You have learned your face recognition algorithms to adapt to face masks, to use at airports for example – tell us more about how this works?
– Actually facial recognition is based on the recognition of your individual face characteristics with another reference such as a picture of you. Each face is different from another so when analysed it can be used as a very efficient identification method.
The objective of our improved solution is to be able to focus on the top part of your face only and not the entire one so it can adapt to the new trend of people wearing a mask. In other words we train our algorithms to be sensitive and accurate enough, to capture all the singular characteristics of a specific part of a face and be able to compare it with a reference template with the same efficiency as if we could capture the entire face for the matching system.
To give a bit of details, thanks to our deep neural network algorithms the technology is more than 99.3 percent accurate when no mask is present, and around 97 percent for mask wearers – as long as the reference photo is of good quality, (such as in a passport or ID document) and does not hide the eyes as there are a lot of very singular information in your eyes!
Hopefully the pandemic will be over in a near future – will people to some extent still wear masks and how much interest will there be in a product like this then?
– People wear a mask because they respect the current healthy restrictions. That’s why we have trained at an early stage last year our algorithms to adapt to this new situation, but this is not an optional feature we added – it is simply and fully integrated to our solution. We have ‘simply’ improved the accuracy of our solutions so it works with masks. So when we’ll be back to a normal situation when masks are not compulsory anymore (we hope very soon) no changes will be required as it will just be part of the overall improvement. Moreover, good to remind that in some countries like in Japan, masks can be part of a traditional population habit. So still other valid reasons out of the pandemic context to get an improved solution.
There are many critical voices against biometric data as it is more sensitive if it’s leaked and so on – how do you approach this and what are your thoughts about the future for biometric data as a security solution?
– All those questions around security and data protection are essential for biometric solutions. That’s why several regulations are being discussed in various countries, including EU and we will continue to make sure our solutions comply to it.
For instance GDPR has categorized biometric data as highly sensitive which is a great step, so these data have to be cautiously managed and protected. That’s why all our solutions for instance are GDPR compliant, so the data are safe and well protected against any theft attempts – as data are encrypted with very sophisticated cybersecurity mechanisms. What’s more, the biometric data as you ‘know them’ your face, your fingerprints are not equally registered in a system, they are digital (Mathematical) representations generated by the system itself – this means that they can’t be exploited out of this system. Biometric is successful as it brings the ideal balance between security and convenience. Biometric solutions have reached top levels in terms of security and enable very smooth and frictionless experiences (ex: biometric boarding, your face to replace your corporate access badge etc). This level of security has also been reached thanks to new tech such as AI/deep learning that enabled to add specific anti-spoofing mechanisms – such as liveness detection where the system is able to recognize a real face from a video or a picture that would be presented by someone trying to use another identity.
Now what’s needed on top is a bit of education, to explain why it is secure, and to inform users and bring as much transparency as possible in the way biometric data are used before asking any consent.
