Biometric Recognition Using Non-Reversible Feature-Based Keys
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current biometric verification systems face challenges in ensuring the non-reversibility, diversity, renewability, and revocability of biometric templates, leading to security risks and management difficulties in large-scale implementations, particularly due to the storage and transmission of unprotected templates.
Innovation Solution
A biometric recognition system that generates a unique and entropic representation of a biometric feature vector using autoencoders and convolutional neural networks, which is compressed, hashed, and used for cryptographic key material without storing the original biometric data, ensuring non-reversibility, diversity, and revocability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If biometric templates are stored for verification, then authentication capability is improved, but security risk increases due to reversibility of template encoding
Solution Approach 1:
The patent extracts only the essential biometric features (minutiae, ridge count, pattern type) from the complete fingerprint image and stores only these extracted features as templates. This extraction approach maintains authentication capability while reducing the amount of stored data and eliminating the reversibility risk associated with storing complete biometric images or their direct encodings.
Solution Approach 2:
The patent creates a simplified representation or copy of the biometric data (template) that contains only the necessary identifying features rather than the original biometric image. This template copy can be stored and compared without exposing the complete biometric, thus maintaining security while enabling verification.
2Measurement precision
If complete biometric data is stored for identification, then recognition accuracy is improved, but storage security and data management complexity increase
Solution Approach 1:
The patent segments the biometric data into distinct feature components (minutiae points, ridge count measurements, pattern type classifications) and stores only these segmented features rather than the complete biometric image. This segmentation maintains recognition accuracy through feature-level detail while significantly reducing storage requirements and management complexity.
Solution Approach 2:
The patent extracts and stores only the essential identifying features from the complete biometric data. By taking out only the necessary features (minutiae, ridge count, pattern type) and discarding redundant information, the system achieves efficient storage and simplified data management while preserving recognition accuracy.
3Object-affected harmful factors
If biometric templates are made non-reversible for security, then template protection is improved, but ability to verify and match identities may be compromised
Solution Approach 1:
The patent creates a simplified template copy that contains only the essential biometric features in a non-reversible encoded form. This template copy can be securely stored and compared for verification without risking exposure of the complete biometric, thus achieving both protection and verification capability.
Solution Approach 2:
The patent transforms the biometric data into a different parameter representation (encoded template format) that is non-reversible. By changing the parameter form from raw biometric image to encoded feature template, the system achieves security while maintaining verification capability through the encoded representation.
Data Source
AI summary
Biometric recognition systems and methods are described, including a biometric recognition system including a user system with a biometric able to scan a finger and provide an image. From the image, a plurality of features may be extracted to create at least one fingerprint feature vector using a convolutional neural network.