Humans Have a Natural Defense Against Facial Recognition: Aging

pcmag.com:

Facial recognition systems have become very adept at identifying people in a crowd, but it turns out they don't cope very well with us getting older.

As New Scientist reports(Opens in a new window), the algorithms used to associate stored facial data with an actual face don't deal well with wrinkles and natural feature changes. In fact, if no new images of a face are captured, it only takes five years of aging before they really start to struggle to identify individuals.

That's the conclusion of a team of researchers(Opens in a new window) led by PhD candidate Marcel Grimmer(Opens in a new window) working at the Norwegian University of Science and Technology. They created 50,000 AI-generated human faces and synthetically aged them to see how facial recognition systems would cope.

As commercial recognition systems don't divulge how their algorithms work, open-source facial recognition tools were used instead as the next best option. What the team discovered through testing was that the accuracy of the algorithms dropped as age increased, with five years being the point at which identification started to fail. 20 years or more of aging and they had very little chance of a positive result.

The age of the person also had a big impact on the accuracy of the facial recognition. Identification of individuals aged below 20 or over 60 had much higher failure rates because aging happens so much more quickly. As Grimmer explains, "Babies will change within two months, so you could probably capture a new photo every month and it would still fail. Generally, even until the age of 20 years, the face still kind of changes." With the over 60's, "the head shape changes again, and you have more pronounced wrinkles," so the failure rate