They develop AI that allows maintaining anonymity in group photos on social networks

A group of researchers has developed a new technique for deepfake which serves to improve privacy in the social networks and maintain anonymity in photographs in which several people appear published on these platforms.

Deepfake is a video technique that uses Artificial Intelligence (AI) to create, alter and falsify images, which in recent years has been used mainly by cybercriminals to carry out their fraud.

LOOK: The shocking images created by two AIs with the futuristic version of Lima in the year 2050

So much so that the FBI announced at the end of last July that there had been an increase in profiles of this type, in which impostors used videos, images, recordings and stolen identities, posing as someone else to obtain a position. of remote work.

Due to the increase in the misuse of this technology, a team of researchers from Binghamton University (United States) and Intel Labs, who have focused on a beneficial use of this masking mechanism to improve privacy in social networks.

As pointed out in the research document, users of these platforms can publish images of people who have not authorized their appearance on them. In addition, they have the ability to record and store these identities regardless of whether or not their owners have tagged those who appear in them.

SEE ALSO: Google works on an AI to generate creative texts and “explore the future of writing”

To promote greater privacy in social networks and that these users have the opportunity to maintain their anonymity, the researchers have based themselves on the motto ‘My Face My Choice’ (My face, my choice) in which their faces are replaced by others. different enough to be totally unrecognizable in these images.

To do this, they have proposed an anonymization system that improves the privacy of those who appear in the image and uses the masking or deepfake technique where it appears to be a group of several people.

To carry out this forgery, the system collects the information from the photograph and modifies the features of the people (eyes, nose or mouth)proposing traits belonging to users of the same gender and approximate age.

LOOK: Shutterstock will offer AI-generated images and pay the artists who created them

In this way, the people in the original image lose their original identity and acquire a new one while maintaining a similar appearance. In addition, the body pose and facial expression of that person are preserved in the images.

On the other hand, the research indicates that these modifications are applied depending on who is viewing the image. In this way, whoever publishes it will always see the original photograph, but it will not be like that in the rest of the cases.

The rest of those who appear in it will see the original faces of the people on their friends list. Those with whom you do not have any type of contact, but that do appear in the photo, will be shown modified through the deepfake technique.

SEE ALSO: It’s not just art anymore, now there’s also AI-generated selfies with DrawAnyone

As a consequence of the application of this technique, users will believe that they are seeing the original photograph and this, in turn, will present new details that are far from the features of each of the people that appear in it.

The researchers have commented in the conclusions section of the study, in which several examples of this technique are presented, that “as a prototype system there is always room for improvement” and that the result of the deepfake depends largely on aspects such as the resolution or the orientation of the faces.

Source: Elcomercio

James

I have worked as a journalist for over 10 years and have written for various news outlets. I currently work as an author at 24 News Recorder, mostly covering entertainment news. I have a keen interest in the industry and enjoy writing about the latest news and gossip. I am also a member of the National Association of Journalists.