Obscuring your face does not conceal you from facial recognition systems, scientists have actually found.
A group from the Max-Planck Institute found that blurred images were still separately identifiable with simply a couple of non-obscured images to train from. With the proliferation of images on socials media especially on Facebook, it is possible that practically anyone’s blurred face might still be recognized.
The researchers said just 10 fully-visible examples of an individual’s face were needed to determine a blurred image with 91.5 percent precision. With approximately just 1.25 tagged images, the system could still properly identify a private 56.8 percent of the time, which is 73 times higher than possibility would enable.
It fared worse in “throughout event” measurements – those in which the lighting, clothes, and context had actually altered – however still performed at 17 times the rate of opportunity, identifying people in 31.9 percent of obscured images.
The scientists also evaluated different obfuscation approaches to see if there was variation in the recognition performance. By trying to identify non-tagged Gaussian blurs, white boxes, and black box images they found that the best method for remaining anonymous is to post all your photos to Facebook with a black box over your face and shoulders. The next most safe would be blocking it out with a white box, then a Gaussian blur.