includes, for at least one test biometric face sample, verifying the at least some of the intermediary outputs and/or result and/or digest in the Trusted Execution Environment.