Facial recognition is used by police in several countries. Deemed intrusive and going against privacy laws, its use – sometimes abusive – is widely controversial. To denounce this practice, but also to fight against police violence, activists are developing their own recognition systems to identify police officers, who frequently perform their duties on condition of anonymity.
In early September, the Portland, Oregon, city council unanimously passed a law banning the use of facial recognition for city departments, but also for private businesses. A first in the United States, while clashes between demonstrators and the police have been rife in this city for several months.
A practice that is not unanimous
Facial recognition technologies are used today in many countries by law enforcement agencies, but their use is very controversial. Out of 25 Member States of the European Union, the NGO AlgorithmWatch identifies at least eleven countries whose police forces use this technology. Eight plan to introduce it in the coming years. Only two countries, Spain and Belgium, are not yet considering it. In August, however, the British justice ruled illegal this practice, considering it contrary to laws on the protection of privacy and not sufficiently supervised. At the same time, the petition initiated by the Ban-Facial-Recognition.EU movement has collected more than 18,000 signatures to date.
In June, two US senators, Ed Markey and Jeff Merkley, take a clear stand on the subject: they propose that facial recognition technology and other forms of biometric surveillance remain out of the hands of law enforcement. An ambitious bill, which aims to curb the over-dependence and abuse of law enforcement agencies on faulty surveillance algorithms. ” Facial recognition technology not only poses a serious threat to our privacy, it puts black Americans and other minority populations in our country at physical risk. ”Said Markey.
This summer, following the death of George Floyd, the city of Portland like so many others was the site of several weeks of tension and protests against police violence. The police chief, who is no longer in post today, then authorized his agents to conceal their name badges, hung on their uniforms, with adhesive tape on which their number was inscribed. An inadequate directive which, from the point of view of civil rights defenders, went against the responsibility of the police. Especially since many demonstrators reported acts of violence by police officers, whom they no longer had any means of identifying.
Faced with this new practice, a growing number of activists are developing facial recognition tools to identify police officers. For activists, reversing the roles in this way is also a way of denouncing the misuse of this technology, which they consider particularly invasive. One of them, Christopher Howell, a self-taught programmer, has thus developed a recognition tool to identify the Portland police officers who hide their names.
It was because he himself suffered a form of violence during the June protests that he began to think about this project. The development of such tools has become relatively straightforward, thanks to standard image recognition software made available in recent years. To create his algorithm, Howell used the TensorFlow platform, developed by Google. The real difficulty lies above all in the learning phase: you have to find enough good quality images from the local police to train the algorithm. News articles posted on the web and social media have been of great help here. Howell was able to collect thousands of images and his neural network is now able to recognize nearly 20% of the Portland police force.
The sprinkler watered
Its technology is not yet accessible to the public, Howell continues to work on it. But his program has already reportedly successfully identified a police officer with whom a friend of his had run into. With the city of Portland having only recently formally banned the use of facial recognition by public bodies and businesses, Howell questioned the legitimacy of its technology. Portland Mayor Ted Wheeler said his plan was “a little scary”, but a city lawyer said the law does not apply to individuals. He can therefore a priori continue to work on his project.
It should be noted that similar initiatives can be observed outside the United States: last year, in Hong Kong, during the demonstrations against the amendment of the extradition law, an activist named Colin Cheung also used photos of police officers extracted from the Web to build a system capable of identifying them. But once his effort was made public, he was arrested and then decided to abandon his project.
Earlier this month, Paolo Cirio, an Italian artist and activist, a staunch advocate for internet privacy, plastered hundreds of posters of police faces on the streets of Paris, which he had retrieved from the web . For his artistic performance, named “Capture”, Cirio created a database containing 4000 faces of French police officers to identify them with facial recognition technology. Following the reactions of the Minister of the Interior and the unions denouncing the endangering of the police, as well as the threats of legal proceedings, the artist preferred to remove the photos from his site.
The demonstrations against Belarusian President Alexander Lukashenko have also been – and still are – the scene of violent arrests. Andrew Maximov, founder of Promethean AI – an artificial intelligence-based technology used for the design of video games – of Belarusian origin and now based in Los Angeles, posted a video on YouTube explaining how facial recognition technology could be used to digitally remove balaclavas worn by police officers. This video has over a million views to date:
Here again, the software exploits the faces of the agents gleaned from social networks. For Maximov, the hour of revenge has come: “ For a while, everyone knew that adults could use [cette technologie] to identify and oppress the little ones, but we are now approaching the technological threshold where the little ones can [faire de même] to the big ones », He rejoices.