Why aren’t all iBeta evaluations equal?

When comparing liveness detection solutions, iBeta evaluations are a key factor in determining their robustness. However, unlike the popular belief, not all iBeta evaluations are equal, and it is essential to understand their nuances to make a wise choice.

This article explores the criteria to consider to ensure the reliability of liveness detection technologies and the pitfalls to avoid when comparing different market solutions. Understanding the subtleties of iBeta evaluations will allow you to choose the solution best suited to your needs.

What is an iBeta evaluation?

iBeta is an independent American laboratory specializing in the performance evaluation of biometric technologies, regardless of the method (facial, fingerprint, voice, etc.) and in compliance with international standards such as ISO or FIDO.

In the context of facial biometrics, iBeta is particularly known for testing "liveness detection" solutions to ensure that these technologies are robust against presentation attacks (PAD). You can learn more about presentation attacks in this article.

iBeta has two levels of evaluation:

• Level 1 iBeta evaluation subjects liveness detection solutions to relatively simple and standardized attacks (such as printed photos or videos on a screen). This is a first filter to assess the robustness of the solution against basic attacks.
• Level 2 iBeta evaluation is designed to evaluate liveness detection systems against more sophisticated attacks, such as 3D masks and HD videos. The requirements are therefore higher, guaranteeing a higher level of security.

The choice of iBeta evaluation level will depend on the use case applied to liveness detection, but a level 2 evaluation will inspire more confidence.

An iBeta evaluation is not a certification...

One of the most common mistakes is to believe that an iBeta evaluation is equivalent to a certification. As a reminder, iBeta is an "evaluation" institution, not a "certification" one! The error comes from the fact that the evaluations conducted by iBeta are "compliant" with the requirements of ISO 30107-3. This means that the laboratory uses ISO standards to evaluate the quality of liveness detection solutions, but in no case does it issue an ISO certificate. Other organizations are responsible for issuing this certificate, and again, a good iBeta evaluation will not necessarily lead to certification. We explain why!

How do iBeta evaluations differ from each other?

To understand why not all iBeta evaluations are equal, it is necessary to define two key terms related to liveness detection: APCER and BPCER.

APCER, or Attack Presentation Classification Error Rate, is an indicator that represents the proportion of presentation attacks incorrectly classified as a real person by a liveness detection system. The higher this rate, the more compromised the system's security.

BPCER, or Boniface Presentation Classification Error Rate, is an indicator that represents the proportion of real people wrongly classified as presentation attacks by a liveness detection system. The higher this rate, the worse the user experience.

Initially, iBeta did not set BPCER limits for liveness detection solutions during evaluations. This means that at the time, if a solution was indeed effective in detecting attacks, we could not know what it was for the user experience.

Since then, iBeta has set a BPCER threshold limit of 20% and then 15%, which implies that some solutions have been able to obtain the iBeta label under much more favorable conditions. The right reflex to have when choosing your liveness detection solution would be to check the BPCER obtained during the evaluation. This information is available on the "Certificate of Conformity" issued by iBeta.

Furthermore, some iBeta evaluations are conducted on specific targets and not others (mobile or computer), casting further doubt on the real effectiveness of the technologies. The universality of a liveness detection solution on any type of device is a necessary condition to cover all remote identity verification journeys and maintain continuity in the user experience. Find out in this article how the user experience in the context of remote identity verification can reduce acquisition costs.

Conclusion

iBeta evaluations are an excellent indicator of the performance of a liveness detection solution. However, it is essential not to rely solely on this label. To make the right choice, you must go further and examine closely the evaluation criteria, the results obtained, and the conditions under which the tests were conducted. You can find  in this article our tips for choosing your facial biometrics solution.

‍