Concerns over technology, ethics as French politicians embrace facial recognition - FRANCE 24

Issued on: 15/10/2019 - 18:20Modified: 15/10/2019 - 18:24

France has seen several initiatives to encourage the use of facial recognition software – much to the chagrin of civil liberty groups. 

Advertising

The French interior ministry unveiled over the summer a smartphone app called Alicem, allowing people with a biometric passport or electronic residence permit to identify themselves on the Internet through facial recognition technology. France would be the first European country to offer such a service, which will be ready in November, according to Bloomberg. But there are technological as well as ethical questions

Cédric O, the Macron government’s minister for digital affairs, announced a plan to “supervise and assess” the possible uses of facial recognition on October 14, while calling for more testing of the technology to allow manufacturers to make their products more reliable and efficient.

In the eyes of privacy advocates, Alicem is the first small step towards a Chinese-style system, where facial recognition is ubiquitous. Even the official watchdog, the National Commission on IT and Civil Liberties, has expressed strong reservations about this app, which raises questions about how biometric data will be collected and used.

Pressure groups have denounced what they see as a situation in which politicians are keen to use facial recognition technology without taking into account its potential pitfalls. For instance, Valérie Pécresse, the centre-right head of the Paris region, advocated in February the use of such technology to “fight against the terrorist threat”. Meanwhile police forces want to use algorithms to quickly identity people in huge databases such as the file of wanted people, Le Monde reports.

Technology ‘improved a lot over past two years’

But aside from ethical concerns the current craze for facial recognition may also be technologically premature. In 2017, the Metropolitan Police tested such software at London’s Notting Hill carnival. It was a resounding failure: the machine was wrong on 35 occasions – even resulting in the arrest of an innocent person.

Nevertheless, that was two years ago, and the capabilities of facial recognition software has “improved a lot in the meantime”, Josh Davis, an expert on the technology at the University of Greenwich in London, told FRANCE 24.

In France, experts argue that facial recognition should be at least 80 percent effective before it can be used: “We’re not there yet; we have to do more experiments,” one police officer told Le Monde.

However, even this figure has to be taken with a pinch of salt. “80 percent is not good enough,” Jean-Luc Dugelay, a special in image processing at the Eurocom engineering college near Paris, told FRANCE 24. He gave the example of airports, where tens of thousands of people would go through a facial recognition security system. If it were just 80 percent reliable, that would create thousands of cases of the technology either misidentifying innocent people or letting security threats go through.

Facial recognition technology only works well “under certain conditions”, Dugelay continued. That is to say that if the lighting, use of makeup and person’s position in relationship to the camera are similar to the image its used against, it is hard to trick the algorithms.

But it is more difficult for the technology to make the comparison between photos in a file and images of anyone who is distant from the camera in a poorly lit place or who has part of their face hidden. “As things stand, most facial recognition programmes are struggling when the photo used for identification is more than six years old,” Davis said.

Furthermore, facial recognition algorithms have to draw on a database of images in order to be effective. Millions of faces have to be fed into the machine for it be able to accurately distinguish facial features. The sample must also be representative of all skin colours and all different shapes of people’s faces and eyes.

With regard to the size of their databases, European countries “lag behind the USA and Asia, which have far bigger databases”. European law also places far greater restrictions on the use of such images than there are in China, for example.

Meanwhile France’s national gendarmerie pointed out in a 2016 report that there are dangerous implications for computer security when it comes to building biometric files.

“The protection of facial recognition systems against computer viruses is a major issue because, whereas PIN or credit card numbers stolen by fraudsters can easily be changed, one can imagine what kind of problems criminals can cause with stolen facial images,” the report said.

Cybercriminals could, for example, replace the face of a terrorist on a blacklist with that of an ordinary, innocent person.

This article was adapted from the original in French

 

https://ift.tt/2oLT5kl