Facial Recognition May Not Work for Age Verification
OSAKA, JAPAN — Only days after a new facial-recognition method of age-verification was introduced on cigarette vending machines in Japan, a reporter demonstrated the technology can be fooled far too easily.The reporter for Sankei Sports was able to fool one machine’s facial-recognition camera by holding up a six-inch-wide magazine photo of a man who appeared to be in his 50s. He fooled another by presenting a three-inch-wide magazine photo of a 30-ish female celebrity. By continuing to downsize the photos, he discovered the magazine images were rejected when they reached the one-inch-wide size.
The machines were installed at street-corner vending locations in several Japanese cities in an attempt to prevent minors from buying cigarettes. According to manufacturer Fujitaka, the camera-equipped machines can tell the age of a buyer based on facial characteristics like number of wrinkles, bone structure and skin texture. In cases where the machines are not able to make a positive age-identification based on facial characteristics alone, they are programmed to require the user to scan his or her drivers license. Fujitaka claims the machines are accurate 90-percent of the time.
After the Japanese reporter’s report, Fujitaka said it would design a more advanced system that cannot be tricked into reading photographs instead of live, human faces.
Similar technology, to be mounted on laptop and desktop computers, has been proposed as an age-verification mechanism for adults-only online activities.