Biometric Self-Key Verification With De-Identified Feature Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional facial identification systems face challenges in balancing privacy and security, with outsourcing solutions violating privacy regulations and local solutions being limited in scalability, flexibility, and power consumption, while also risking data leakage and device theft.
Innovation Solution
An identity verification system that performs secure identity verification by capturing biometric features locally, de-identifying the data using deep learning models, and generating self-keys dynamically, which are used for verification without storing sensitive data on third-party systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If facial identification data is outsourced to a central server, then verification efficiency is improved, but user privacy is compromised and security risks increase
Solution Approach 1:
The system divides the verification process into two independent parts: (1) local extraction of biometric features and generation of verification data on the user device, and (2) verification of this data on the server without accessing original biometric information. This segmentation allows efficient server-side verification while protecting user privacy by never transmitting or storing sensitive biometric data on external servers.
Solution Approach 2:
The patent introduces an intermediary verification data that acts as a mediator between the user's biometric information and the server's verification process. This verification data contains sufficient information for accurate verification while being inherently anonymous and non-reversible, thus serving as a privacy-preserving intermediary that enables efficient verification without compromising security.
2Reliability
If biometric data is stored locally on user devices, then privacy protection is improved, but scalability and flexibility are limited
Solution Approach 1:
The system extracts only the essential verification characteristics from biometric data while removing all personally identifiable information. This extraction process creates compact verification data that can be stored locally for privacy protection yet transmitted efficiently to servers for scalable verification, thus resolving the contradiction between local storage benefits and system scalability requirements.
3Measurement precision
If deep learning models are used for de-identification, then processing accuracy is improved, but computational power consumption increases
Solution Approach 1:
The system performs preliminary extraction of biometric features and generation of verification data locally on the user device before transmission to the server. This preliminary action reduces the computational burden on mobile devices during actual verification, as the heavy deep learning processing is completed in advance when the device has adequate power resources, thus balancing accuracy requirements with power consumption constraints.
Data Source
AI summary
Disclosed are an identity verification system and method. In the method, a user device captures a first biometric feature of a user by using a first biometric feature capturing device, performs de-identification processing on the first biometric feature to obtain first de-identified data, transforms the first de-identified data into a feature vector including de-identified features and stores the same in a storage device, and dynamically generates a self-key including the feature vector in response to an activation operation. A verification device captures the self-key by using a data capturing device, captures a second biometric feature by using a second biometric feature capturing device, performs the de-identification processing on the second biometric feature, transforms the result into a feature vector including de-identified features to compare with the feature vector in the self-key, and verifies the second biometric feature according to a comparison result.