FAR and FRR in the context of TalkingID

Informative > Useful Clarifications

What is TalkingID doing in comparison to other biometric methods?

Introduction

Biometric authentication methods are becoming increasingly popular and are mainly used when targeted personal recognition is required. A password is right or wrong. Biometric parameters, on the other hand, are read and interpreted. One of the most important quality characteristics of biometric authentication is therefore the false acceptance rate (FAR) or false rejection rate (FRR). Please see description below.

Biometric Method	FAR in %	FRR in %
Fingerprint	0,001 … 2	0,1 … 5
Face Recognition	0,5 … 2	0,01 … 3
Hand Geometry	1 … 5	0,01 … 5
Iris Recognition	0,0001 … 1	0,001 … 2
Voice Verification TalkingID	0,0001 … 1	0,001 … 2

The False Acceptance Rate (FAR) indicates the probability of a person being falsely accepted. The False Rejection Rate (FRR) is the counterpart and indicates the probability with which a person is falsely not accepted.

FAR and FRR are thus important parameters for describing the effectiveness of a biometric recognition system.

For example:

Iris recognition has a FAR value of 0.0001%, which means that under 1 million fraudulent access attempts one access is allowed.

Fingerprint has a FRR value of 0,1%, which means that under 1 million access attemps 1000 attemps are rejected.

But how good are current systems really?

SmartBiometrics provides a qualitative assessment of various methods in the figure and classifies the accuracy (FAR and FRR) of iris scans and voice verification as very high. The older the available statistics are, the less accurate and more falsified the respective classification is. However, considering the current hardware technology and Deep Learning Artificial Intelligence in the application, audiovisual methods are the leading systems in recognition.

It also shows that quality and costs do not have to be on the same level.

Biometric Method	Fraud	Security	Accuracy	Usability	Costs
Fingerprint	high	low	medium	high	medium
Face Recognition	low	medium	high	high	low
Hand Geometry	very low	high	high	medium	high
Iris Recognition	low	high	high	low	high
Voice Verification TalkingID	very low	high	high	high	very low

Note: FAR and FRR were adopted following (DANG 2020) and Conti, Militello and Vitabile (2017). SmartBiometrics has taken the liberty to adapt the values regarding safety and costs to the current state of the art in hardware and software.