Google Patents Facial Recognition For Android Security
Imagine whipping out your Android phone and getting past the lock screen by making a funny face rather than connecting a series of dots with your finger. Facial recognition is the key to making that happen, and it appears Google is on it, as evidenced by the fact that it filed a patent for detecting facial gestures when determining whether or not to deny authentication to a computing device.
It's true that a similar technology already exists in Android 4.1/4.2 Jelly Bean, which has users blink to unlock the device. The problem there is that it's relatively easy to thwart by taking an image of the device's owner, painting over the eyes in Photoshop with the same color as their surrounding skin, and then flashing the two photos to simulate a blink (we're not condoning this behavior, but it's a well known hack and one of the reasons why Google is still working to improve facial recognition).
The technology Google is trying to patent is a little different. It uses a two-part authentication system, the first being the recognition of at least one facial feature, such as that of an eye, eyebrow, the mouth area, forehead, or noise. Secondly, it looks for a gesture. You could be frowning, wrinkling your nose, or sticking out your tongue, to name a few possible examples.
"The device may emit a light beam in the form of a bright flash. The device may cause the flash to be sufficiently bright to cause a user to blink involuntarily," Google outlines in its patent application. "The device may detect the blink (one or both eyes closed either fully or partially) in at least one captured authentication image. In other implementations, the device may display a message instructing the user to blink within a particular time frame. In one such implementation, the device may capture multiple authentication images, and verify that the user blinked in an authentication image corresponding to the time when (s)he was instructed to blink."
Google also describes emitting a low light beam at either captured image to help detect whether or not the camera is looking at a real person or a photograph that someone is holding up.
Not all patents come to fruition, but given Google's past attempts at integrating face recognition technology into Android, we wouldn't be surprised if these methods found their way into Key Lime Pie.
It's true that a similar technology already exists in Android 4.1/4.2 Jelly Bean, which has users blink to unlock the device. The problem there is that it's relatively easy to thwart by taking an image of the device's owner, painting over the eyes in Photoshop with the same color as their surrounding skin, and then flashing the two photos to simulate a blink (we're not condoning this behavior, but it's a well known hack and one of the reasons why Google is still working to improve facial recognition).
The technology Google is trying to patent is a little different. It uses a two-part authentication system, the first being the recognition of at least one facial feature, such as that of an eye, eyebrow, the mouth area, forehead, or noise. Secondly, it looks for a gesture. You could be frowning, wrinkling your nose, or sticking out your tongue, to name a few possible examples.
Google's facial gesture technology is also able to detect a figure's pitch, which can be worked into the authentication process. When the software is finished analyzing everything, it then determines whether to grant a user access to the device or keep them locked out.
What about tricking the device? Google explains some of the anti-spoofing methods that might be included.
What about tricking the device? Google explains some of the anti-spoofing methods that might be included.
Google also describes emitting a low light beam at either captured image to help detect whether or not the camera is looking at a real person or a photograph that someone is holding up.
Not all patents come to fruition, but given Google's past attempts at integrating face recognition technology into Android, we wouldn't be surprised if these methods found their way into Key Lime Pie.
