Cybersecurity experts over at Sensity AI have shared their findings on a disturbing AI-based bot that uses popular messaging app Telegram as a platform. The new “deepfake ecosystem” allows users to “strip naked” clothed images of women; the service is also free, which has resulted in the targeting of around 104,852 women as of the end of July 2020.
According to the report, most of the affected women are “private individuals”, and the images were taken from their private social media accounts. What’s even more worrying is that a “limited number” of images being shared across these channels are of underage women.
“A limited number of images also appeared to feature underage targets, suggesting that some users were primarily using the bot to generate and share pedophilic content.”
What is a deepfake?
Deepfakes are the result of AI-based human image synthesis. Basically, the technology works by superimposing existing images or videos onto source images/videos via machine learning techniques. In a nutshell, deepfakes are AI-generated videos that can be manipulated to look very real.
According to Sensity, the Telegram bot in question probably uses an open-source version of DeepNude software—a paid service that “undressed” images of women for a price. DeepNude has been taken down since, although it appears that we’re seeing something similar here.
However, the deepfake bot on Telegram is free to use. Users can still pay for premium benefits, such as removing watermarks and skipping the user queue, but the basic service of “stripping down” images is free. These offending images are then shared across the deepfake bot’s ecosytem of seven affiliated Telegram channels—which have a total of 103,585 member (as of end July 2020).
Of these users, 70 percent are from Russia and other former USSR countries, with the U.S., UK, Canada, Australia, and other English speaking countries making up three percent. Spain and Latin America users, meanwhile, make up six percent of the user base. Despite Telegram’s ban in several of these countries, the bot’s creator reportedly encourages the use of a VPN service to bypass the restrictions.
Telegram isn’t the only platform that the bot operates on. VK, a Russian social media platform, reportedly hosts links to the bot and several affiliated channels, and users have been spotted posting adverts for the service on the platform.
Perhaps most worrying is the fact that the vast majority of users, according to a poll on the bot’s central hub Telegram channel, are using the service for images of “familiar girls, whom I know in real life”. Deepfakes have traditionally been used to doctor/fabricate videos of known celebrities and public figures—but here, private individuals are being targeted.
In general, the danger of deepfake videos are that they can be used for public shaming and extortion. According to Sensity, 96 percent of deepfake videos found online have pornographic elements to them, with high profile celebrities often targeted. At the moment however, it is “unclear” how the bot-generated deepfakes on Telegram are being used.
Telegram, for their part, has yet to issue a direct response to the findings. Suffice it to say, you should always be careful with your photos and videos online. If you’re an avid social media user (and poster), you can set your accounts on private to ensure that only known individuals can see your photos.
For the full report, click here.
[ SOURCE ]