Face Recognition Authentication With Offline Edge Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing facial recognition systems face challenges with interoperability between edge devices, require network connectivity for authentication, and compromise user privacy due to server-dependent operations.
Innovation Solution
A face recognition system that generates consistent facial encodings on both server and edge devices, allowing offline authentication and reducing network dependency, while using computer vision and neural networks for accurate user identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If facial recognition encoding is generated on server during authentication time, then authentication can be performed centrally, but network connectivity is required and authentication becomes time-consuming
Solution Approach 1:
The system generates facial encoding on the edge device during registration time rather than during authentication time. This preliminary action allows the encoding to be stored locally on the edge device, enabling fast authentication without network dependency. The encoding is generated in advance and stored in a database on the edge device, eliminating the need for real-time server communication during authentication.
2Loss of time
If facial recognition system works offline without network connectivity, then authentication speed improves, but interoperability between different platforms becomes difficult
Solution Approach 1:
The system uses a universal facial encoding generation mechanism that works consistently across different edge devices and platforms. The same encoding algorithm and database structure are deployed on all edge devices, enabling them to互认 each other's user encodings. This universality allows offline authentication to work across multiple platforms while maintaining interoperability.
3Measurement precision
If DNN models are trained using multi-class classifier with large-scale training data, then face recognition performance improves, but computational power requirements increase and user privacy is compromised
Solution Approach 1:
The system extracts and stores only the facial encoding (a condensed representation) of users on the edge device rather than storing raw training data or complex DNN models. This extraction approach allows the edge device to perform fast authentication using simple encoding comparison without requiring the computational power needed for DNN training. The complex training is performed centrally during registration, and only the simplified encoding is stored locally for fast authentication.
4Extent of automation
If server-dependent authentication is used, then centralized control is achieved, but user privacy is violated due to data exposure to service providers
Solution Approach 1:
The edge device performs authentication autonomously using locally stored facial encodings without requiring server involvement. Each edge device maintains its own database of user encodings and can independently verify user identity. This self-service approach eliminates the need for users to trust centralized servers with their biometric data, thereby protecting user privacy while maintaining automated authentication control at the edge device level.
Data Source
AI summary
The present disclosure relates to a method and a system for authenticating a user via face recognition. The disclosure encompasses: capturing at least one image(s) of the user; analyzing the at least one captured image(s) of the user to detect a face corresponding to the pre-defined parameters being conformed; cropping out the detected face from the at least one captured image to form a cropped face image; extracting one or more features from the cropped face image; comparing the extracted one or more features with a set of data associated with a plurality of users, and authenticating the user based on a positive matching of the extracted one or more features of the cropped face image with the set of data corresponding to the user.