They deceived several European politicians through video calls with a deepfake

In recent days, various European politicians have been deceived by users who, using a deep fake, pose as a Russian opposition figure in video calls. The victims, Members of the European Parliament, thought they were talking to Leonid Volkov, an ally of Russian leader Navalny, but it was all a hoax.

Among the deceived politicians are Rihards Kols, head of the Latvian foreign affairs committee, Tom Tugendhat, chairman of the UK foreign affairs committee, as well as other Estonian and Lithuanian politicians. The Guardian.

Kols uploaded a photo of Leonid Volkov, a Navalny ally, and a screenshot taken from the video call. Volkov said the two looked virtually identical. And he wondered how they’d managed to put a face on the call via Zoom. He concluded with a forceful sentence: “Welcome to the era of deepfake”

The deep scythes which allow, thanks to the use of neural networks, to carry out assemblies where it is difficult to distinguish an original from an artificially created product. In recent years, this type of technology has advanced by leaps and bounds. There are even apps that allow any user, in just a few steps, to create their own content where they swap a face on another’s body and make the character sing or dance to their liking.

These versions are intended to create parodies but there are still more sophisticated tools based on the same deep learning technology that allows you to create increasingly realistic content. The risk of all of this is that they are used to spread disinformation or create hoaxes of the type mentioned above.

On March 16, politicians were contacted by someone posing as Volkov, who requested a meeting with the foreign commission to discuss the attempted assassination of Navalny and political prisoners in Russia. The meeting took place with Kols on March 23, although previously there would have been virtual meetings with Lithuanian politicians. But it was quite a trap. They thought they were talking to the Russian opponent but it was nothing more than a deep fake, as detailed Engadget.

Kols said someone claiming to be Volkov wrote him an email and asked him to make a video call to discuss support for Russian political prisoners and Russia’s annexation of Crimea. After a while, he realized he had been part of a hoax.

“It is clear that the so-called era of the decadence of truth or post-truth, and subsequent events can seriously threaten the security and stability of countries, governments and local and international societies,” stressed the politician, as published by the British news site.

Volkov accused a Russian duo called Vovan and Lexus of being behind this call. However Lexus, Alexei Stolyarov’s nickname, has denied using a filter to look like him.

European police had previously warned of the growth of this type of technology and the implications it could have in the future. The FBI recently did the same. The agency warned that the circulation of manipulated digital content will increase in the next 12 to 18 months and said that this type of deceptive material could be used to carry out various cybercrimes such as phishing campaigns, i.e. Identity theft: This involves theft of someone else’s identity to obtain confidential information or steal data.

A tool to detect deepfakes

There are different developments to help identify these types of hoaxes. One of the most recent was developed by researchers at the University of Buffalo. It is a system that automatically identifies fake photos by analyzing light reflections in the eyeballs of the users depicted and is, according to its creators, 94% effective.

“Both eyes must have very similar reflective patterns because they are seeing the same thing. This is something we normally don’t notice when we look at a face», Explained, in a publication published by the university, the main author of the research, Siwei Lyu, PhD, Professor of Innovation SUNY Empire in the Department of Engineering and Computer Science.

However, most AI-generated images, including images produced by antagonistic generative networks (GANs), fail to do so accurately, possibly because a multiplicity of combined photos are used to generate the wrong image.

It is not the first time Lyu focuses on such development. It also investigated and found that users blink less or inconsistently in deepfake videos, which may also be a clue to identify this type of content.

He also helped create the Deepfake-o-meter site, which analyzes videos and helps identify if a clip is real or a deep fake. It is not 100% effective, but one more tool to help evaluate digital content.

KEEP READING: