Myths surrounding biometrics, often exaggerating its capabilities and the risks associated with personal data usage, can create user wariness and hinder the adoption of new technologies. This is further fueled by the unrealistic depictions in science fiction films and a general lack of information. This article aims to shed light on the collection and usage of biometric data, providing a clearer understanding of the technology.
In the modern world, there is a fear that people will have no personal space left where they can be alone, their images are captured by biometric systems and thus become available to third parties who can use them for any purpose.
When biometric systems "capture" an image of a face or other body part, they process the information and convert it into special digital objects called biometric templates. Therefore, photographs and other images are not stored in biometric databases, like fingerprint images in police systems. Biometric templates can be compared to complex formulas that are completely unreadable to humans. When processing an image, the system compares the template to others, for example, one already stored in the database or was simultaneously read from an access card. Essentially, it is about determining the close correspondence of two mathematical formulas in some common measurement space. They do not contain any personal information: last name, first name, date of birth, address, etc. Therefore, it is difficult to extract any benefit from them without linking them to personal information.
This is not always and not entirely true. On the one hand, biometric systems are indeed a "big" AI technology. They perform a huge amount of mathematical operations every second, and such performance simply cannot be cheap. On the other hand, biometrics is rapidly becoming mainstream and widespread, which leads to a rapid decrease in the price of an individual device or software solution. These include access systems for an apartment or office or customer identification systems with analytics collection in convenient cloud services.
Modern biometric algorithms are tested on huge amounts of data. In some extreme cases, the entire population of the Earth can be placed in one large database, and biometric technology will quite cope with a successful and accurate search in this database - provided that the quality of the initial data is high and even. If, during operation, the biometric system receives data of sufficiently good quality and detail, it will be reliable enough for any business.
In the past, facial recognition systems could make mistakes due to lighting conditions, weather conditions, changes in the user's appearance, or clothing that obscured the user's face. Over the past six years, the accuracy of facial recognition by neural networks has increased 50 times, reaching a rate of over 99% for databases containing hundreds of millions of faces.
Why not 100%? Because any recognition based on neural network technologies is a probabilistic method - the comparison of biometric templates to establish their similarity is always done with some finite accuracy, in other words, the level of strictness of the algorithm.
This myth is based on real cases where the facial recognition system in smartphones was successfully tricked using a very high-quality 3D mask. In all cases, biometrics significantly increases the level of security. It is used to prevent unauthorized persons from acting on behalf of the user, either by direct forgery or by using captured passwords and documents. For example, it increases the reliability of two-factor authentication, when the password is confirmed by a biometric feature calculated "on the spot" based on a fingerprint, face image, or iris.
In this article, we’ve described common myths about facial recognition technology, revealing it as a helpful tool rather than something to fear. From everyday conveniences like unlocking phones to streamlining business processes and enhancing public safety, facial recognition plays a positive role in various sectors of society.