28 Jun 2024

Buyers Guide

Complete playbook to understand liveness detection industry

Learn More

What is the False Acceptance Rate (FAR) in Facial Recognition?

Author: teresa_myers | 28 Jun 2024

Identity Verification solutions are critical to an organization’s regulatory compliance and act as a security measure against different threats. Technological advancements in the identification systems for people are the lifeline of digital security in an organizational setup. Normally, these solutions are termed ‘Identity vendors’. 

The prevalent technologies in identity proofing include:

  • Biometric identity verification
  • Fingerprint
  • Facial Recognition
  • Voice Recognition
  • IRIS, and Retina Scanning.
  • Multi-Factor Authentication (MFA)
  • Near Field Communication (NFC)
  • Digital Certificates 
  • Passwords/PINs
  • Security Questions

We have seen that password-based authentication, security questions, and MFA (without biometrics) are proven as ‘gappy’ as they can now be easily spoofed as compared to the Biometric Identity Verification on the other hand has a strict and personalized approach offering more control to the users over their digital identity and freeing them from worrying about identity fraud attempts. 

The accuracy and speed of a biometric recognition system especially the facial recognition system depend upon multiple factors in 2 phases:

  • The ‘Enrollment Phase’ is where biometric facial identities are stored in a system’s database for the first time through facial recognition. 
  • The ‘Identification Phase’ is after the enrollment is complete and users are verified against the identity database using their biometric facial identity.

False Acceptance Rate (FAR) & its Impact on Biometric Facial Recognition? 

False Acceptance Rate is a biometric identity verification measure that refers to how much an identity verification system falsely accepts an identity. It is abbreviated as FAR. 

Another definition of FAR is:

The percentage error of a biometric identity verification system in accepting wrong identities.

False Acceptance Rate affects a biometric system’s accuracy and can cause serious implications.

False Acceptance Rate (FAR) Calculation

FAR in biometric systems is generally calculated as below:

False Acceptance Rate (FAR) in the given formula is calculated as the number of false acceptances divided by the total number of imposter attempts multiplied by 100.

Given the above illustration, the false acceptance rate depends on the total number of false positives and the total number of fraudulent identification attempts. The term false positives is often used instead of false acceptance rate. Sometimes it is also referred to as False Match Rate. But there is a slight difference between them:

  • A false acceptance rate is the number of wrongly accepted identities by a biometric identity system as a genuine user.
  • False positives are the missed-out identities that were authenticated by the system where they should’ve been flagged as negative.
  • A false match rate (FMR) is when an unauthorized person is falsely recognized during the feature comparison stage.

FAR in Facial Recognition

False Acceptance Ratio in face recognition solutions is evident but the numbers are low. This is because of the unique identity parameters and highly efficient facial recognition technology. Since facial biometric identity solutions are designed to detect and prevent anomalies, the algorithms in the back end work to patronize deepfakes, mask attacks, and other spoofing attacks. But rapidly advancing AI and its availability to everyone is causing threat vectors like deepfakes to become highly realistic and difficult to detect. This causes the facial recognition systems to have an increased FAR too. 

Reasons for FAR in Facial Recognition

False Acceptance Rate can occur in any biometric system. Distinctively, it can occur in facial recognition systems due to the following reasons:

Limitations in the System
  • Outdated or weak Algorithms in an AI facial recognition tool can cause increased FAR.
  • Both the Feature Extraction Accuracy and Face Matching through an AI-driven tool can have increased FAR due to its limited technical capabilities.
Facial Image Quality 
  • Different aspects of a face image quality affect levels of FAR including
  • Lighting conditions, brightness, shadows, and exposure.
  • Resolution & clarity impacts the overall face-matching tool’s performance.
Database Issues
  • A database of facial identities can also cause a biometric false acceptance rate due to
  • Duplicate entries of facial identities in a single database.
  • Size & Quality of database storage in terms of speedy access and response while face matching in both 1:1 and 1:N face comparison. 
Similarity Threshold
  • The similarity index of face matching also plays a critical role in the occurrence of False Acceptance Rate in biometrics.
  • A lower similarity index causes increased FAR but a higher similarity index may also cause increased False Rejections.
  • Objects like glasses, hats, or masks hinder the detection of facial features and may also result in FAR.
Intentional Deception
  • The creation of fake identities using deepfakes, morphing, and mask attacks can also cause an increase False Accept Rate.

A Discussion on Benchmark for Reduced FAR in Biometrics

The National Institute of Standards and Technology (NIST) has a sophisticated and structured approach to facilitating achieving perfection in biometric identity verification. NIST has not set a benchmark standard as naturally, the ideal state of FAR in biometric systems would be a perfect zero. Instead, NIST provides guidelines, standards, and benchmarks for evaluating the performance of biometric systems, including FAR, through various publications and testing programs. The acceptable FAR can vary depending on the application and security requirements. 

For facial recognition identity solutions, NIST has divided its research and testing initially known as FRVT (Face Recognition Vendor Testing) into two categories:

  • FRTE (Face Recognition Technology Evaluation)
  • FATE (Face Analysis Technology Evaluation)

Here’s a summary of key points on NIST Standards and False Acceptance Rate (FAR) for Facial Recognition:

In this chart, NIST evaluates different biometric modalities under FRVT and FRTE where application-specific requirements are set as per environment quality from high to low including the Trade-off between FAR & FRR where Benchmark values of FAR are set for facial recognition and fingerprints.

What Businesses Employing Facial Recognition Should Do?

A biometric identity verification solution focused on facial recognition requires the perfect intersection between False Acceptance Rate and False Rejection Rate. It is because no identity solution has yet reached a perfect zero. After all, if a facial recognition system tries to minimize FAR, the FRR will increase. 

So, here are three practical considerations for Identity solution providers to comply with NIST standards and enhance their solutions in terms of reduced FAR and False Rejection Rate (FRR).

  • Refer to NIST Reports: Utilize NIST’s evaluation reports to select biometric systems that meet their specific FAR requirements.
  • Perform In-House Testing: Conduct internal testing to validate the system’s performance in the actual operational environment.
  • Balance Security and Usability: Adjust the FAR and FRR to achieve a balance that aligns with the application’s security needs and user experience.