The announcement that Aadhaar will make facial recognition a mandatory part of authentication has pushed understanding this technology to the top of everyone’s agenda. Unfortunately, despite many recent advancements, misconceptions about facial recognition still reign, like the ideas that facial recognition is “racist” and that it’s easily spoofed by pictures of faces or masks. This is simply not true.
In all fairness, the public has good reason to hold these beliefs. Legacy solutions – and even some contemporary (inferior) solutions – are plagued with issues that could make anyone wary of facial recognition’s effectiveness and trustworthiness. However, new, state-of-the-art solutions do not suffer these flaws.
The task then, for the government or business in question looking to implement facial recognition is to find the right solution-provider. With that in mind, here are some of the most worrying issues one can still find in second-rate facial recognition technologies – and how more advanced solutions overcome them.
- Doesn’t work in real-world lighting conditions. The scientists and engineers developing facial recognition technologies have had an unfortunate tendency to only test their solutions in optimal lighting conditions, in which many solutions work quite well. Put these solutions into a very dark room or in front of a bright window, however, and they fail, either not allowing you to access your device (bad) or just giving up letting anyone access it (worse).
Fortunately, some facial recognition developers have discovered ways to extend the situations in which the technology is effective. SensibleVision’s “Facebright” technology, for example, lights up the user’s to authenticate them in a dark room common situation. Bright lighting is still tricky, but nonetheless significantly less of an issue with modern 3D facial recognition, particularly Time-of-Flight solutions.
- Only works on European faces. Headlines often claim that “facial recognition doesn’t work on darker skin tones,” or even “facial recognition is racist,” full-stop, as if it were true of all facial recognition solutions. In fact, this is an issue only confirmed in the few solutions tested in the two or three studies actually conducted on the topic. The issue with these solutions is most likely a lack of comprehensive face data.
SensibleVision, as a long-time global provider of facial recognition technologies, has access to face data from around the world, and uses it. The need to avoid biases toward certain skin-tones is also why we use so many distinct depth data points of reference for authentication – more than 10,000 all of which are color blink and not affected by most lighting conditions, in fact. In short, despite their connotation of “intelligence,” AI algorithms are only as good as the data you feed it – so make sure your facial recognition provider feeds it with images of diverse faces.
- The user experience is clunky. This is the myth that I have most trouble understanding. (Surely looking at a screen isn’t hard?) The myths that facial recognition requires you to make a specific face or take your glasses off are just that – myths. Newer technologies are smarter than that. In fact, facial recognition in most cases improves the user experience of the device, application or activity in question. SensibleVision, for example, uses facial recognition to automatically and instantly customize the settings of a shared device to whoever is looking at it. And consider a wallet-free supermarket, like Amazon Go, but without even having to scan your phone or download an app: Just walk in and walk out. That’s what facial recognition can do for the user/customer experience.
- Facial recognition is easily spoofed. Again, it is perhaps the legacy of bad or early technologies that cause many to think that all facial recognition is easily spoofed. Modern solution-providers, however, like SensibleVision, use machine learning – fed by millions of points of face data – to train their facial recognition software to recognize someone with incredible accuracy. And they measure thousands of unique data points on the user’s face upon each authentication to ensure a match. The result is that 3DVerify by SensibleVision has a false acceptance rating of 20-million-to-1, and its false rejection rate is approaching zero. And with techniques like liveness detection, printed pictures and even masks don’t stand a chance.
To summarize, many worries about facial recognition simply reflect unfamiliarity with modern facial recognition. It’s through no fault of their own that many people hold these beliefs – it’s thanks to bad technologies with overhyped claims that myths about facial recognition persist. But considering that Aadhaar currently uses fingerprints as its primary form of biometric identification – which very much can be spoofed easily – facial recognition, as long as it uses the technologies and techniques mentioned above, is sure to be an improvement, both reducing fraud and increasing the speed of authentication.
George Brostoff is CEO of SensibleVision, which makes facial-recognition software. Brostoff has seven US patents and has developed technology used by Dell and other major companies.
By George Brostoff, CEO, SensibleVision