Biometric Analysis

Daon is independent of all biometric technology and vendors and our only priority is our customer's particular needs. We continuously analyze the performance of different biometrics so that we can best advise our customers on the current state of biometric technology.

We constantly evaluate, select, configure and combine the most suitable algorithms and solutions for different use cases. In addition to integrating cutting-edge third-party algorithms we also develop our own algorithms which can be easily tailored to fit specific scenarios.

Common biometric authentication errors

No biometric authentication system is perfect. The most common errors are:

  • failure to register a person's biometrics in the system (Failure to Enroll)
  • failure to correctly verify a genuine user attempt (False Non-Match)
  • incorrectly accepting an impostor (False Match)

In a presentation attack detection subsystem (anti-spoofing) the corresponding error rates are: mistaking a genuine user as an attacker (Bona Fide Presentation Classification Error Rate) and failing to detect a real attack (Attack Presentation Classification Error).

Understanding the source of these errors and taking the appropriate steps to reduce them to acceptable levels is a key element in the successful deployment of a biometric authentication system. Factors that affect biometric performance include the capture device, capture application, population, environment, transaction protocol and the algorithm settings.

How does Daon approach biometric analysis?

The basis of any biometric analysis work is to measure these errors for all possible configurations of an algorithm for a specific scenario. A trade-off can then be made between the different error rates in order to derive the optimal configuration. For example, lowering the False Match Rate (FMR) to increase security will usually result in an increase in the False Non-Match Rate (FNMR). Daon's analysis produces accuracy statistics which are graphed to visualize these configuration trade-offs. The graphs show how algorithm accuracy changes across all the different decision score thresholds.

Analytics chartIn evaluation performance, Daon follows best practices for biometric testing and reporting as defined in the standards ISO/IEC 19795 and ISO/IEC 30107. We developed our DaonAnalytics software to automate much of this testing and analysis, taking large biometric datasets as input and outputting accuracy statistics, graphs, and performance metrics.

Ongoing testing allows us to identify, incorporate and improve on the latest biometric innovations to offer state-of-the-art biometric solutions for any given scenario. Find out more about the R&D work of the Daon Biometric Research Lab.