Facial recognition technology is quickly becoming the cornerstone of modern technological security solutions. Be it from simply unlocking your smartphone to registering access in and out of a secure office block. There’s no denying that simply lifting your phone to your face is much more convenient than finding the correct fingerprint alignment.
However, it’s not necessarily the most secure security option, especially when based on basic color camera technology that’s rather easy to fool. Cost-effective but basic camera-based recognition can be duped with photos and other mock-ups, so it’s not something you want to trust when trying to secure super sensitive data. It’s one of the reasons why we still recommend that the security-conscious use their phone’s fingerprint scanner rather than the face unlock feature. Similarly, the image-based technology found in most smartphones doesn’t adapt to changes in clothing habits, such as wearing new glasses or a face mask.
Of course, there are more secure 3D face mapping products already on the industrial and gadget markets. Smartphone examples include the Apple iPhone 12 series and as far back as Huawei’s Mate 20 Pro. 3D technology is much more secure against anti-spoofing, but the extra hardware is comparatively quite expensive. Hence why so few manufacturers include the technology. 3D face mapping is also limited to short distances, and, again, it can struggle to adapt to changing facial characteristics either and doesn’t always work as well in brighter lighting conditions.
Ideally, we need something a little more robust and flexible. One such company trying to do just that is Suprema — a leading global commercial biometric solutions provider. The South Korean company has some very interesting next-generation facial recognition technology in its portfolio which could hint at the future of face unlock on our gadgets.
Rather than using color camera image data like traditional face unlock methods, Suprema’s facial recognition system scores classic color and near infra-red (NIR) images of your face. Unlike some other systems on the market that use infra-red data just for anti-spoofing, the use of black and white near infra-red images (sorry, no heat-map Predator vision here) forms a key part of the system’s score-based verification system. AI-based visual recognition ties the system together, which analyses this unique enhanced spectrum facial data for key facial profile features. So it can’t be fooled by pictures, sculptures, etc., and works well in all lighting conditions.
The combination of AI and multiple images also allows the technology to recognize mask-wearing faces, albeit with increased false acceptance and rejection rates. In fact, the system can be used to enforce mask wearing when used for unlocking doors, etc. Impressively, the user doesn’t even need to enroll a mask-wearing picture for this feature to work. The AI-based recognition system can still verify a face based on visible features, such as eyes.
Currently, Suprema’s NIR tech is only found in its FaceStation 2 and F2 terminals (showcased in the video embedded above). When asked about scaling the technology down to smaller gadgets, Suprema noted that power consumption is the biggest obstacle to bringing this technology to battery-powered devices. Currently, Suprema’s AI algorithms run on a graphics processor, but lower power consumption could be achieved by bringing these algorithms to neural processors instead. These capabilities already exist in mid-tier and premium smartphones. Importantly, biometric data and AI processing are currently all handled on the edge, ensuring that sensitive information doesn’t leave the device. Camera hardware only requires a 720p resolution with a reasonably wide field of view, which is a very reasonable requirement by modern technological standards.
Of course, Suprema isn’t the only horse in the race to improve facial recognition technology. Trinamix unveiled a face unlock sensor that can be hidden under a phone’s OLED display in early 2021. The company also offers skin detection capabilities to assist with anti-spoofing. This skin detection idea consisting of an infrared imaging camera and a light projector, which, combined with smart algorithms and 3D sensing technology, identifies and classes materials such as skin. Trinamix has also been working with Qualcomm to run its mapping algorithms on Snapdragon Hexagon DSPs. Proving that it’s possible to run advanced classification algorithms on a wide range of Android smartphones out there.
In theory, IR facial recognition technology could bring convenient and secure facial securing to TVs, laptops, smartphones, and more, without the caveats and costs of current-gen implementations. The technology will almost inevitably scale down to lower power and cost points in time. Not to mention that improvements in battery hardware and AI processing is expanding possibilities for more demanding processing technologies, even in mainstream product markets. As we’ve seen plenty of times in the past, security features have a habit of making their way to smartphones. PINs from massive ATMs made their way to lock-screens, and fingerprint scanners from bulky accessories and laptops are now small enough to embed under a phone’s display. Watch this space.