Provided they are blessed with eyesight, every person on this planet uses facial recognition in their day-to-day lives.

 

Facial recognition is as old as humanity itself. It is the oldest fastest and most reliable way that the human species has of identifying other people. 

 

Recognising another person is a basic survival necessity. We decide instantly if we know them to be a friend or a foe. Or we pick up on characteristics that experience has shown us might be warning signs of trouble to come. 

 

So why are some people raising concerns about something we all use in our everyday lives? Perhaps we should go back to the start… 

 

To recognise someone, and in a millisecond decide if they pose a threat or not, everybody has to store information about the image of the person and behaviour that they associate with that appearance. Whether the person you recognise is a loved one or a known threat, memory is what allows you to get to a very quick conclusion about what your next action should be. 

 

To identify a person, you need to retain a profile in your mind. It is memory and the immediate assessment of the possibility of a threat posed by that person, based upon retained information, that makes recognition such a powerful tool. With no ability to associate past behaviour with a face you were looking at, you would not recognise a threat. A memory of a face is as important as the ability to see the face in the first place.

 

It is this need to have a recollection of an appearance or visual traits leading to a certain conclusion that people struggle with. And unfortunately, some highly vocal pressure groups, who perhaps could be accused in some cases of misunderstanding, misinterpreting or simply scandalising the subject in order to get attention has led to confusion about what facial recognition is all about, what personal data of that person is stored and where.

 

Reading some of the media you would think that face recognition systems store full images of whole families, and instantly knows who they are, where they are from, their lifestyle choices and bank account passwords! That is wrong.

 

However, whilst there is a lot of nonsense talked about face recognition, I do think it is fair to say that there is good face recognition and there is bad face recognition. Let me explain…

 

We have all of us become accustomed to living with CCTV. Few people would expect to visit a modern city and not be observed on a CCTV camera. And people accept it as a good thing. Of course, the CCTV camera simply “sees” you. It does not recognise you. 

 

It is what is on the other side of the camera that some people have become concerned about. They do not like the idea of being “watched.” However, this is to misrepresent what the person, or indeed the software, on the other side of that camera is doing.

 

Everyone is familiar with the idea of a security guard sitting in a room with a bank of CCTV monitors. If you see cameras in operation in say a retail environment, most people I assert would assume that there is a person in a control room somewhere watching the screens. I say “watching” but of course a multiplicity of screens means that attention is divided. So, one person watching 30 screens has a limited capability to spot potential problems. We have all watched films where the hero is visible on CCTV, but the guards still do not spot him entering the vital area. Enter the super recogniser…

 

Luxury retailers around the world employ individuals with a special skill. They are known as super recognisers. They have a level of observational ability way beyond the norm, and they are “wired differently” to us normal folks. As a result, they can identify people in crowds that are heavily disguised by the smallest of tells, even if wearing sunglasses, a head scarf and a collar turned up so that hardly any of the face is visible. 

 

Originally, super recognisers were used to spot high spending customers entering department stores. Later super recognisers were used to spot crooks.

 

What both a security guard and a super recogniser have in common is a memory. They store information about people they know in their brains. Nobody would object to someone being able to recognise a criminal or someone who has volunteered their information for help.  So why do people get concerned about the storage of a memory not in a person, fallible and flawed as that is, but in a machine?

 

The truth is that people are not worried about being observed by a camera but perhaps by having their information stolen. The issue is around what data is stored about them and what can it be used for. Facial recognition has been roundly and erroneously demonised as a system that gathers information. That is not the case. However, in the past some systems have stored data that can be attached to a particular person. And that is the difference between good face recognition and bad.

 

The software behind facial recognition is complex. However, the basic working of good face recognition is simple. Like a fingerprint, key points of a person’s face do not change over time, despite the ageing process, fashion, etc. Just like a fingerprint, which today people use for banking, immigration, biometric entry to their homes and phones, a face has key points that, all taken together, make a unique combination of measurements.

 

Good face recognition systems, like my company Vix Vizion’s Imagus system, DO NOT STORE DATA ON EVERYONE. They simply have a set of measurements – stored as long strings of 1s and 0s and known as “metadata” - that relate to some people’s faces. Specifically, only people who have consented to be entered into the face database or people of interest from a criminality point of view. Each packet of metadata relates to one individual. We are familiar with fingerprints, but this is of a face – a “faceprint” if you will. The system just matches the faceprint to the face seen by the TV camera if that person is spotted. And then it alerts the system operators… just like a security guard recognising a known thief entering a mall.

