From Flakey Fingerprint Readers to Bullet Proof Biometrics

Date: 17.02.2012

I had a demonstration today of a new biometric reader and it really opened my eyes. Thankfully it wasn't iris recognition - I'm not sure the “pun-ometer” could stand that.

To be frank I have tended to steer clear of anything to do with fingerprint reading in the past, as it has had a reputation for being unreliable.

The issue is that the optical sensor technology used looks at the surface of the fingerprint, which is easily altered by dirt, dust, grease and other common substances.

Today I witnessed a reader reliably recognising a finger through a layer of latex gloves which adds a whole new dimension of convenience and robustness.

Multi-Spectral Imaging

Multi-spectral imaging captures radiation at a range of different frequencies, not just visible light which optical sensors are confined to.

Just as radio telescopes allow us to see deeper and deeper into space, multi-spectral fingerprint sensors allow us to see deeper into the surface of the finger - up to 4mm deep in this case - so the reader is not just looking at the surface fingerprint.

The results of this are low false rejection rates (about 1 in 1000) and extremely low false acceptance rates (around 1 in 10 million) well suited to high security applications.

Resistance to Spoofing

ATEC have had a lot of experience with hand geometry readers, a technology popular on construction sites as it deals well with dirty hands. It is however quite bulky, needs an inconvenient housing when fitted outside.

More importantly to some it can be easily spoofed - we managed to fool one in our lab with a false hand made for the purpose. Because the multi-spectral technology allows us to see deeper inside the finger, it allows for some robust anti-spoofing measures, including looking for oxygenated blood which would be difficult to fake.

Data Protection Concerns

Fingerprints are a sensitive data protection issue. People are quite rightly concerned if their biometric data is going to appear on a database in readable form. Often this is addressed by putting the biometric data an access control card in encrypted form.

In use the template is loaded from the card and the reader performs a 1:1 match. This requires less processing than a 1:many match and is more secure because it requires two credentials (card and finger) but is fundamentally less convenient for the same reason (it's difficult for most of us to forget to bring our finger to work).

This reader uses the 1:Many matching principle, but crucially doesn't store a fingerprint as such. It takes key points from the finger print and stores those.

Also the finger data is never stored in the same place as the personal details - the two are only associated with a card number. This should address privacy concerns but sometimes perception can get in the way of the facts and there may still be some barriers to overcome.

Summary

A good biometric reader should offer convenience and robust performance in the designated environment. With the help of multi spectral imaging fingerprint reading technology seems set to leave it's flaky reputation in security applications behind.

Go Back