With the expansion of biometric registration for a wide range of personal uses, it seems that we are one step closer to the future. After all, authentication with our biometrics is more efficient, since our fingerprints are, by nature, unique and difficult to copy. But all this progress raises other questions: are there risks in using facial recognition?

after the recent episode of Cufa, we spoke with Christian Perrone, coordinator of the Law and Technology area at the Institute of Technology and Society of Rio de Janeiro (ITSRio). In addition to clarifying the implications of the case, the expert explained to Vida Celular a series of questions surrounding what is really at stake with facial recognition.

Facial recognition is like a virtual memory

As we published around here previously, facial recognition as an individual identification is not that old, and it was only in 2011 that the technology began to be incorporated into our smartphones. In its most basic structure, the technology has two parts: one that identifies and maps, and a biometric database of faces to compare. Without one of the sides, it does not work properly.

Perrone explains that facial recognition technologies work, in a way, like our own memory. “It’s like looking at an old photo album and knowing that the person in there is a childhood friend.” The expert explains that detection occurs in three stages.

First, the machine, an artificial intelligence, identifies that the image contains a person's face. After that, the machine maps this face, transforming it into what is called biometric data. Finally, it compares the facial map with a database to determine the specific person in the photo. That is, if the person is in the database being compared.

The problem with digital memory is that it also fails

Today, this “digital memory” plays an important role in device security. Devices like the iPhone, which now use the feature to unlock their devices, can now use biometrics to authorize payments. Facebook, for example, is studying to implement the technology in your smart glasses and even on WhatsApp.

However, facial recognition is not without its risks. In the documentary Coded Bias, MIT researcher Joy Buolamwini highlights how visual biometrics are not as effective for faces of people who are not male or white, generating what experts call “racial bias.”

“Part of her study showed that many of these tools did not recognize that they were female faces,” Perrone explains. “For example, most of these technologies have much more difficulty recognizing black women, even with images of celebrities like Oprah or Michelle Obama.”

Recognized in the field of artificial intelligence, racial bias is one of the most recurring difficulties of machine learning. To work properly, databases need to be trained to be diverse — and to do so, work with a wide range of profiles. Facebook released a series of videos for programmers to test the recognition and identification capacity of their algorithms.

Inability to adequately recognize impacts public services

In Brazil, facial recognition involves two challenges at once — implementing the technology without racial bias, and in the correct way. In certain circumstances, biometric inaccuracy can pose risks to civil society.

“Could it [facial biometrics] not be having a form of exclusion as a consequence? Could it not be impacting one population group more than another? This technology being used in a subway, there are people who cannot stop using this means of transport. So, it could have a disproportionate impact on this population.”

In fact, the Brazilian government already applies the use of facial biometrics in some of its programs, such as, for example, Proof of Life, for the benefit of the INSS. In this scenario, technology is a mechanism to prevent fraud, recognizing whether the person is, well, alive. Facial recognition with racial bias can pose risks to access to these public services.

Misuse threatens individual freedom

On the other hand, improper facial recognition can threaten individual freedom. In the past, a use of facial biometrics for reaction detection on the São Paulo subway improperly collected data from passengers. When used in an advertising campaign, the technology invaded the privacy of users who, if they did not want to participate, would have to refrain from using public transport.

The risks become even greater when facial recognition is implemented in the area of ​​public security. “This allows the State or whoever is behind using the technology to be aware of people’s steps, where they are, what they do,” explains Perrone.

In the United States, civil entities have already spoken out against the application of facial biometrics in public security. The expert claims that the inappropriate use of facial recognition ends up resulting in the creation of a surveillance state.

“Imagine the impact on people’s freedom of movement; on the possibility of peaceful assembly; of protesting; and even on matters of personal privacy, such as a teenager discovering his or her sexuality. All of this can not only be captured in images, but you can literally know who the person is, what they are doing and who they are with.”

General Data Protection Law works, but could be better

In Brazil, some of the risks of facial recognition are mitigated by the General Data Protection Law (LGPD). The clause treats facial biometrics as sensitive information, and its use must be clear to participants. However, there is still room for progress.

“Likewise, there is still no clear regulation regarding the use of facial recognition for a number of purposes, such as public safety,” concludes Perrone. “The question here is less when it should be used and more when we believe the balance between risks and opportunities favors its use in a safe, responsible and healthy manner.”

Image: filadendron/iStock