New technology protects your online photos from face recognition algorithms

New technology protects your online photos from face recognition algorithms

7. Juli 2020 0 Von Horst Buchwald

New technology protects your online photos from face recognition algorithms

Singapore, July 7, 2020

The human eye can only scan a few photos in a second. Computers, on the other hand, are capable of performing billions of calculations in the same time. With the explosion of social media, images have become the new social currency on the Internet.

Today, Facebook and Instagram can automatically tag a user in photos, while Google Photos can group its photos using Google’s proprietary image recognition technology to identify the people in those photos.

So dealing with threats to digital privacy today extends not only to preventing people from seeing the photos, but also to preventing machines from harvesting personal information from the images. The boundaries of privacy protection must now be extended to machines.

Under the leadership of Professor Mohan Kankanhalli, Dean of the School of Computing at the National University of Singapore (NUS), the research team at the Faculty of Computer Science has developed a technique that protects sensitive information in photographs by making subtle changes that are barely perceptible to humans but make selected features unrecognizable by known algorithms.

The visual distortion with currently available technologies ruins photos because the image has to be changed a lot to fool the machines. To overcome this limitation, the research team developed a „map of human sensitivity“ that quantifies how people react to visual distortion in different parts of an image across a variety of scenes.

The development process began with a study involving 234 people and a set of 860 images. Participants were shown two copies of the same image, and they had to pick the one that was visually distorted. After analyzing the results, the research team found that human sensitivity is influenced by several factors. These factors included things like lighting, texture, object feel and semantics.

Applying visual distortion with minimal disturbance

Using this „map of human sensitivity“, the team refined its technique to apply visual distortion with minimal disturbance to the image aesthetics.

It took the NUS team six months of research to develop this novel technique.

„Our solution allows the best of both worlds, as users can still safely put their photos online before the prying eyes of an algorithm,“ explains Prof. Kankanhalli.

End users can use this technology to mask important attributes on their photos before they put them online, and there is also the possibility that social media platforms will integrate this as standard in their systems. This will introduce an additional level of privacy protection.

The team also plans to extend this technology to video, another prominent type of media often found on social media platforms.