To fight against the sale of gambling to minors, the FDJ is testing a solution to check the age of players using facial recognition. But what about the reliability of the technology and the confidentiality of personal data?

To fight against the sale of gambling to minors the

To fight against the sale of gambling to minors, the FDJ is testing a solution to check the age of players using facial recognition. But what about the reliability of the technology and the confidentiality of personal data?

La Française des Jeux (FDJ) continues its fight against the illegal sale of gambling and games of chance to minors. Indeed, according to a study by the Sedap association carried out in 2021, 35% of young people between the ages of 15 and 17 have played gambling at least once, proof that the law is not always applied by the 23,500 or so tobacconists in France. This is why, within the framework of the International Cybersecurity Forum (FIC) which was held on April 5, the FDJ announced that it had tested, for the past three months, a new solution making it possible to verify the majority of customers thanks to the facial recognition, as reported The Digital Factory.

FDJ facial recognition: terminals to verify the majority

Tobacconists have installed terminals equipped with facial recognition software, developed by the start-up Yoti – whose technologies have been validated by the British government and German digital regulators – to determine whether the customer is major or not. He positions himself in front of her and a photo is taken and then analyzed to determine his age, and therefore to confirm their majority.

On her blog, Yoti explains that “our technology can accurately estimate a person’s age by analyzing their face. We designed it to provide everyone with a safe and private way to prove their age in different daily life scenarios: from age verification on social platforms and online stores, to automatic checkouts in supermarkets, and more. by bars and nightclubs.

FDJ facial recognition: fast and accurate results

This solution is already used in British hypermarkets to verify that a customer is old enough to buy alcohol, but also by Instagram, which now requires recording a video selfie which is then submitted to an algorithm (see our article). Its use for the FDJ is still only at the prototype stage, but it still raises questions about the reliability of the technology and the confidentiality of customers’ personal data…

The start-up promises that the result can be displayed in less than 2 seconds, with an accuracy rate of around 1.6 years for those under 20 – below this age, however, the trader must request the identification papers. player identity. But what is it really? According to one external studyYoti’s tool is 98.89% reliable in determining whether a person is over or under 25 years old. AT writing, we tested the technology demo version, which is available online, and the age displayed turned out to be consistent with our actual age. However, it is quite possible to trick the software by trying to look older, with make-up for example.

FDJ facial recognition: reliable and confidential results?

Of the tests carried out in 2019 show that the margin of error can increase depending on different audiences. Thus, skin color can influence the result, especially for women with dark skin, for whom the average error can go from 1.6 years to 4.8. This would be explained mainly by the under-representation of the group in the data with which the software was trained. During the conference, Yoti’s representative himself more or less acknowledged that the system was not infallible, explaining that a life on the streets prematurely ages the face, for example. Facial recognition cannot therefore completely replace human control for the moment.

There is also the question of privacy and data protection, even if Yoti promises that the photos will be deleted from the server immediately after use and that the technology cannot recognize an identity – only detect the age – in order to guarantee the anonymity of the person. Still, the start-up must respect its commitments in this area. For example, it could use it to train its software…

It is not impossible that this technology will soon be adopted by the Government as a short-term solution in order to prevent minors from accessing pornographic sites, until it sets up the double anonymity system (see our article ). Indeed, the CNIL is considering verification by facial identification and Yoti is a member of the Laboratory for Child Protection, set up by the Élysée in November 2022. However, it is still quite possible to deceive the tool simply by showing a photo to the camera or asking someone older to take the video selfie…

ccn5