The ghost of Deepfake arose: Detecting 14,678 celebrity face transplant videos, 96% of which were sexually suggestive

The development of technology has led to many of the consequences and the most dangerous of them, Deepfake – the technology that uses artificial intelligence (AI) to fake identity, in particular here is fake. Accordingly, this technology uses data on facial features, shapes that have been collected, then merged by artificial intelligence (AI). From there, Deepfake can assign one person’s face to another, with realism beyond imagination, leaving viewers unable to distinguish between real video and live video.

Worth mentioning, this software is built to show off the power of artificial intelligence, but is being used to attack others, and the victim can be anyone: From celebrities, politicians, senior officials … even ordinary people.

96% of deepfake videos on the internet are pornographic, the victims are all female artists

According to Forbes, creating videos with Deepfake technology is easier than ever, as the software used to create these fake videos is becoming widely available on source code-sharing websites like Github, or famous online forums like 4chan. Any user with a bit of technological knowledge and powerful machines can now download the tools and create their own deepfake videos.

The authenticity of Deepfake technology makes it difficult for viewers to distinguish true / fake

In 2017, the first deepfake videos appeared on a subcategory of social networking platform Reddit, with the content being celebrities swapping faces into high school movie stars. However, after 2 years, the number of videos with similar content is increasing rapidly. In fact, users can easily find a series of deepfake videos with sensitive content posted on social networking platforms.

Censorship of these videos is difficult, as they can easily reappear soon after being deleted by the social network admin team. Websites including Pornhub, Twitter and Reddit have banned AI-generated pornography, but videos with 18+ content can still be easily found just by searching on Google.

Gal Gadot is just one of hundreds of famous artists who are victims of Deepfake

According to Deeptrace, an Amsterdam-based company that discovers content created by artificial intelligence, 96% of deepfake videos have depraved content. Most of the victims are famous singers and actresses who have their faces taken into pornographic films.

The company also said there are currently 14,678 deepfake videos on the Internet, up 84% compared to 7,964 videos as of December 2018. To make statistics, Deeptrace uses an automated tool to monitor online forums and social networks, where users often exchange and discuss related content.

” We found that the top four pornographic websites that publish deepfake videos have a total of 134 million views. Hundreds of actresses, singers and celebrities worldwide are victims (” face transplant) “ Deeptrace CEO Giorgio Patrini said in a report. ” This huge audience indicates the existence of a market where websites create and host pornography videos using deepfake technology. This trend will continue to grow unless measures are taken to prevent it. block given . ”

Notably, Deeptrace also noted the proliferation of video creation services Deepfake, when those in need simply spend money to ‘order’ these videos.

Video creation tools using Deepfake technology are increasingly popular

The report from Deeptrace also emphasized that AI technology has been abused to target women by creating non-consensual porn videos. ” The impact of Deepfake has been felt globally ,” the company said. ” We hope this report will further stimulate discussions on this topic and emphasize the importance of developing a range of countermeasures to protect individuals and organizations from applications. harmful of Deepfake. ”

However, the risk from Deepfake does not stop at mere statistics. In an interview with CNBC, an associate professor of computer science at the University of Southern California, Hao Li warned that people would be able to witness Deepfake videos using ” true to perfection ” within from the next 6 to 12 months.

” It is still very easy for you to identify with your own deepfake videos today,” said Mr. Hao Li. ” But there are also really, really convincing examples. Soon, there will come a time when there is no way for us to really detect deepfake anymore, so we will have to find other solutions “.”

Reference Forbes