According to the american institute of NIST standards and technology (www.nist.gov) and reported by the magazine TheVerge.com masks are a very good defense not only against the spread of microdroplets of Covid-19, but also against the systems of facial recognition.
Thus, the more a mask covers the face and the nose, the more it thwarts recognition algorithms. “The rate of error varies between 5% and 50% depending on the types of systems of facial recognition “.
The institute NIST focused on ” the manner in which an algorithm developed prior to the pandemic could be affected by subjects wearing facial masks. Later this summer, the institute plans to test the accuracy of algorithms that have been intentionally developed to decode the masked faces “.
As the masks of the real world are different, the team at NIST has offered variations, including differences of shape, colour and coverage of the nose.
In regards to color, the black masks déjoueraient easier algorithms than the equivalent blue.
And because they remove most of the visible features of the faces, the masks reduce the accuracy of the algorithms of facial recognition.
Speaking of another system of facial recognition that millions of owners use each day, the recognition system the Face ID of Apple is more effective because the iPhone phones use sensors of 3D depth for added security – provided you have of course installed the update iOS 13.5.
Not all the algorithms
If you are a little paranoid in terms of monitoring, do not shout victory too soon to travel incognito.
Note that these are not all the algorithms of face recognition which have been scrutinized by the NIST that focused only on the “model match one-to-one” – known as the one-to-one matching –, which is used at border crossings for passport control.
The monitoring systems of the masses through ” agreement one-to-many known (one-to-many) have not been tested by the NIST.
That said, the us agency adds that ” if the facial masks break systems one-to-one, they break probably the algorithms one-to-many with a frequency more or less equal “.
According to the magazine Quartz, the chinese SenseTime, considered to be one of the most important societies of artificial intelligence in the world, said at the beginning of the year that she had developed a facial recognition system that integrates thermal imaging cameras to detect people with an elevated body temperature (fever).
In spite of wearing a mask, this system is also able to identify persons with high accuracy.
The NIST brief
A word about NIST, this government agency is responsible for assessing the accuracy of these algorithms (as well as many other systems) to the federal government and the classification of the different suppliers is very influential. It is this same agency that has studied the dynamics of the collapse of the twin towers of the World Trade Center in New York.