Seamless authentication is a method where users are authenticated without having to provide any credentials. While the concept itself is nothing new, recent developments in machine learning, image recognition etc. has made it more viable than before, and it is something Rakuten Institute of Technology is exploring actively. The ultimate form of this is using biometrics, such as voice, fingerprints or facial recognition to identify a user, as this enables access without the need of anything else than your own body.
In a step towards this, we started a pilot project called Secure Facial Payment together with Rakuten Pay, Rakuten’s payment solution business. The system allows users to simply use their face and a PIN number to safely pay for items in stores, without the need for cash, cards, phone or anything else.
While all options come with pros and cons, facial authentication has some strong characteristics as a biometric authentication method. One of its strengths is that faces are universal and easy to collect, and it is generally well accepted by users as an authentication tool. In addition, the method of capture, in this case cameras, are not very expensive. The downside is that traditional face authentication has poor accuracy compared to other biometrics.
However, thanks to recent developments from eigenfaces through deep learning and convolutional neural networks, facial recognition technology has evolved continuously. We won’t get into too much technical detail here, but the latest models show over 99% accuracy by adopting specific loss functions called triplet loss. While the accuracy of facial recognition has improved dramatically, it has still not reached the level necessary to safely use for authentication on its own.
In order to overcome this weak point, most services use some other information as a second factor, like the user’s phone-number. However, this is also not secure, as phone numbers are often more or less public information that people share freely. Our solution to this is the addition of a personal PIN number as a second factor, making sure that the user identified is the one it seems, and initial tests have been very promising.
We just successfully finished an internal alpha test at a Rakuten Café in the Rakuten offices, and Rakuten Institute of Technology and Rakuten Pay will continue to pursue these collaborations to accelerate the innovation in the Fintech field. The next step is a larger scale internal test, and then from there we will investigate further how we can bring it to the wider public.
For now, you can see a video and some photos from the tests below.