The air turbulence created when breathing could be used as an effective biometric test for unlocking smartphones or other devices 鈥� one with the morbid advantage that, unlike other tests such as a fingerprint scan, it cannot be passed by a dead person.
at the Indian Institute of Technology Madras, Chennai, and his colleagues started experimenting with breathing data recorded by an air velocity sensor out of curiosity. They had hoped that an AI model could interpret the data to identify people with breathing difficulties who might benefit from medication. But they found that the data actually told them a lot more.
They recorded 10 breaths from each of 94 human test subjects, using an air velocity sensor to take readings 10,000 times a second. That data was then fed into the AI model.
Advertisement
During their experiments, the researchers discovered that the model, once it had analysed breath data from a particular subject, could聽verify that a new breath either did or did not come from that individual with more than 97 per cent accuracy.
In a second test where the model was asked to identify a breath without first being told who was logging in, it was able to correctly identify the person as one of two possibilities more than 50 per cent of the time.
Panchagnula says the AI model is detecting unique patterns of turbulence caused by the different shapes of each person鈥檚 extrathoracic region 鈥� made up of the nasal and oral passages, pharynx and larynx.
He adds that the results can more than likely be improved drastically with refinement, but that the initial findings show the concept is viable. And while we arguably already have enough biometric systems, Panchagnula says a breath test would have one unique advantage: it only works on living people. Some news reports suggest that the , but the new system would be safe from this vulnerability.
鈥淵ou can鈥檛 make a dead person pass this test because you need the person to exhale.鈥�
arXiv
