Encrypted Biometric Embeddings for Searchable Privacy Authentication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional biometric systems face limitations in performing one-to-many searches on encrypted biometric data, are vulnerable to faked or replayed biometric signals, and struggle with large class problems in accuracy and scalability, often resorting to human intervention for low probability matches.
Innovation Solution
A neural network architecture that generates encrypted feature vectors to train a classification network for authentication, incorporating liveness validation and post-classification validation to enhance accuracy and security, using encrypted biometric and behavioral data for privacy-enabled authentication.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional biometric systems store and search encrypted biometric data, then security and privacy are improved, but the ability to perform one-to-many searches and matching is lost
Solution Approach 1:
The patent introduces homomorphic encryption as an intermediary that enables computations on encrypted data. The system uses encrypted feature vectors that can be mathematically operated upon without decryption, allowing one-to-many searches to be performed on encrypted biometric data while maintaining security. The homomorphic encryption scheme acts as a mediator between the security requirement (encrypted storage) and the search requirement (one-to-many matching).
Solution Approach 2:
The patent transforms the biometric data into a different parameter space using neural network feature extraction. Instead of working with raw biometric images or signals, the system converts them into feature vectors with specific mathematical properties that enable homomorphic operations. This parameter transformation allows encrypted searches while preserving the ability to perform one-to-many matching.
2Ease of operation
If conventional systems accept any biometric input for authentication, then ease of use is improved, but vulnerability to faked or replayed biometric signals increases
Solution Approach 1:
The patent implements liveness detection as a preliminary action before authentication. The system performs liveness verification by analyzing characteristics of the biometric input to determine if it originates from a living person rather than a spoof or replay attack. This preliminary check filters out fraudulent inputs before they reach the authentication stage, maintaining both convenience and security.
Solution Approach 2:
The system incorporates feedback mechanisms where the authentication process provides information back to the user about the validity of their input. Liveness detection feedback helps users understand whether their biometric input is being accepted or rejected, allowing them to correct improper attempts while maintaining security against spoofing.
3Measurement precision
If neural network training uses large datasets for improved accuracy, then authentication accuracy is improved, but training time and computational resources increase
Solution Approach 1:
The patent performs preliminary data preparation and feature extraction during the enrollment phase, transforming raw biometric data into feature vectors before storage. This preliminary action reduces the computational burden during authentication, as the system only needs to compare pre-processed feature vectors rather than processing raw images or signals in real-time.
Solution Approach 2:
The patent segments the authentication process into distinct phases: enrollment (feature extraction and storage) and authentication (feature comparison). By segmenting the workload, the system can perform computationally intensive operations during enrollment when time is not constrained, and perform only lightweight comparisons during authentication, improving overall efficiency.
Data Source
AI summary
A set of distance measurable encrypted feature vectors can be derived from any biometric data and/or physical or logical user behavioral data, and then using an associated deep neural network (“DNN”) on the output (i.e., biometric feature vector and/or behavioral feature vectors, etc.) an authentication system can determine matches or execute searches on encrypted data. Behavioral or biometric encrypted feature vectors can be stored and/or used in conjunction with respective classifications, or in subsequent comparisons without fear of compromising the original data. In various embodiments, the original behavioral and/or biometric data is discarded responsive to generating the encrypted vectors. In another embodiment, distance measurable or homomorphic encryption enables computations and comparisons on cypher-text without decryption of the encrypted feature vectors. Security of such privacy enabled embeddings can be increased by implementing an assurance factor (e.g., liveness) to establish a submitted credential has not been spoofed or faked.