A Breath of Fresh Air in Biometric Authentication

You can ditch the passwords and breathe on your smartphone to log in to your accounts, thanks to this research out of Kyushu University.

Nick Bild
3 days agoMachine Learning & AI
(📷: Kyushu University)

Remembering passwords for dozens of different accounts can be a real headache, especially when those passwords have to be long and contain many special characters. And if those passwords need to be changed frequently, we all know they are going straight on a Post-it Note on the monitor, against all of our better judgment. For all the trouble we go to to remember all of those passwords, they are still vulnerable to being stolen or guessed in a brute force attack. One of the ways that we might get out of this mess is through the use of biometric authentication methods. These methods use a person’s physical characteristics — a fingerprint pattern or their facial appearance, for example — to allow access to a secure system. This eliminates the need to remember passwords, or to have to create a different secret for each system that one wants to access.

This sounds great in theory, but unfortunately, many physical characteristics can be spoofed to allow unauthorized access to protected resources. In the search for a new biometric measurement that may be more difficult to spoof, a group led by researchers at Kyushu University's Institute for Materials Chemistry and Engineering have done some research that may breathe new life into the field. Using an artificial nose, they have explored the possibility of uniquely identifying individuals by examining the chemical makeup of their breath.

The artificial nose contains a 16-channel chemiresistive sensor array that is able to capture a detailed chemical fingerprint of the exhalations of an individual. From this data, a total of 28 compounds that have been identified as viable options for identification purposes are measured. It is nonobvious how these compounds, and their concentrations might be used to uniquely identify an individual, so the team designed a machine learning algorithm to help with the task.

After training the machine learning model, it was tested on a cohort of six subjects. With this small group, the team found that they could accurately identify individuals 97.8% of the time on average. After these encouraging results were uncovered, they decided to run another test with a larger group of participants. Increasing the sample size to 20 individuals showed an equally high rate of accuracy. The group was diverse in terms of age, sex, and nationality, which bodes well for more widespread use of the device, however, larger scale trials will be needed before we know if breath authentication would be likely to work well at uniquely identifying individuals in a large population group.

The present work also required subjects to fast for six hours before using the breath identification system, which is clearly impractical for real-world use. Fortunately, however, the team learned during the course of their work that this obstacle could be overcome by adding more sensors to the artificial nose, and collecting more data to train the machine learning model. You may not be able to unlock your smartphone with a breath just yet, but with additional work, this method may one day be a viable option to eliminate the need for those pesky passwords we still have to remember.

Nick Bild
R&D, creativity, and building the next big thing you never knew you wanted are my specialties.
Latest articles
Sponsored articles
Related articles
Latest articles
Read more
Related articles