Maybe one day we’ll be able to breathe on phones to unlock them, a study suggests: Researchers have developed a prototype scent sensor that, combined with machine learning, can sniff out people’s identities based on their breath. The “artificial nose” already achieves a high level of reliability, but the concept still needs to be refined before it is ready for the market, the developers point out.
Access control to banking, to computers and mobile phones or to smart vehicles: Biometric recognition methods can offer a convenient and secure alternative to traditional authentication methods in the field of information technology. Various personal aspects have already been used as access keys, including the individual characteristics of fingerprints, irises, face or voice. So far, however, all these methods have been based exclusively on the physical aspects of a person. “These techniques have disadvantages: the features can be copied or affected by physical injuries,” explains Chaiyanut Jirayupat from the University of Tokyo. Therefore, he and his colleagues have now explored the potential of a human biochemical characteristic as a new method for the biometric security toolkit: the personal signature in the breath.
Recognition based on the individual breath odor
In recent years, researchers have already shown that interesting information can be found in volatile substances that people utter. It is well known that dogs can not only easily identify people by smell, but can even sniff out certain diseases – including in the air we breathe. Although technical noses have not yet reached the level of animals, odor sensors can also detect even small concentrations of volatile substances in the air. As Jirayupat and his colleagues explain, breathing air could be particularly suitable for biometric authentication because, in contrast to perspiration from the skin, it has significantly higher concentrations of odorous substances.
To explore the potential of the concept, the team first started with a basic analysis of subjects’ breath to find out which connections could be used for biometric authentication. As the researchers report, 28 compounds emerged as useful indicators: people release them in a concentration distribution pattern specific to them, which is apparently linked to their genetic predisposition. On this basis, the scientists then developed an olfactory sensor array with initially 16 channels, each of which can identify a specific range of compounds.
With the previous prototype of the detection system, you blow into a bag. This breath sample is then pushed through the olfactory sensor array. The captured data is then fed to a machine learning system that analyzes the composition of the human’s breath and develops a profile from it. The corresponding odor signature can then be used to identify the respective person, the scientists explain.
On the way to the mobile phone with a nose?
The researchers tested the system with breath samples from up to 20 people and found an average accuracy of the identification system of 98 percent. “It was a heterogeneous group of people of different ages, genders and nationalities. It is encouraging that we have already achieved such high accuracy,” says senior author Takeshi Yanagida from Kyushu University in Fukuoka. As he and his colleagues explain, it is already becoming apparent that the system could also guarantee absolute detection reliability by integrating additional sensor channels.
Nevertheless, the team emphasizes that some development work is still needed before the process can enrich new generations of smartphones and the like. So far, the question remains open as to what extent it is susceptible to disturbances caused by food intake. Because the artificial nose must of course also be able to identify someone by their breath who, for example, has just eaten a garlic baguette. “Previously, we had our subjects fast for six hours before the test,” says Yanagida. But the team is confident: “We have laid a good foundation. The next step will be to refine this technique. Our results already show that any problems can be overcome by adding more sensors and collecting additional data,” says the scientist.
Source: Kyushu University, professional article: Chemical Communications, doi: 10.1039/d1cc06384g