Facial recognition software is one of the most hotly debated technological advancements of our time. The possibilities of misuse brought many companies to put the development and use of facial recognition on hold. But new start-ups, and the enormous demand of public and private security organizations have made facial recognition a spooky reality. There are many cities around the world that employ facial recognition in public spaces to spy on their citizens, with the pretext of keeping them safe. Of course there are examples where facial recognition succesfully detected criminals. But people’s privacy is massively and inadvertently compromised through analyzing the movements of identifiable individuals. This blog will try to give a concise overview of possiblities of evading facial recognition. I will first focus on techniques in the digital realm. These techniques relate to facial recognition in public spaces in the way that you can avoid your face entering facial recognition database. Then I will shed light on a few techniques of how to conceal your identity and fool facial recognition in the analogue world. I will only give a small introduction to each technique and provide links if you are interested in specific ones.
A new facial recognition software called Cleraview AI is being used by many US law enforcement agencies. Equivalent softwares are used by law enforcement but also by corporations, allegdly for security purposes all around the globe. Clearview scrapes thousands of websites to build a face database. Here are some ideas how to avoid being identified online and to avoid your “real” face entering a database.
Disguise your face or never upload Photos
Methods for disguising you face will be discussed below in the analogue methods part. Never uploading any photos is possible but not an option for many people.
There are various online tools that allow you to add noise to your pictures, which makes it more difficult for facial recognition software to detect your face.
Image cloaking with software
University of Chicago invented a software (it already exists but is not publicly accessible due to security concerns) which places a digital mask over pictures of your face that you upload. For the eye, the images look the same, but key facial biometric points have been edited.
There is also a free to use software which allows you to edit videos, and put the face of another person on top of yours.
Change Settings in, for example Facebook
Not all Facebook users seem to have access to the feature of (de-)activating automatic facial recognition. Settings-> Edit->facial recognition-> select No under “Do you want Facebook to be able to recognize you in photos and videos”-> close. If you do not have this function available you can file a complaint to the Facebook support team.
In public spaces facial recognition is used to identify people as they are passing by or spending time outside. Cameras in public spaces mean that you cannot flee from them, only if you never went outside again. This seems obvious, but is a meaningful insight, as the goal has to be not to avoid the cameras but to fool them. Here are some techniques to fool facial recognition and keep parts of your privacy.
Make-up and Hair
Contrastingly to how make-up is usually applied in western societies, lightening the eye sockets and to darkening nose, eyebrows and cheekbones can fool facial recognition. Assymmetry through make-up may also confuse the software, as it is based on the (close to) symmetry of human faces. You should hide any unique features of your body like tattoos, moles or scars. Facial recognition focuses a lot on lips, so you can use lip colored make-up to add volume to your lips, or skin colored make-up to conceal parts of you lips. You can style your hair in way that it covers part of your face, also leading to asymmetry. Check out some styles here.
Clothes and an Image
Obviously you can wear clothes that obfuscate your face, like a baseball cap or a hoodie. Covering an eye is extremely effective but inconvenient. In addition, it is possible to buy reflecting clothes, which disturb cameras as they starkly reflect any source of light. An interesting finding was made by students at the University Leuven in Belgium how an image can obfuscate the YOLOv2 software which detects human bodies, rather than only faces. In this video you can see how holding a simple picture of people holding umbrellas can fool the software. The “Adversrial T-shirt” was designed by students of MIT to evade person detectors in the physical world.
Accessorizes and Gadgets
Sunglasses can be used to cover you eyes and especially helpful are those that refelct infrared light, which create a glare around your face region. There are also goggles with infrared lights attached to them. This light is not visible to humans but can completely blind a camera. Infra red lights can actually be added to any accessorize or clothing. Another interesting invention is the HyperFace, a cloth which minimizes the difference between your face and the surrounding.
In places where its legal, you can wear actual masks. Here are some examples of masks that are less intrusive and lie in a legal Grey area, as they do not actually cover the face. Designer and inventor Jing Cai-Lui developed a projection mask.
As you can see there are many ways to escape surveillance by facial recognition. This blog does not aim at providing an exhaustive list and some techniques were left out. At this point I also want to stress that the context of all these techniques matters substantially. You should always consider in what situations you use what technique as results may vary. My final remark is that I openly want to caution of the growing use of surveillance by facial recognition. It may lead to a perverted society of paranoia as the message of cameras everywhere is: bad people are among us, always be weary. The practical use of facial recognition in public spaces is debatable. Advocating privacy concerns and any other preventive measures are the last techniques I will mention that counter a future of surveillance by facial recognition.