Multi-Distant Weighted Scoring for Biometric Authentication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing biometric authentication methods face challenges in accurately distinguishing authorized users from impostors due to high false acceptance and rejection rates, particularly when relying on behavioral data that is not repeatable and varies significantly across different interactions.
Innovation Solution
A method involving pairwise comparisons of biometric samples against templates, using allowable variability ranges, nonlinear transformations, and weighted scoring to determine whether a sample corresponds to the same user as the template, by calculating proportions of successful matches across multiple scaling factors and applying these through specific mathematical equations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If simple statistical methods or neural networks are used to compare biometric samples, then the system is easier to implement without special hardware, but the false acceptance and rejection rates increase
Solution Approach 1:
The comparison process is divided into multiple independent stages: initial distance calculation, binary comparison at multiple scaling factors, proportion computation, nonlinear transformation, and weighted scoring. Each stage processes specific aspects of the biometric data independently, allowing complex authentication to be broken down into manageable steps that maintain high accuracy without requiring specialized hardware.
Solution Approach 2:
The patent introduces multiple scaling factors (different dimensional perspectives) to compare the same biometric sample against the template. By evaluating distances at various scaling dimensions and combining results through weighted scoring, the system captures behavioral patterns from multiple angles, significantly improving authentication reliability while remaining implementable on standard computers.
2Adaptability or versatility
If behavioral biometric data is used for authentication, then no special hardware is required, but the data is not repeatable and varies significantly across interactions
Solution Approach 1:
The system dynamically adjusts the allowable variability ranges for each biometric element based on the specific sample being analyzed. Rather than using fixed thresholds, the method computes scaling factors and weighted scores that adapt to the natural variation in behavioral data, allowing the authentication system to accommodate legitimate variations in user behavior while maintaining security.
Solution Approach 2:
The patent transforms the raw biometric data through multiple parameter changes: calculating distances at different scaling factors, applying nonlinear transformations to proportion values, and computing weighted sums. These parameter transformations convert unstable, variable behavioral data into a stable scoring system that reliably distinguishes authorized users from impostors despite natural variations in typing patterns.
Data Source
AI summary
Methods of comparing a plurality of measurements to a template are described. Measurements are compared piecewise (element-by-element) and a proportion of successful comparisons at each of a plurality of distance scaling factors is calculated. The proportions are subjected to a nonlinear transformation, then normalized and combined into a weighted sum. The weighted sum is compared with a threshold value to establish the result of the comparison. Software and systems to implement embodiments of the invention are also described and claimed.