Posted on 5. December 201821. January 2021 by Ann-Kathrin Freiberg

Deep Learning Revolutionizes Biometric Authentication

Why the new NIST face recognition vendor test implies to use Biometrics as a Service (SaaS)

Every few years, the NIST face recognition vendor test (FRVT) compares various biometric companies concerning their performance. Traditionally, established vendors like NEC and Cognitec lead the field. But, this year’s results indicate a massive change for the whole industry: Surprisingly, vendors that had been at the forefront for decades were surpassed by far concerning the accuracy of their biometric recognition. To be precise, the winning participants were 20 times (!) as good as the industry standard of the last NIST Face Recognition Vendor Test. A major reason for this extreme development is advances in the field of artificial intelligence (AI) and deep learning.

The test results imply a major consequence: A biometric authentication system bought 3 years ago, is completely outdated now – unless you had decided back then to use Biometrics as a Service in a SaaS model. With SaaS, vendors which constantly enhance their offer can easily let their customers be part of the innovations. Industry-changing developments like the FRVT’s results can dynamically be dealt with through auto-updates in a centralized server solution. For companies, using Biometrics as a Service additionally brings the advantage of high scalability and cost efficiency. From the customer’s point of view, a convenient and frictionless user experience is central. In a cloud solution, one biometric face enrollment is enough for all devices used for authentications. With the rapid developments happening right now in the biometrics industry, this is the way to keep your systems up-to-date and your customers satisfied.

Big Data and AI in Face Recognition

Innovations in deep learning technology are revolutionizing the way we think, the way we work and the way we use data. For instance, the winners (e.g. Microsoft) take great advantage of the new possibilities generated through Big Data and AI. To train the algorithms for face verification, giant data sets of millions of faces were used. The important difference between traditional vendors and companies like Microsoft is the access to huge amounts of data. This opens unknown opportunities for training recognition engines. The ultimate result is an extreme discrepancy between the established vendors and the new players. Deep convolutional neural networks for image classification are the key to generating new levels of accuracy in face recognition.

Finally, there’s one good news and one bad from the customers’ point of view: On the one hand, the results have shown that biometric authentication, like any other industry, is subject to change depending on technological advancements. Massive changes in standards and processes can occur. On the other hand, there’s no need to keep away from biometric authentication software as it constantly evolves and gets better. This can be dealt with using Biometrics as a Service in a SaaS model. This way, companies can leverage from the constant enhancement of the whole biometrics industry. This way they stay at the forefront of innovations automatically. Biometrics as a Service is the natural answer when it comes to deciding for a certain technology in a rapidly moving industry like ours.