Affordable Biometrics

[BackstageTop contributor]

Affordable Biometrics
A team at Yahoo Research Labs - including Sri Lankan Senaka Buthpitiya – are adapting the low-resolution scanning capability of a smartphone’s touchscreen to work reliably as a biometric scanner

By Isankya Kodithuwakku.

Published on June 16, 2015

Biometric technology, which can recognise people from their fingerprints and irises, is widely used in access control systems for buildings. High-end smartphones use fingerprint-scanning technology as a passcode to recognise the device’s owner. The technology requires a fingerprint reader to work, a miniature version of which is costly, which limits it to the most expensive smartphones.

Three Yahoo Research Labs scientists, including Sri Lankan Senaka Buthpitiya,have collaborated to adapt a smartphone’s touchscreen capability to work as a biometric scanner. The project is called Bodyprint. The challenge was that since a smartphone flat screen scans at a low resolution, it’s unable to reliably scan fingerprints. So the three scientists have adapted it to scan the contours of an ear or palm instead.

Tests on a dozen users have shown 99.5% reliability of allowing access only to the right person. The sensor algorithm, however, also turned away the right person 26% of the time. They are working on improving the technology.

Buthpitiya did his undergraduate studies in Computer Engineering at the University of Peradeniya after completing his secondary education at Trinity College, Kandy. At university,he worked on several research projects and became interested in pursuing research as a career. He completed his Ph.D. at Carnegie Mellon University, Philadelphia.

yah3How did Bodyprint come about?
At Yahoo, we have many mobile app users, and we care about data security. We came across studies that showed that users rarely protect their mobile devices through PINs or passwords. But that changed when biometric sensors, like fingerprint scanners,were introduced on a small number of
high-end mobile devices. That got us thinking: how do we bring the convenience and reliability of biometric authentication to all mobile devices?

It occurred to us that touchscreens themselves scan using the same capacitive technology as fingerprint scanners – at a lower resolution. All we had to do was to repurpose a smartphone touchscreen as an image sensor and scan part of the user’s body large enough to be recognized by a touchscreen. Hence, the name Bodyprint. Eventually, we settled on users’ ears and hands, which proved to be very reliable for identification.

What kind of work went into it?
A touchscreen reports (x,y) touch positions. Initially, we didn’t have any images to work with. So we developed our own touchscreen interface to read a series of capacitive measurements from the touchscreen in 2D, which finally gave us an image. To identify people using this signal we recreated “what the screen sees” from this signal. We had to do some signal processing work and then some image processing work to create a moving image from the signal.

To identify people, we needed to turn this stream of images into a few snapshots, so we essentially collapse the video into a series of still images maintaining as much of the detail as possible while avoiding blurring. Then we do some more image processing focusing on the important areas of the images.

Finally, we use some machine learning techniques to identify the defining features of the body part, the ear for example, to learn to reliably identify these points over time.

When can we expect this technology in devices?
In its current scope, Bodyprint is a research project at Yahoo Labs, and we continue to investigate interesting research projects on user authentication.

Tell us about working at Yahoo Labs?
Working at Yahoo Labs is great. We get to define our projects, pitch the ideas to the company and then work on them. We have a lot of freedom in defining projects and driving them and publishing our work. The best part is the freedom to work on cool research ideas with the backing of a company with huge resources. The free food, gym, recreational areas and activities don’t hurt either.

What are the other research areas you’re interested in?yah
While I can’t go into specifics here, I’m focusing on big data to improve personalization.

What do you expect to see in smartphones of the future?
Devices in the near future will get a lot more personal. At the moment, phones have very little personalization, but over the next few years, I think we’ll start to see devices begin to understand the user and his surroundings better. Devices will begin to know what you like, what you don’t like, who you’re with, where you are, what you’re doing and provide a customized experience based on these factors.

What’s the future of Bodyprint?
At this point, Bodyprint is a research prototype that demonstrates the feasibility of our approach. We explored the feasibility of bringing biometric user identification to smartphones that don’t have a dedicated biometric sensor. That is, to work on any commodity smartphone with a capacitive touchscreen, which are the most common smartphones available in the market now. We accomplish this by using the touchscreen itself to scan users’ biometric features when they hold up their phone against the ears, knuckles, palms, or fingers.

We also show that ear-based identification on today’s devices yields stable biometric features allowing us to achieve a high authentication precision (99.8%) with a low false-rejection rate of only 1 in 13.

Will this be another intrusion on privacy, scanning and saving your skin data,etc?
Bodyprint currently operates on the device. The information is stored locally and not shared, so all biometric data is securely on the phone

What are the minimum smartphone specifications to run this technology?
yah2One big advantage of Bodyprint is that it uses the capacitive touchscreen to sense biometrics. Since all current smartphones incorporate a touchscreen with this scanning technology, Bodyprint has the potential to run on a wide variety of mobile devices.

Fingerprint scanners provide reliable user identification. Unfortunately, they are expensive and available only on high-end devices. Bodyprint in contrast uses the touchscreen for scanning, a component that is part of every existing smartphone and therefore has the potential to bring biometric
authentication to every smartphone.

What is the engineering challenge with this technology you face now?
Bodyprint produced an impressive identification accuracy of 99.98% with the 12 participants. To understand if the Bodyprint’s approach would scale, we need to test our prototype on a larger group.

Is Yahoo funding all the research of your team?
Since this is a Yahoo Labs research project, resources are provided by the Lab. Bodyprint was developed by the project’s three authors Christian Holz, Marius Knaust and myself.