Private Authentication with Helper Networks for Spoofing Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional biometric authentication systems using procedural programming fail to achieve the required level of data filtering and accuracy, leading to compromised training and prediction accuracy due to the inclusion of bad data, and are unable to effectively handle presentation attacks or spoofed information.
Innovation Solution
Implementing helper networks as pre-processing neural networks to filter out bad data and validate identification information before training, using machine learning models to enhance accuracy and prevent spoofing, while operating in an encrypted space to maintain privacy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If procedural programming is used to filter biometric data, then the system is simple to implement, but the filtering accuracy and ability to detect spoofing is insufficient
Solution Approach 1:
The patent applies preliminary action by implementing helper networks that pre-process and filter biometric data before it reaches the main authentication model. These helper networks perform initial validation, spoofing detection, and data quality assessment, removing bad data early in the pipeline. This preliminary filtering improves the accuracy of subsequent authentication operations while keeping the main model relatively simple.
Solution Approach 2:
The patent introduces helper networks as intermediary components between raw biometric data and the main authentication model. These intermediary networks specialize in specific tasks such as detecting presentation attacks, validating data quality, and filtering out corrupted samples. By delegating specialized filtering functions to these intermediary helper networks, the system achieves high filtering accuracy without requiring the main authentication model to be overly complex.
2Measurement precision
If more data filtering is applied to improve accuracy, then authentication precision improves, but processing time and computational resources increase
Solution Approach 1:
The patent segments the data filtering process into multiple specialized helper networks, each responsible for specific aspects such as spoofing detection, quality validation, and corruption detection. This segmentation allows parallel processing of different filtering tasks, improving authentication accuracy through comprehensive filtering while managing processing time by distributing computational load across multiple specialized components rather than using a single monolithic filter.
Solution Approach 2:
The patent implements partial filtering through helper networks that selectively process only the aspects of data necessary for authentication. Rather than applying exhaustive filtering to all possible data aspects, the helper networks focus on critical filtering tasks such as detecting presentation attacks and removing obviously corrupted samples. This partial action approach achieves sufficient filtering accuracy for authentication purposes while avoiding the computational overhead of complete exhaustive filtering.
3Reliability
If helper networks are added to filter data, then spoofing detection capability improves, but device complexity increases
Solution Approach 1:
The patent designs helper networks with multi-functionality, where each helper network can perform multiple related tasks. For example, helper networks are configured to simultaneously detect various types of presentation attacks (masks, deepfakes, photos), validate data quality metrics, and identify corrupted samples. This universality allows the system to achieve comprehensive spoofing detection capability while using a relatively compact set of helper networks rather than requiring separate specialized components for each function.
Solution Approach 2:
The patent implements a nested architecture where helper networks are integrated within the broader authentication system framework. The helper networks are embedded as sub-components that process data flows within the main authentication pipeline. This nesting allows the spoofing detection functionality to be incorporated into the existing system architecture without requiring completely separate external systems, thereby improving reliability while managing complexity through integrated design.
Data Source
AI summary
A set of 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 other embodiment, helper networks can be used to filter identification inputs to improve the accuracy of the models that use encrypted inputs for classification.