Biometric Authentication Types

Of the different types of biometric authentication that we see, there are also different types of concern that people may have – that we may have as a provider of the authentication service, and also, as a customer or citizen, that they may have when using the biometric service. So for facial recognition, typically it is medium accuracy. There have been some known, as we said, people have 3D printed copies of faces. So it’s medium accuracy. In terms of invasiveness, it tends to be very low. To log in using facial recognition just requires you to look at the device, typically. For acceptability, for this reason, acceptability is medium.

There are some privacy concerns that people typically raise around taking copies or creating a digital template of your face. And we saw recently in a news story in China that a journalist was identified as jaywalking from her picture being recognized on the side of a bus using an automated facial recognition system. So we want to keep the acceptability in mind. The throughput is, again, medium. It’s not the fastest process. Although, in recent years, the throughput has increased significantly. For fingerprints, thought to have high levels of accuracy. But again, this varies depending on the number of points, number of samples that we are taking.

The accuracy can be reduced through poor implementation or through a lower number of points used deliberately within an implementation. For invasiveness, again, we see here some concerns over the use of fingerprints, partly because historically fingerprints have been used as part of tracking criminals. And so people are sometimes a little wary of giving their fingerprints, and they assume that this information may be shared with public bodies. Acceptability, for that reason, is medium. But fingerprint scanners, the throughput now is relatively high. It’s become, over the last five years, a very, very quick process for authentication. For iris and retina, and the two are different.

Iris is the colored disc around the retina, and the retina is the black dot in the centre of the eye. For the iris, we’re looking at the pattern of that colored area. For the retina, we’re looking at the pattern of the blood vessels within the eye. Both are very similar, though, in terms of accuracy, invasiveness, acceptability, and throughput. Very high accuracy. Perceived as being very invasive because, for registration and also in use, you have to physically put your face against a reader, your eye against the reader. Sometimes these shine light in your eye to improve the reading and the registration process, which creates discomfort for the user. So for that reason, acceptability tends to be very low.

And also, throughput is very low. We’ve seen this in some airports that have been using retinal scanning, where we’ve had long queues for the biometric-processed passports. For voice, accuracy tends to be medium. Again, examples of it being compromised. Invasiveness is low. This can be passive, and you see banks increasingly using this to maintain authorization through a telephone call, for example, through a transaction. Does the voice continue to match? Acceptability, for that reason, is high. People don’t tend to be too concerned about their voice being taken and used as an authentication medium. Throughput, though, is medium. Throughput, we typically need the individual to speak a few words before we can process an authentication request.