Fujitsu reveals tool to counter biometric facial authentication fraud
Fujitsu Laboratories has announced a new technology capable of detecting attempts to fool biometric facial recognition systems.
The tool can be installed on conventional cameras and can set apart photographs from actual face footage.
According to Fujitsu, the new biometric system has been developed to encourage greater trust in remote authentication as the technology becomes increasingly widespread around the world.
From infrared cameras to determination models
The main factors that trigger traditional facial authentication spoof detection systems are reflections or blurring displayed on a smartphone screen, for example.
However, reflections caused by fluorescent lights or sunlight and blurring caused by facial movement could be perceived similarly by facial recognition software, undermining these systems’ efficiency and calling for higher levels of security.
Near-infrared cameras or depth cameras are often used to measure the distance between the subject and the camera to catch typical signs of forgery. These systems can be expensive, however, and not always completely accurate.
Fujitsu’s new biometric anti-spoofing technology is based on a three-steps approach. First, a face image is captured by the camera and separated into various elements, its various reflection and shape elements calculated.
Image processing technology is then utilized to digitize the characteristic features of forgery for each of the separated elements.
Finally, the characteristics of each element are combined to generate a characteristic for judgment, based on the original photo and not live footage, which could potentially cause focusing issues.
The camera captures the image, and the companion software runs the analysis against a previously taken photo. This way, no additional complex or expensive machinery is required.
Commenting on the news, Fujitsu said its new biometric anti-spoofing technology has proven at least as secure as existing facial recognition systems in detecting forgery attempts.
The company said it is looking at improving the accuracy of the new tool further and is thinking about a public release by March 2021.