 

Now here is the critical bit. When a person of interest is spotted that is the point at which a system will start to capture data for future needs. If the person of interest is believed to be a known thief, then the details of the sighting and the corresponding faceprint may be retained for a period of time – perhaps to show what happened in a subsequent investigation or in a court of law. However, the stored data on the face recognition system is still basically metadata. So, you know when a face matching the faceprint was spotted and where. And that information can be stored for as long as the system operator wishes.

 

So good face recognition stores only faceprints of people - as packets of 1s and 0s that make up the metadata - that are unique to a specific person’s face that allow it to “act” when that face is spotted for a specific reason, like in the case of criminals or people who have volunteered their faces. It does not know or access any other information about them. It simply spots the person and responds accordingly. 

 

Bad face recognition on the other hand stores images of people and tries to match them for marketing or other reasons they have not consented to.  Clearly if an individual wants to be spotted – for example, when they walk into a bank, they want the teller to know for certain who they are so they can access their accounts – that is the individual’s choice. But storage of images and associated information without the permission of an individual is bad face recognition. And there have been high profile legal cases and massive fines dealt to organisations who have tried to do this in the past. 

 

How do you know if an operator is using good or bad face recognition. There is a simple rule of thumb… 

 

If you are a law abiding citizen and it is obvious that an organisation has linked your faceprint to personal information about you – such as lifestyle preferences or interests – without your permission, then this is bad face recognition. However, where an organisation is alerted to your presence - whether you are a person who has consented to be included in a face database (such as in the case of gaming self-exclusion), or you are a person of interest or a crook - but the system does not store personal information about you, that is good face recognition (although if you are a bad person any face recognition is bad news!).

 

Because of the amount of publicity afforded to face recognition in recent months in connection with retail and the apprehension of thieves, many think of this as the main use of face recognition. Actually, that is only a small part.

 

Recently publicity has swirled around the use of face recognition in sports stadiums. Much of the publicity has been misguided suggesting that everyone has their faces stored on a computer system somewhere so they can be observed. Having read the above you will now realise that is simply not true, because good stadiums practice good face recognition.

 

Arguments have been made that use of face recognition enables stadiums to identify those who have been banned from attending sports events because of their criminal or anti-social behaviour. And that is true. The faceprints of those people will be stored on the system and it will keep a look out for them. However, there are other predators in stadiums.

 

For example, some police forces have discovered that pop concerts aimed at young teenagers are a mecca for convicted paedophiles. Using the expectation that some young children would be accompanied by parents as cover, these adult offenders would mix freely with children assumed to be parents ensuring the safety of their kids. As many as 43 paedophiles have been linked to one night of a concert. A chilling thought. Had face recognition been in place, the arena security would have identified these people and challenged why they were there BEFORE they got inside the arena. And the same would be true of other criminals, like drug dealers. 

 

Face recognition systems may be used to check that only people with faceprints matching those on the system are in an area, such as a high security area, making sure only those who should be there are there. Or it may match faceprints to a list of those who should be excluded, such as known thieves in a retail environment.

 

Interestingly a growing area of face recognition is where people are enrolling in systems to protect themselves. For example, those with an issue with gaming.

 

And in a world where terror threats have to constantly be guarded against, face recognition systems on critical infrastructure allow operators to be alerted to an intruder. The system can guarantee that only those people with appropriate security clearance enter high security areas, and at times at which they are supposed to do so.

 

In conclusion, I hope that this article has achieved three important things.

 

First to allay fears that face recognition stores personal data about individuals seen on camera. It only stores faceprints of those who have consented to be put into a face database or people of interest associated with criminal activity. Faceprints are made up of metadata incomprehensible to anyone but the computer system.

 

Second, that there are two types of face recognition systems. There is good face recognition and there is bad face recognition. And using my rule of thumb, you can easily spot them.

 

Finally, the purpose of good face recognition is overwhelmingly to keep people safe or to enrich their lives. It is a huge force for good.

​

Simon Herron

Vix Vizion CEO