Does facial recognition on social media lead to privacy violations?

In recent years, large IT companies have been actively engaged in technologies that allow users to recognize faces. Face recognition algorithms are designed to simplify and speed up authentication, as well as make the use of popular services and sites truly convenient. However, many users fear that the introduction of new technologies will lead to a violation of privacy rights, especially in cases where facial recognition is used on social networks and on popular sites. Should you be afraid of new technologies? Let's try to understand this issue. But first, let's turn to history ...


How it all began? Face recognition in the twentieth century


Woodrow Bledsoe, an American scientist in the field of artificial intelligence, made significant contributions to the theory of pattern recognition and automatic proof


Despite the fact that the media became interested in face recognition technologies relatively recently, many algorithms that allow you to find a human face in an image and determine who it belongs to were developed in the last century. In the mid-1960s, Woodrow Bledsoe, commissioned by the US government, developed one of the first comprehensive face recognition solutions for photographs.



The hardware and software complex consisted of a tablet from RAND and a program that reads the coordinates of faces in images. The process of the program was not fully automatic: the user manually indicated where the lips, eyes, nose, etc. are in the photograph, the result of the work was recorded in a special database. After that, the program could find people in photographs whose faces had been analyzed earlier. New images were transferred to the program by digitizing them using drum scanners (the first such scanner was developed back in 1957). The recognition accuracy was low - even with a slight change in head tilt or facial expression in the image, the algorithm gave incorrect results.


In the 70s and 80s, the algorithms for calculating the geometry of faces became more accurate. In particular, a method has been developed that makes it possible to accurately recognize faces by certain control points. However, the breakpoints, as in Woodrow Bledsoe's design, were still manually set by the user.


In 1991, American scientists Matthew Turk and Alex Pentland developed mathematical models, the use of which theoretically made it possible to automatically recognize faces in digitized photographs. The practical application of these algorithms became possible already in the 21st century - in the early 90s, the power of computers for such complex calculations was not enough.


In the second half of the 90s, large companies first thought about commercial applications of face recognition. The result of the efforts of sponsors and researchers was the FERET database, which contained photographs of several hundred people. These images were used to test photo comparison algorithms. By the late 1990s, based on this data, scientists had achieved relatively high accuracy in face recognition for the first time.


New millennium. From intelligence agencies and police to social media

Of course, law enforcement agencies are interested in such achievements. In 2002, US police used algorithms developed by researchers to analyze footage of the US National Football League final. With the help of modern technology, people who were previously wanted for serious crimes were discovered in the stands. However, later it turned out that the face recognition algorithm often made mistakes due to the simultaneous presence of hundreds of people in the frame, so the results of the work were additionally checked by police officers.


In 2006, new algorithms were developed that calculate a three-dimensional model of a face from a flat image. The use of new mathematical models made it possible to reduce the likelihood of erroneous results by 10 times compared to the results of the above-mentioned analysis of video recordings in 2002. And soon the testing of neural network face recognition algorithms began. Now Internet companies are finally able to fully use face recognition algorithms for commercial purposes. Therefore, the creators of social networks and search engines became interested in the developments of scientists.


Era of Facebook

While researchers and law enforcement were involved in the development, the likelihood of an unauthorized invasion of privacy was practically zero, since facial recognition algorithms were used either to ensure public safety or in the interests of science. In these cases, the research results usually remained in the archives of police departments and universities. However, in the late 2000s, the situation began to change. And since 2010, Facebook has begun to implement facial recognition algorithms into its social network.


According to American IT specialist Ben Parr, Facebook killed the very concept of privacy (note that these words were published in 2011). According to the specialist, the very ability to share almost any information already creates significant problems, in particular, due to the fact that the rules of social networks prescribe the use of real usernames. Also, the technology of face recognition was introduced by Google (it is used in image search), which means that many users can "share" their personal data with anyone, just by getting into the camera lens.


No more privacy?

This brings us to the answer to the question in the title.


The use of software, search engines and social media with facial recognition algorithms can indeed lead to privacy breaches. Moreover, it is often not the software developers who are to blame for such violations, but the users themselves.


Many users of social networks often publish personal data that allows them to determine their place of residence, social circle, etc. Of course, this information may be of interest not only to friends and acquaintances of users, but also to attackers (from cybercriminals to robbers). At the same time, face recognition algorithms become "smarter". In December 2017, Facebook representatives announced a new function of the social network. Now Facebook will be able to guess that the user is captured in the photo, and send him an alert, as well as offer to tag himself in the photo. The function can be turned off, as well as inform the social network that the photo shows another person. Company representatives assure that the new features will not lead to serious privacy problems. However, other users may by mistake or deliberately provoke a leak of confidential data: the human factor has not been canceled.


However, the likelihood of data leakage does not exist only when using Facebook. Back in 2016, photographer Yegor Tsvetkov tested an application that indexes images from VKontakte and finds user pages from photos. The photographer quietly filmed strangers in the Moscow metro and found their profiles on the social network. Tsvetkov's work consisted of two adjacent photographs: photographs in the subway and in the social network. Thus, the photographer gained access to the personal data of random people, while without violating the law (let's leave the ethical side of the issue aside).


How to prevent data leaks?

Immediately, we note: do not fall into paranoia and delete pages on social networks. Firstly, social networks are a fairly convenient means of communication, and secondly, it is almost impossible to completely prevent the appearance of information about yourself on the Internet. However, there are a number of steps you can take to make automatic face recognition on social media harmless.


Forbid tagging yourself in photos. Most social networks allow you to do this: for example, Facebook has a section for chronicle and mark settings, in Odnoklassniki - "Publicity settings", etc. True, VKontakte does not have such an option, but users who mark you in photos can be added to the black list. 

Check your privacy settings carefully when adding photos to albums. Let your personal photos be available only to close friends and family. 

Publish on social networks and on popular sites only the data that you could communicate to any unfamiliar person without consequences - for example, name, gender, personal website address, etc. In this case, even if someone comes to your page using neural networks, they will find only trivial information. 

Make your profile photo "unrecognizable" for automated face recognition. Head tilt, hood, sunglasses, image manipulation with artistic filters - all these lead to automatic recognition errors. At the same time, people will be able to recognize you without any problems.

In general, we note that face recognition is a rather useful function if it is used deliberately, in compliance with basic safety rules. However, it should be remembered that while recognition algorithms can be wrong: a social network can confuse you with a stranger, and Windows Hello authentication can be passed  using a specially prepared photograph. Therefore, it is still desirable to protect critical information, for example, using two-factor authentication (password + code from SMS).

No comments