You are unique. Your breath too, according to research in India. To the point that breathing could become a biometric tool to unlock your phone. A future alternative to fingerprint scanning or facial analysis?
In the future, maybe you'll unlock your smartphone not by typing a PIN or scanning your fingerprint, but simply by breathing next to the mic. In any case, this is the perspective that we can draw from reading the article by New Scientistpublished on January 12, 2024 and which echoes research work carried out in India.
Researchers from the team Mahesh Panchagnula, from the Indian Institute of Technology in Madras, initially worked on a project consisting of identifying people who suffer from respiratory problems, in order to take care of them. It was about designing an artificial intelligence model to automate the process.
A unique airflow, influenced by the respiratory tract
But by examining the data collected so far, they noted a possible new direction: what if breath analysis could also be used to positively identify a given person? Not by focusing on breath and its composition (there are also works on this theme), but on the air flow itself.
This air flow is, in fact, influenced by the morphology and shape of the respiratory tract — pharynx, larynx, nasal cavity. Therefore, we do not exhale quite in the same way as others. In any case, this is the path that Mahesh Panchagnula's team has embarked on, report NDTV And Koriiwho cite New Scientist.

As part of the Indian experiment, 94 people were asked to take 10 breaths each in a device capable of recording atmospheric variations up to 10,000 times per second. They then fed this data into an AI model.
Once it was trained, it was subsequently able to determine whether or not a new breath corresponded to a person already identified, with quite significant precision – we are talking about a success rate of 97%. A subsequent test, with more difficult parameters, gave an accuracy of 50%.
In this other test, the model was tasked with identifying a breath, but without having prior knowledge of the person concerned. He then had to associate this breathing with that of one of the two people offered to him. Artificial intelligence is said to have been right one out of two times, it is reported.
One of the advantages of biometric recognition based on breath is that it is closely linked to the person's situation. This is what New Scientist points out: a dead person does not breathe. From then on, the biometric key that relies on his breath is irrevocably lost — unless it can be copied somehow.




What practical use for recognition through breathing?
Biometrics is a technology used to identify what belongs to a person. We know the biometrics common in smartphones, to unlock the home screen. This may involve facial analysis, scanning a fingerprint, or reading the iris of the eye. This is eminently personal data.
Other biometric signals may exist: voice, silhouette, gait, physical attributes, behavior, gestures, and even the venous network. The chemical composition of the breath is also a key to recognizing a person, like the variations and intensity of the breath, according to the work of the Indian team.
There remains the question of the reliability of the device and the engineering work that will need to be accomplished to hope to integrate this type of sensor into a device as thin and small as a smartphone. It's one thing to carry out an experiment in the laboratory, with dedicated equipment. It's another to give it a concrete application.


However, this alternative to traditional biometric recognition may not be aimed at the general public or, at least, not at this type of device. Why not, for example, consider it for a car to verify that the person behind the wheel is indeed the owner of the vehicle? This is a possible direction.
This work raises questions that research has obviously not (yet?) raised: can the system isolate a given breath if there are several people nearby? What also happens in the event of disturbed breathing – shortness of breath, for example? A negative answer would then greatly limit the practical interest of such technology.