you are here: HomeNewsTechnology

Deepfake bots create fake nudes of women, aid public shaming and extortion

Telegram users have been asking bots to create fake nudes and sharing the manipulated pictures, an investigation by a cybersecurity firm has found. Their victims could include children.

November 07, 2020 / 08:33 AM IST

The users of messaging app Telegram used bots to create fake nudes of women by manipulating the pictures they posted on various social media platforms and circuclated these photographs, an investigation by a cybersecurity firm has found.

More than 100,000 deepfake images, which undressed women by using artificial intelligence, were generated on request by bots and shared on Telegram, researchers at Sensity, which describes itself as a "visual AI company", found. More than 70 percent of these came from Russia and its neighbouring countries. Victims could also include minors, said the report, a copy of which is with Moneycontrol.

The pictures were shared for revenge, extortion, or public shaming, Sensity said, highlighting the challenges the technology poses to safety, privacy and personal information.

Derived from the words “deep learning” and fake, deepfakes use artificial intelligence to manipulate photos, videos and audios that appear to be real.

Deep learning is a field of artificial intelligence where machines learn “by experience”, which means improving performance by going through a large amount of data, and performing tasks that a human brain does.

Close

The technology can manipulate images to make them look real. Deepfake photos and videos have aided the spread of misinformation and were thrust in the limelight during the bitter campaign for the US presidential election.

There have also been worrying reports of deepfake journalists driving online propaganda and spreading misinformation.

Bots doing the dirty work 

Social networking sites like Facebook, TikTok and Twitter have banned the use of deepfakes but not Telegram.

The deepfake bot is easy to use. Users upload a photo to the bot and receive a processed image within minutes.

The bots gained traction in the last three months, increasing by 198% the number of processed images, Sensity said referring to the May-July period.

DeepNude library was earlier restricted to a website (DeepNude.to). From April, the algorithms were integrated in Telegram, giving its users anonymous access.

By paying a “service fee” of 100 rubles (about 95 rupees), a user can get the bot to remove the watermark and skip the queue, says the report. The bot is operational through Telegram, where messages and chats are encrypted and almost impossible to track.

Seven Telegram channels for the AI deepfake bot had more than 1 lakh members, the report said. The primary channel had more than 45,000 unique members. A poll conducted in one of the channels revealed that two-thirds of the users abused the tool to create deepfakes of girls they knew in real life.

The bots use a technology called DeepNude. The first batch of pornographic deepfakes landed on the internet in 2017 as fake celebrity videos.

Similar apps started popping up on the internet that assisted users. The majority of victims of DeepNude are women. And in an interview with Motherboard's Samantha Cole, the creator said he did it for "fun and enthusiasm" and wanted an "economic return from the algorithm". The creator, who goes by the alias Alberto, remains anonymous and law enforcement hasn’t been able to track him.

US House Speaker Nancy Pelosi was one of the first high-profile victims of deepfake when her doctored video, which showed her impaired, was widely shared on Facebook. The motive was to embarrass and discredit her.

Similarly, Facebook CEO Mark Zuckerberg's deepfake video was also created and stayed on the platform for a very long time. It was in January 2020 that the social media giant finally banned deepfakes.

For most people, it's impossible to tell a deepfake from real. Individuals, companies and governments are worried but don’t know how to deal with it.

Sensity shared its findings with Telegram but didn’t get a response.  Media reports, however, have said the Italian Data Protection Authority had opened an investigation into Sensity's findings.

Telegram is known for its encrypted messages and makes it easy for a user to create an anonymous account that is almost impossible to track. While these safety features are meant to help people evade surveillance and protect privacy but the app has often been misused.

Privacy concerns have mounted with the advances in technology. Deepfakes are an early sign that law enforcement needs to come to terms with this new reality. Criminals are now equipped with state-of-the-art encryption and AI and they can’t afford to just play catch-up.

(The author writes on technology, aviation, and mobility.)
Shivam Vahia
first published: Nov 7, 2020 08:33 am
Sections
ISO 27001 - BSI Assurance Mark