Biometric Matching Using Encrypted Euclidean Feature Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems face challenges in searching encrypted biometric data without decryption, particularly in large databases, which is essential for protecting privacy and ensuring secure authentication.
Innovation Solution
The use of a neural network to translate biometric vectors into Euclidean measurable feature vectors, allowing for matching and classification without decryption, using techniques such as convolutional layers and Frobenius algorithms to perform searches in O(log(n)) time, ensuring secure and efficient identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If encrypted biometric data is stored in databases for secure authentication, then data privacy and security are improved, but the ability to search and identify users efficiently deteriorates
Solution Approach 1:
The patent introduces an intermediary representation layer between the encrypted biometric data and the search query. The system transforms biometric inputs into a standardized representation space where encrypted templates and live captures can be compared without decryption. This intermediary representation enables efficient similarity searching while maintaining encryption, resolving the contradiction between security and search efficiency.
Solution Approach 2:
The system changes the parameter space in which biometric data is stored and searched. Instead of storing raw biometric data or requiring decryption for search, the system transforms data into an encrypted representation that maintains mathematical properties suitable for similarity computation. This parameter transformation allows O(log n) search complexity while preserving encryption, simultaneously achieving security and efficiency.
2Reliability
If biometric data is encrypted to protect against unauthorized access, then security against data breaches is improved, but the speed of authentication processes deteriorates
Solution Approach 1:
The system performs preliminary transformation of biometric data into encrypted representations during the enrollment phase. By pre-computing and storing encrypted templates in an optimized structure, the actual authentication process avoids the need for full decryption operations, significantly reducing authentication time while maintaining security. The heavy computational work is done beforehand when data is less sensitive.
Solution Approach 2:
The patent replaces the traditional mechanical decryption-comparison-verification process with a direct comparison in encrypted representation space. Instead of decrypting data to compare, the system uses mathematical operations on encrypted representations that directly yield similarity metrics, eliminating the decryption step and accelerating authentication while preserving security.
3Reliability
If traditional search methods are used on encrypted biometric databases, then data protection is maintained, but search complexity and computational resources required increase significantly
Solution Approach 1:
The system segments the search process into two distinct phases: an indexing phase where encrypted biometric templates are pre-processed and organized into a search structure (such as a tree or hash index), and a query phase where new biometric inputs are transformed and compared against the indexed structure. This segmentation reduces search complexity from O(n) to O(log n) by dividing the database into manageable segments that can be efficiently traversed without decryption.
Solution Approach 2:
The patent introduces a new dimensional space for storing and searching biometric data - the encrypted representation space. By transforming biometric data into this additional dimension while preserving encryption, the system enables efficient similarity searching without compromising data protection. The transformed space maintains mathematical properties that allow for fast computation while keeping data encrypted.
Data Source
Figure 1
Figure 2A
Figure 2B~2C
AI summary
Computer implemented systems and methods for matching an encrypted biometric input record with at least one stored encrypted biometric record, and without data decryption of the input and the at least one stored record.