Should we ban facial recognition?
Biometric identification technologies often appeal to the imagination of people and usually quickly tap into all kinds of emotions. Large volume implementations of especially finger print recognition have however proven the value of biometric systems and have led to a wider acceptance of biometric identification.
It seems however that facial recognition is causing much more debate than other biometric identification technologies. San Francisco has even decided to ban the technology. Other cities are considering to do the same thing. Why is that?
The obvious reason of course is that if there is one biometric technology that reminds us of Orwellian Big Brother scenario’s, it is facial recognition. And that is understandable. But is that a good reason to ban it’s use?
The three steps of biometric identification
Any biometric technology is using the same principle (simplified single factor scenario):
- Enrollment: An image of the biometric characteristic is taken, processed and converted into a template with distinct features of the image and that template is enrolled and stored into a system.
- Scanning: An image is taken when someone needs to be identified or when an identity needs to be verified and that image is also processed and converted into a template with distinct biometric features.
- Identification / verification: The two templates are compared and will lead to an identification result of a certain accuracy.
This process is used for finger print, iris, palm vein and all kinds of other biometric recognition systems. Usually people focus on the third step: What is the accuracy of the system? What are the chances of false acceptance or false rejection?
The unique features of facial recognition
To understand the public debate about facial recognition better, it is however more relevant to look at the first two steps and focus on the differences between facial recognition and other biometric identification systems.
Facial recognition systems are completely based on optical recognition. Which means that the enrollment of a facial image could be digitized. Which means that a good picture of someone could be enrolled into the system without the consent and physical presence of that person at the time of enrollment. An example of this would be the cases where missing children are retrieved using facial recognition. This is only possible if facial images of those children would be available and these images are enrolled into a facial recognition system. This would be more difficult to achieve with technologies like fingerprint or iris recognition.
And facial recognition is also different in nature in the second step. Facial recognition software today already can run on many types of devices. No specific cameras are needed. And facial recognition SDK’s (Software Development Kits) are widely available, including in open source projects. We do not see this separation of hardware devices and the identification software with other biometric technologies. There are impressive steps being taken in finger print recognition, but in those cases the enrollment and scanning are not as easy and non-intrusive as with facial recognition software, because it takes much more effort to take a picture of the minutiae of the finger print (and forget about veins) than of a face.
Another difference that is related to the second and third step of the process, is that facial recognition is far better equipped for simultaneous identification of people. The comparison algorithms can run on powerful server platforms and cameras of sufficient quality can cover areas with multiple people. In that sense it resembles technologies like ANPR which also sometimes raises privacy concerns.
Should facial recognition be banned?
In the public debate about facial recognition we sometimes hear arguments used which not always seem relevant. Sometimes people refer to accuracy of the system and claim that it is dangerously inaccurate. But accuracy is depending on many things (setup, topology, situation, etc.), the technology is also continuously developing and for many applications a lesser accuracy level may still bring value, so perception of the accuracy of the system would be a far-fetched reason for governments to consider a ban.
Privacy related concerns sometimes concentrate on the storage of images. Most facial recognition systems however do not necessarily store the actual facial image of the face, but a digital representation of the facial features, which cannot be converted back to an image. Storage of CCTV footage is already under legislation and it seems that current and upcoming privacy laws provide better protection of privacy than the explicit banning of the use of a technology.
Facial recognition technology exists, is here to stay and it’s performance is continuously improving. A large infrastructure of video surveillance systems and in addition millions of handheld devices with cameras exists globally and a large portion of this infrastructure potentially could be used for facial recognition. The easy enrollment and scanning is the strength of the technology, but is also considered a weakness in the public debate about privacy.
Privacy related concerns are valid and should not be neglected. It does however not seem wise to jump to conclusions based on misunderstandings and misconceptions. Any technology can be used for good or bad and banning it’s use by governments is also preventing them from using it for purposes that serve public interests. Finding the right use cases and support them with legislation that prevents abuse but supports those valuable use cases would make much more sense.
Facial recognition is one the least intrusive biometric identification systems. In the security industry we already see great applications of the technology, like identity verification in access control and attendance registration in educational organisations. So how do you feel about this topic? Let us know!