18 Jul 2024

Buyers Guide

Complete playbook to understand liveness detection industry

Learn More

Knowing the Equal Error Rate (EER) in Biometrics

Author: admin | 18 Jul 2024

So far we have explained the error rates that have trade-offs between them as attempting to reduce one increases the opposite one. For example, if we discuss the reduction of FMR (False Match Rate) and FNMR (False Non-Match Rate) together, it is impossible to reduce both at the same time due to trade-offs. But reaching the lowest intersection point is considered ideal which is also called the Equal Error Rate (EER).

EER In Biometrics

Equal Error Rate is the convergence point of both the False Acceptance Rate (FAR) and False Rejection Rate (FRR). In other words, it is the threshold of a biometric identity verification system to incorrectly accept an imposter as a legitimate user at an equal rate of incorrectly rejecting a legitimate user identity flagging it as an imposter.

The Importance of Equal Error Rate (EER) in Biometrics

As stated earlier the trade-off between FAR and FRR causes both a non-zero intersection between them. Hence, the Equal Error Rate (EER) is important to calculate as it is the ideal state of performance of a biometric identity verification system. EER determination is important for two major goals:

  • Maintaining a balance between security and user convenience as lower EER rates indicate better user convenience as well as enhanced security through biometrics.
  • Comparison of different biometric identity verification technologies and scaling their performances to make informed decisions about implementing the best-suited biometric system.

Equal Error Rate Formula and Calculation

1. Data Collection

Large datasets are gathered having samples of the biometric identity verification system under study, let’s say a face recognition system has a database of facial images, selfies, and live videos of faces. This database comes from the relevant population with high diversity to increase the system’s capability and train its models for higher efficiency rates.

2. Feature Extraction 

Biometric facial data is extracted focusing on features such as nose, eyes, lips, etc. This data is further used for face matching and comparison in 1:1 and 1:N matching domains by implementing a matching algorithm that can generate intelligent similarity scores.

3. FAR & FRR Calculation 

EER is determined after False Acceptance Rates and False Rejection Rates are calculated being dependent on the two.

The graph illustrates False Match Rate (FMR) and False Non-Match Rate (FNMR) with an intersection between False Acceptance Rate (FAR) and False Rejection Rate (FRR) as Equal Error Rate (EER). Both curves show the Imposter attempts and Genuine user attempts accordingly while the decision threshold is where a biometric solution finds its current performance standing.

Use-cases of EER

In a biometric security system such as a facial recognition tool (FRT), the Equal Error Rate (EER) has the following use cases:

1.  Fine-tuning Biometric System’s Performance 

It helps in improving the system’s accuracy and efficiency. If EER levels are high it indicates that the system needs improvement in some areas and as the EER levels are lowered the system’s credibility and value increase automatically.

2. Analyzing Digital Security through Biometrics 

EER is the quantitative measure determining and pointing towards the biometric security level offered by a tested system. Lower EER means that the biometric screening system is highly secure and can be used in sensitive areas.

3. User Journey

User acceptance increases if EER rates are lower. It is because legitimate users can easily and swiftly get verified as the system’s error rates are low. This means that biometric technology is convenient and offers users more value and frictionless experience.

How to Lower EER Rates?

In biometric facial recognition technology which is the only unconstrained biometric trait and is considered to be the only fully accepted biometric technology in the future for identification, Lowering EER is highly important. 

Here’s a brief overview of how EER can be lowered for improved accuracy and speed in an FRT.

  • Improving image quality by using high-definition cameras and quality equipment.
  • Enhancing lighting conditions and angles in which facial image is recorded.
  • Trying to improve FAR and FRR rates and attempting to reach the lowest intersection point by improving the overall system.
  • Improving the AI and Machine Learning algorithms used in a facial recognition tool.
  • Enhancing user experience through runtime instructions and making the UI more user-friendly.