Russian trolls outing porn stars and prostitutes with neural network facial recognition app

A new app that uses a neural network combined with facial recognition software to help put names to faces in random photographs by scanning social network data is being used in Russia to identify and harass young women who have previously appeared in pornographic films.

Trinity Digital, an app developer in Russia, released a free iOS and Android app called FindFace in February that enables people to identify people by taking a random photograph and using its neural network to figure out the person's name, location, occupation and other details.

A neural network is a huge network of computers that are trained using computer algorithms to solve complex problems quickly, as well as improving artificial intelligence by gaining a deeper recognition and understanding of art and the world around us.

On 9 April, members of a disreputable imageboard called Dvach, which is like the Russian cousin of 4chan, launched a campaign to deliberately try to locate and identify actresses who appear in pornography, as well as women listed on Intimcity, a Russian website advertising prostitution and escort services.

Using FindFace, the Dvach users not only identified the women, but they also shared archived copies of their profiles on Vkontakte (the Russian version of Facebook) publicly online and repeatedly spammed the families of the women letting them know that they had been outed as porn stars and prostitutes.

According to Global Voices, the users claimed that they were conducting the doxing campaign against the women due to moral outrage, but at the same time, the Dvach users also complained that the type of women who are porn stars and prostitutes usually ignore the men that make up Dvach's audience.

Neural networks being used for bad

Trinity Digital says that the app is meant to facilitate making more friends, but FindFace has gained popularity with the Russian public primarily due to coverage showing how fast and effective the technology can be at allowing any member of the public to locate another person.

In March, software engineer Andrei Mima wrote a post on Vkontakte (the Russian version of Facebook) about FindFace. He said that six years previously, two women had asked him to take a photograph of them in St Petersburg. However, Mima forgot to get their contact information, so he was never able to send them the photograph.

He uploaded the photo to the FindFace app and was able to instantly identify the women within one second and contact them, and he was incredibly impressed that the algorithms worked so well. His post quickly went viral.

After reading Mima's post, a local Moscow artist called Egor Tsvetkov decided to carry out an experiment, photographing random travellers on the St Petersburg subway. He then used FindFace to instantly identify the people in order to point out just how invasive the technology could be when used in the wrong hands.

Unfortunately, Russian media reports about Tsvetkov's work have only served to further inflame negative interest in the app. Looking at the official FindFace app website, it seems as if the app developers are fine with this, given the imagery used and the app's tagline, which reads: "This will find any Vkontakte member! Innovative people search service for photos."

"We are making every effort to protect all Vkontakte users from potential malicious acts," FindFace founder Maxim Perlin told Russian news site TJ Journal, pointing out that distributing pornography in Russia is a felony.

"And we're prepared, if necessary, to provide any information needed to find the users responsible for this harassment."

Neural networks, while still in their infancy, are capable of great things due to the sheer amount of amassed computing power, and researchers have only tapped the surface of what they can do.