Surveillance cameras are everywhere today: outdoors and indoors, at airports, railroad stations, offices, and shops. You cannot escape the all-seeing eye of the cameras even in the wild, making George Orwell’s refined imagination blanch in the face of reality.
For most monitoring systems, video is recorded in cycles “just in case;” and does not go anywhere further. Lately it has become more frequent that this video is sent to various data analysis systems, therefore, it could be used to track some specific people.
It goes without saying that Big Brother may violate our private lives. You can accept that evil from the government, as these people are here as if to maintain order. Yet today, biometric systems attempt to use common businesses that act against our pockets and our right for private life. And this is a horse of a different color. For example, you are shopping for a winter coat. At the same time, the surveillance system of the store checks you with the pictures of known robbers and adds one more record into your buyer profile.
Alternatively, you go to a car dealer in order to take a look at new cars. Immediately upon your entrance, he finds out your name and all there is to know about you. Including the fact that you cannot afford a new car.
Isn’t it nice? Not so much, but there is nothing criminal about this. What will you say if every detail of your private life, collected by different companies, comes to light one day on the Internet? Unlike it has been with the story about hacking the Ashley Madison site, there will be no doubt that it is you — here are your relevant pictures and videos.
The laws of most countries still do not truly suppress the usage of facial recognition for commercial purposes, just as it is not forbidden, for example, to take pictures of people in the streets. It does not come as a surprise that more and more people are wondering how to hide from the all-seeing eye in these conditions. To understand how it can be done better, a couple of words should be told about contemporary image analysis methods. Under certain conditions, it is possible to highlight two common approaches.
May I see your face please
The first, approach is based on comparison of some markers that are assignable in a picture and a prebuilt database. The markers can be the distance between the eyes, the nose measurement results, lip shape evaluations, and such.
This approach is similar to identifying a person by his fingerprints. The sample fingerprints should be taken beforehand and saved into a database. After this, we compare to what extent the papillary lines of an unknown person match one of the samples. Therefore, the prerequisite for facial recognition is adequate-quality pictures (full-face and with good even lighting) of the right people. Where can these pictures be obtained? The sources can be different. Maybe we’ll be required to look into the camera when creating a discount card, or maybe someone will scan some documents with your picture.
It is relatively easy to cheat the classic facial recognition system. The easiest way is to lower your head and not look into the camera. Most markers can be measured only from a certain full-face angle so that the picture at an angle will not provide the required data, most likely. If you wear a peaked baseball cap upon your head, then the cameras located above (they are usually installed somewhere high enough) become totally useless.
Some experts advise to make faces when you pass by a lens. Maybe it works quite well, but it attracts too much attention. A pair of dark glasses is what you need. The advantage of sunglasses is in covering the eyes, which is one of the most useful areas on one’s face for recognition systems. Common transparent dioptric glasses do not distort the required details of a picture well, and advanced algorithms can cope with that quite easily. However, large opaque glasses are a serious challenge for the classic systems. So are mirror models that blind the camera with the reflected light.
And the way you look tonight…
The second approach to human recognition that is actively developed by, for example, Facebook and Google, works in a different way. It is based on machine learning algorithms and automatic sample data download and upload to compare it with all of the available online sources. This is a much more flexible thing that is much harder to trick. Even a gas-mask covering your face does not guarantee remaining incognito, as similar systems do not require strictly preset markers.
They can use any available data for recognition: the shape of your leg or your bald patch, your tattoo, your bearing, your clothes, etc. An experimental development by Facebook can identify a person from any angle with 83% accuracy with a sufficient number of sample pictures. The key point here is a sufficient number of pictures for the comparison. If there, at the other end, is only one image of you, even a high-quality one, then the possibility of successful recognition plummets. This is why Big Data technologies and fast Internet search algorithms are brought to the forefront.
Here comes the sore point: should we openly publish pictures of ourselves online? We can stand the fact that Facebook or Google uses them for their own marketing goals, as you cannot hide from these “Big Internet Brothers” anywhere at all. Yet, nothing stops any company from digging up the required data online if they are in free access.
Let’s assume that your Facebook page is closed with the “Friends Only” privacy setting. What about random pictures of you in different posts of other people? What about your profile on LinkedIn? It is very hard to cut off all of the sources even by complete social network abstinence. The security solution to this is yet unclear. Likely, it can be a more strict regulation of the biometrics market from the government side and a more conscious attitude from society.
So, it is time to get used to the thought that our personal pictures are almost the same private thing as our document or credit card scans. Flaunting and flashing them everywhere is absolutely not recommended.