Biometric Set Relationship Scoring for Presentation Attack Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing presentation attack detection (PAD) systems are ineffective in identifying high-quality fake biometrics, biometrics from multiple individuals, or identical biometrics, leading to erroneous classifications.
Innovation Solution
A method that analyzes a set of N biometrics to determine genetic relationships between them, using feature extraction and comparison functions to generate a relationship score, enabling detection of presentation attacks, including high-quality fakes, multiple individuals, or identical biometrics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If individual classification of each single biometric is used, then the system is simple to implement, but it fails to detect high-quality fake biometrics, biometrics from multiple individuals, or identical biometrics
Solution Approach 1:
The patent merges individual biometric classifications with relationship-based classification. Instead of classifying each biometric independently, the system combines them into a set and evaluates relationships between biometrics using comparison functions, thereby detecting fake biometrics, multi-individual submissions, and identical biometrics while maintaining implementation feasibility
Solution Approach 2:
The patent introduces relationship scores and comparison functions as intermediaries between individual biometric classifications and final authentication decisions. These intermediaries analyze relationships between biometrics in a set, enabling detection of sophisticated attacks without requiring complete redesign of the authentication system
2Reliability
If relationship-based classification of biometric sets is used, then the detection accuracy improves, but the system complexity increases
Solution Approach 1:
The patent segments the classification process into distinct stages: individual biometric classification, relationship score calculation, and final set classification. This segmentation allows each component to be optimized independently and facilitates integration with existing authentication systems, reducing overall system complexity
Solution Approach 2:
The patent applies partial relationship analysis by focusing on specific comparison functions and relationship scores most relevant to detecting common attack types. Rather than analyzing all possible relationships between all biometrics, the system selectively evaluates key relationships, reducing computational complexity while maintaining detection effectiveness
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention provides a presentation attack detection method comprising: - obtaining (300) a set of N biometrics, N being an integer equal to or greater than 2; - for each biometric of the set, extracting (301) at least one feature of the biometric; - determining (302) at least one relationship value for the set of N biometrics by applying a comparison function to the extracted features; - determining (303) a relationship score based on the at least one relationship value, the relationship score being representative of a genetical relationship between the biometrics of the set; - determining (304) a first classification of the set of N biometrics based on the obtained relationship score, the first classification indicating that the set of N biometrics is genuine or that the set of N biometrics is a presentation attack.