Google has filed a patent suggesting users pull a series of faces, like sticking out their tongue or wrinkling their nose, in place of a password to unlock their Android phones or tablets.
Google says requiring specific gestures could prevent the existing Face Unlock facility being fooled by photos, reports BBC News.
The document - which was filed in June 2012 but has only just been published - suggests the software could track a "facial landmark" to confirm a user not only looks like the device's owner but also carries out the right action.
It says examples of the requests that might be made include, a frown, a tongue protrusion, an open-mouth smile, a forehead wrinkle, an eyebrow movement.
It says the check would work by comparing two images taken from a captured video stream of the user's face to see if the difference between them showed the gesture
had been made.
The filing also notes several ways the software might check that the device was being shown a real person's face rather than doctored photographs.
These include studying other frames from the captured video stream to check that the person had made a sequence of movements to achieve the commanded gesture, and confirming all of the frames actually showed the person's face.
In addition it says the software could monitor if there were changes in the angle of the person's face to ensure the device was not being shown a still image with a fake gesture animated on top.
Such efforts might help address criticism that its current face detection software is insecure.
Google says requiring specific gestures could prevent the existing Face Unlock facility being fooled by photos, reports BBC News.
The document - which was filed in June 2012 but has only just been published - suggests the software could track a "facial landmark" to confirm a user not only looks like the device's owner but also carries out the right action.
It says examples of the requests that might be made include, a frown, a tongue protrusion, an open-mouth smile, a forehead wrinkle, an eyebrow movement.
It says the check would work by comparing two images taken from a captured video stream of the user's face to see if the difference between them showed the gesture
had been made.
The filing also notes several ways the software might check that the device was being shown a real person's face rather than doctored photographs.
These include studying other frames from the captured video stream to check that the person had made a sequence of movements to achieve the commanded gesture, and confirming all of the frames actually showed the person's face.
In addition it says the software could monitor if there were changes in the angle of the person's face to ensure the device was not being shown a still image with a fake gesture animated on top.
Such efforts might help address criticism that its current face detection software is insecure.
Post a Comment
Note: only a member of this blog may post a comment.