Man versus machine: which provides the highest assurance levels?

May 5, 2020 by Joe Bloemendaal

BiometricsDuring the coronavirus lockdown, as millions are working and shopping from home, verifying people digitally has never been more vital.

It’s not just banking. Government benefits, health services, online education, dating companies and gaming are just some of the sectors witnessing a huge surge in demand for digital know-your-customer services. The outbreak is also proving fertile ground for fraudsters exploiting the global rise in onboarding. This is expanding the use of digital authentication.

Many claim that fully automated processes are the fastest way to alleviate risks and reduce fraud. However even in 2020, most forms of ID are stuck in the analogue world. There is no comprehensive, interoperable digital identity that exists anywhere in the world. Fraudsters are hot on our heels creating fake IDs and we estimate that about 20 per cent of the techniques they deploy each year are new. You cannot account for all types of ingenious fraud in an automated system.

The science behind digital authentication has become more complex and automated than at any point in history. Machine-learning algorithms can digest multiple security features in a passport or driving licence and know whether it’s fake. This can be combined with a selfie taken by the consumer on their cellphone which verifies them using liveness detection.

Fully automated systems can be faster and are sufficient for many requirements. Computers can be better at facial recognition, and artificial intelligence-powered technology can recognise a valid ID as genuine and match the ID photo with a selfie using biometric facial comparison. But while algorithms are powerful, they are not perfect and there are exceptions that the technology has not yet been trained to recognise. This is where agent- assisted solutions come into their own. Human experts are still better at spotting fraud in some forms of paper ID. For this reason, we still need people to assist in digital onboarding. With over 36 years of experience, Mitek knows the importance of the human expert who knows what the latest fraudulent activity looks like.

Man and machine must work together to reach the highest levels of assurance. Machines don’t teach themselves; they learn from experts. Capturing the latest fraudulent ID’s and feeding these into powerful analytical systems allows computers to spot fake documents, continually improving the machine’s abilities. It’s not man or machine, it’s about using the best of both to achieve the highest possible assurance levels.