Mass surveillance has become a part of life. Whenever we go out in public, to travel, shop or just meet friends in a park, it’s almost impossible to avoid becoming a victim of mass surveillance. And one of the most common ways we are surveilled is through facial recognition surveillance. Once a passive, often unnoticed intrusion of our privacy, facial recognition surveillance is increasingly under focus from fundamental and digital rights groups and the general public.
While legislation against facial recognition surveillance has been slow in coming, action has been taken by rights groups, citizens, and even companies. There is currently a public campaign to force the EU a ban on the technology, for example. But one of the most interesting attempts to counter facial recognition is through the use of anti-facial recognition masks.
These masks are not a long-term solution to the threats of facial recognition surveillance. But their proliferation is yet another sign that people are tired of being spied on, and studies about the efficacy of these masks may offer some interesting insight into how facial recognition works, and how this form of AI has steadily developed.
What is an anti-facial recognition mask?
Anti-facial recognition masks are any accessory worn over part of the face to thwart facial recognition software. Designs often try to minimize the amount of the face that needs to be covered, and many designs don’t immediately say ‘mask’. For example, Polish designer Ewa Nowak’s design features three brass pieces that rest over the wearer’s cheeks and forehead – areas of the face that are often mapped and analyzed by facial recognition software.
Other efforts range from little more than face paint to large masks of clear, ribbed plastic that distort appearance. Scarves and other pieces of cloth are common anti-facial recognition masks. The fabric is often patterned in a way that is designed to fool facial recognition software. Some versions even cover the entire face save for the eyes. And some masks actually reproduce a human face, so the software will scan a ‘face’, but it won’t be your face.
While the form of anti-facial recognition masks varies greatly, the function is always the same: to change one’s appearance so much that the facial recognition software is unable to get an accurate image of the face, thereby preventing identification.
How does facial recognition work?
Facial recognition technology uses computer algorithms to identify distinctive details of a person’s face – say the rise of their cheeks, the shape of their jaw, or the distance between their eyes. These features are converted into mathematical representations that the computer can then compare against other faces in a database. Facial recognition surveillance is a form of biometric mass surveillance, which involves capturing certain physical characteristics – facial features, eyes, thumbprints, etc. – and analyzing them for the purpose of identifying you or making assumptions about you.
How precisely facial recognition works or catalogs faces could depend on its purpose. Governments use it as a purportedly cost-effective way to provide security, though there’s little hard evidence that this sort of mass surveillance makes us any safer. Companies are the biggest pushers of facial recognition surveillance. Tech companies develop it to make money, while other businesses use it to help them understand their customer base and sell us stuff. In such a scenario, making a positive identification of you might be less important than analyzing your features in order to make assumptions about you.
What is wrong with facial recognition?
The problems with facial recognition surveillance are manifold. There are questions about its efficacy and value. And there are significant problems with its impact on fundamental rights. This is especially alarming given how widely this form of AI is used. In the US alone, it is estimated that more than half of the adult population – some 117 million people – are in a facial recognition database.
The fundamental right most obviously violated by facial recognition surveillance is our right to privacy. The capture, analysis and storage of our facial data without our consent – or even knowledge – is a clear violation of privacy. It also exposes us to data theft if the facial imagery database is hacked. Or it can be sold on to third parties, again without our knowledge.
But there are many less obvious concerns. There is evidence that facial recognition technology exacerbates systemic racism and sexism in our societies. Studies have shown that facial recognition technology misidentifies people of color more often than white people. And it’s sexist – one study found that it misidentifies women 18% more often than men.
The threats to our rights go further still. The use of facial recognition surveillance is intimidating and makes people less likely to exercise their right to peaceful protest. It can even be used to track and police whole ethnic groups, as is done to Uighurs in China. This makes people afraid to participate in public life, to join in public debate or even to leave their homes.
But what if you don’t like to protest, and you don’t mind companies taking a picture of your face? Facial recognition surveillance is still a threat to you. That’s because companies can end up using it for all sorts of things. It could help determine whether you should get a job or a place in a study program. Or whether you can travel somewhere, or get a loan or a mortgage.
This affects our freedom to choose because the options that are available to us will be determined by assumptions that are totally unfair. This prevents us from taking our own directions in life, from having the freedom to live how we want. The options that are available to us in life should not be based on assumptions but on our experiences, preferences abilities and individual suitability.
How are anti-facial recognition masks thwarting surveillance systems?
Anti-facial recognition masks attempt to thwart AI by obscuring certain features of your face. This confuses facial recognition software and, in theory, makes you impossible to identify. Of course, walking around with your face entirely covered is neither practical nor stylish. So companies and individuals continue to test new ways to use face coverings to thwart facial recognition technology.As said, some masks attempt to fool the cameras using certain fabric patterns. These may appear to have eyes or other features of a human face. Or they might use “metalliferous fabrics” that attempt to frustrate facial recognition software. The general idea is not to completely hide oneself – we have to go out in public, after all, and we need to see where we’re going – but to change our appearance through anti-facial recognition masks so much that the software is unable to process the image or, at very least, unable to make a positive identification.
What is the success rate of anti-facial recognition masks?
It is difficult to say exactly how effective anti-facial recognition masks are. Facial recognition technology varies significantly both in terms of hardware and software. It’s also continuously improving, meaning the masks that worked yesterday might not work today. Moreover, there are myriad anti-facial recognition masks out there, so it goes without saying that some are more effective than others.
That said, there is evidence that anti-facial recognition masks can foil facial recognition software. One study tested 89 commercial facial recognition algorithms and found that they had error rates of between 5% and 50% in matching digitally applied face masks with photos of the same person without a mask. That study did attempt to account for the differences in masks and applied a number of different face covering options to test against the software.Other findings are less positive for anti-facial recognition masks. The United States Department of Homeland Security conducted a test of facial recognition software used in airports and found that it had a median success rate of roughly 77% when used against people wearing masks, while the best-performing system was up to 96% effective.
And the technology will only get better. Moreover, the increased number of people wearing masks because of the coronavirus pandemic has only made tech companies, corporations and the authorities more aware of the shortcomings of facial recognition software, meaning they are working as hard as ever to make sure that it is as successful as possible.
Will these be the accessories of the future?
Even if facial recognition masks become more successful in thwarting the AI, it would be bad for them to be seen as the solution. Facial recognition surveillance itself is the problem. Instead of focusing on how to hide from it, we should be working to get rid of it. Anti-facial recognition masks may seem like a neat solution, but they don’t actually address the root problem. And we cannot hope to keep pace with the technological developments in AI, meaning we’d always be playing catch-up.So working to get rid of biometric mass surveillance, and thus facial recognition surveillance, is the only way to actually solve the problem. To do this, we must continue to remind people of the everyday dangers posed by this surveillance, and how its use can cause real harm to even those people who support heightened surveillance powers and believe such surveillance causes no harm to them.
How long can artificial intelligence be fooled with anti-facial recognition masks?
It’s important to reiterate that attempting to fool artificial intelligence through the use of anti-facial recognition masks is a fool’s errand. Even if some masks are successful now, they almost certainly will not be in the future. The technology is often extremely good, and it’s only getting better. As said, seeing masks as a solution would be committing ourselves to a perpetual game of catch-up, and one in which we can never even be sure that we’re ahead of the curb at any given moment.
And who wants to go through life having to don a mask every time they go outside? That’s not the sort of society we want to live in. We want to be free to go where we want and meet whomever we want without the constant fear of being watched, of being profiled, and of worrying that our personal information could be shared or stolen. To do this, we must root out facial recognition and other forms of mass surveillance instead of becoming DIY mask makers in our spare time.