Facial recognition will not ensure public safety and here’s why…
Video surveillance with facial recognition won’t really benefit our security, but it could turn our daily lives into a Black Mirror-ish dystopia. These last few months, it has been increasingly used to secure access to popular events, airports or schools. But this technology is neither ethical nor reliable to ensure public safety, and here’s why …
(TW: mention of school shooting, police brutality, transphobia)
This year, Rio’s carnival was under massive surveillance, as never before. Security cameras posted everywhere had their eyes scrutinizing the streets, where Samba dancers were enjoying the most important cultural event of the year. It was about spotting people with arrest warrants, in this joyful crowd. The cameras were supposed to identify the suspects, combining facial recognition with the Police Department’s database.
At the end of the week, out of the 25 000 people that were arrested, only 2 of them were actually charged and prosecuted.
This is one concrete example of how dangerous facial recognition can be when it comes to urban safety, if not used properly.
How does facial recognition work?
Facial recognition (as a tool for monitoring public space) is operated by an artificial intelligence. Through the city’s cameras network, it detects human faces, analyses various spots on it, and through calculating the space between the dots and their position, it builds a mathematical formula of a face, called facial signature. The AI then compares it with a broad database of pictures. If your faceprint and your private data (such as your name, birth date, address etc.) are in the database, it quickly identifies you.
How do you get a database of faces? How is it built? For now, it is mainly made of police files. Soon, it might be our passports photos, but there’s more. Every day, we willingly post petabytes of selfies on social medias : whoever needs to train an AI to facial recognition might just go there and feast.
What are the underlying political motives behind it ?
How do leaders of big cities justify the use of this technology? The main goal, they claim, is to reassure the public : knowing that cameras are everywhere should make you feel less at risk to be assaulted. Another use presented to us would be to find lost children or missing elderly people. If a grandpa disappears (let’s say voluntarily escapes, who could blame him) from his retirement home, his relatives would just have to send a picture of him to the surveillance service. The old man would probably be found safe and sound, instead of wandering for days, amnesic, in the streets. This is the best case scenario.
Except, these examples are a negligible percentage of the actual uses.
The interest for the governments are of course elsewhere, and that’s where we have to look.
In a demonstration video, the police of Nice, France, proceed to arrest a wanted woman, following her location and identification via cameras. Therefore, the goal is very clear : facial recognition is a tool to identify persons of interest and to arrest them.
It’s easy to understand why mayors of big cities are tempted to use biometrics control systems, for example to regulate accesses to public transportation. No need for a pass anymore! This kind of control system is already operating in many airports, such as Paris’, Helsinki’s, Hong Kong’s… and coming soon in 20 US airports, for all international flights. The passengers’ faces are scanned to verify if they match their travel documents informations, and it’s also about making the control lines more fluid.
Beyond that, politicians would also like to prevent people identified as threats from boarding planes, trains and subways. Under what judicial supervision? Who should be banned from transit? In what circumstances? For how long?
Sounds more like a campaign argument to win votes over security anxiety.
1.4 billion of Chinese citizens under surveillance, what about us?
Let’s take a closer look at China, the leader in the market of facial recognition technologies. It is also the country that uses it the most, in every aspects of life, whether it is taking cash at the ATM, checking in at the hospital or paying your order at the fast food : everything is done with facial recognition systems.
Detecting faces means also detecting emotions: if you look worried, angry or nervous, the machine will spot it, and think maybe, you’re up to no good…
If you are yawning during classes, it means that you are bored : it’s time for your teacher to improve their performance or they’ll soon be fired.
Beyond the commercial uses in the private sector, it’s above all the Chinese government that has taken over facial recognition.
The ultimate goal is to develop a large surveillance network, in order to keep an eye on 1.4 billion citizens.
The police are also equipped with glasses providing the same service : real time identification of suspects. If you leave in one of these Chinese cities under surveillance, each one of your moves, actions and emotions could be monitored.
In a strategy of public name and shame, the “troublemakers” (not people who snatch old ladies’ purses but who simply cross the street when the light is red) see their faces posted on giant screens, with their names and their addresses.
Therefore, the dystopian nightmare really begins when we link private data and facial recognition.
Facial recognition can lead to discrimination and serious abuses
The main risk of using this technology is obviously a restriction of our personal freedoms. But other abuses are likely to occur, such as false positives: you can be tracked down and arrested by mistake, which is a pretty traumatic experience.
A good example to understand the limits of this technology is its use in high schools. In the State of New-York, the Lockport school district was one of the first to use facial recognition, in order to prevent unauthorized intrusions. If your face is not in the students and staff database : red light.
It’s not much about catching intruders than about preventing school shootings, the absolute horror that occured way too often, again and again, in the past twenty years.
Of course, our priority is our children’s lives. Of course we should do whatever it takes to prevent school violence. But facial recognition is not the tool for that.
This technology has proven its weaknesses and flaws, being unable to tell one face from another, having a ridiculously high rate of false positives etc. In the same time, it has revealed algorithmic racist biases (and questions the training of AI) : for example, the software were unable to tell the difference between two black students. Joy Buolamwini, studying at the MIT, discovered that a facial detection software was unable to map her face until she wore a white mask.
Similarly, the American Civil Liberties Union has laid out a study demonstrating the flaws of Amazon’s facial recognition software, Rekognition. It incorrectly identified 28 members of the US Congress as criminals already convicted. And guess what? The 28 members mentioned were black people.
Algorithms can reproduce the worst discriminatory biases, if AI are not properly trained with a database of as much diverse faces as there are in humanity.
Facial recognition can also be dangerous for transgender people. The software can mistake a person for another, because of non-conforming gender features. It could trigger an alert that could lead to abusive arrest and police brutality, while the transgender community is already extremely vulnerable and at a higher risk of violence and assaults in general.
The London police has been sued eight times between 2016 and 2018, for abusive arrests due to facial recognition software wrongly matching innocent people with photos of criminals in the database, for a final 96% rate of false positives.
As for securing high schools, let us not forget that the perpetrators of mass murder were students, so they were allowed to enter.
Unfortunately, facial recognition surveillance will not prevent the criminal actions. But, it stays a profitable market for the companies who sells these services : $2.7 billion are spent every year on security in US schools.
A European regulation is necessary
Given the ethical issues we mentioned, facial recognition must be limited to hunting highly dangerous profiles, when they are a real threat for the public safety.
It is certainly not up to us to tell China or the US what to do or not to do, and to force our ethics into their practices. We are not entitled to impose our moral criterias.
But it is urgent to demand a specific regulation in Europe. We must never allow, for example, that facial recognition software matches our faces with our private data.
It is about preserving our fundamental freedoms.