The sinister side of facial recognition

With an estimated one CCTV camera for every 14 people in London, it would appear to be one of the most surveillance-minded cities in the world outside of China. It has been calculated that during the working day of an average person in the UK's capital would be caught on around 300 different cameras. Certainly CCTV cameras have a place in society nowadays for monitoring traffic and supposedly acting as a deterrent to criminal activity. Most of us have learned to live with them but whilst we are monitored like this as we drive and walk around or simply hangout with friends and family, we should not ignore the fact that our personal privacy is at risk. The trouble is, these are public spaces and currently there is not much that we can do about it.

But do they really reduce crime? Studies in the UK suggest that the use of cameras has had little effect as regards crime prevention in town centers. The UK Home Office commissioned a study by a local university of 14 CCTV systems and only one showed a drop in crimes committed. Car parks were the only real winners with a decrease in crime due to the presence of cameras. The reliance upon humans to constantly watch a vast array of monitors seemed to be the weak link in the chain.

Then along came facial recognition. This works by mapping facial features, mainly the eyes, nose and chin, by identifying the light and dark areas then calculating the distance between them. This information is then matched against others on a database.

London's Metropolitan Police (the Met) have been setting up vans outside shopping centers which are known for crime in order to scan the faces of the general public and compare the results with a database of 5000 registered villains and troublemakers. Admittedly the police had a signboard up saying that live facial recognition scanning (LFR) was in progress. Its effectiveness is certainly questionable, with one survey showing that 93% of those stopped were inaccurately identified and an independent review by Essex University found an accuracy in just 19% of cases.

Privacy advocates are certainly not taking kindly to this move by the Met as, unlike fingerprinting and DNA, there are no regulations currently in place governing how this biometric data can be snatched, stored and used. Since this data is being taken without the individual's consent, this will have negative consequences as regards the attitude of the general public towards the police force and legal challenges concerning this practice may well arise. Use of this technology in such an indiscriminate manner is like a virtual identity parade, with all the good people amongst us being lined up with the bad in an effort to get a match. We don't want to move down the path where those attending legitimate protests and meetings are identified and targeted as troublemakers.

In the US and Europe, LFR has been shown to be more accurate when it comes to identifying white people thus leaving ethnic minorities vulnerable to false positive readings. In the States, cellphone cameras are used by police to compare suspects with databases which hold disproportionately more African Americans. However, AI is only as good as the subject matter upon which it has been trained and in the US it has been predominately set up using white faces. In China, the AI has been trained on Asian faces and therefore works a lot more efficiently in that environment. Studies have indicated that people have difficulty recognizing faces of another race and this kind of bias could be moving over into AI. LFR can at times struggle with the lack of contrast on darker skin and also with women using makeup or wearing their hair differently.

But now in the UK one major concern is if the technology could soon make the jump from a simple arrangement on a van outside a shopping mall to operating via a city-wide network of cameras. If it goes unchecked, this will just be the start of more and more surveillance as the authorities push the boundaries as to what they can get away with.

An adviser to the Home Office is suggesting that such use of LFR is something for Parliament to decide upon and a legal framework needs to be in place. It does not stop there though as other technologies such as voice recognition, gait and iris analysis will require legislative control, as will other biometric technologies which will no doubt emerge in the future.

A leak of an internal memo from the European Parliament showed that it was considering the use of LFR to provide "security and services" to its members. This did not go down well with MEPs and thus the Parliament was forced to drop its "digital transformation programme". Taking into consideration the EU's data protection laws, it has now transpired that they are debating a ban on LFR in public places such as stations, stadiums and shopping centres. Good news indeed for the vast majority who value their privacy.

In China the technology is far more advanced and the shift has already been made with many aspects of people's lives being subjected to LFR. In the PRC it is not just a means of catching criminals but a formidable method of social control. From banks to airports, from hotels to public toilets, it is the police and state security who are the most enthusiastic about it, with the plan being to interconnect all private and public CCTV cameras into one vast surveillance network. So in China, who you are, where you go and with whom you socialise will soon be stored on your own government-customized record thanks to LFR technology. The eyes of the masses will also be used with neighborhood committees and other informers able to watch security camera footage on their mobiles and TVs. The advances in the technology in China have been quite something since their LFR is now able to classify targets by their gender, clothes and hair length, plus they can be followed from one camera to the next based upon their faces alone. This has been particularly useful for the central government in order to keep a tight control over the population in the Xinjiang region. The ultimate surveillance state has now been formed.

From a privacy point of view, one particularly dangerous piece of facial recognition software is Clearview. The developers have been gathering up images from social media accounts for some years now and all users' activities, interests and networks are sitting in the Clearview database. From this, it can retrieve a complete profile of a person just based upon one photo when compared against the over three billion which it claims to hold on file. In a recent in-depth analysis of Clearview and its Australian founder, Mr Hoan Ton-That, the New York Times described it as "the secretive company that might end privacy as we know it". Law enforcement, however, take a different view and are certainly very upbeat, describing Clearview as "the biggest breakthrough of the last decade". They have been able to identify criminals using the software and it is thought that some 600 police departments in both the USA and Canada are using it. No doubt it will also have caught the attention of the FBI and Department of Homeland Security. However, a powerful tool like this in the wrong hands can be extremely dangerous. It says a lot that even Google back in 2011 declined to go down the path of creating something akin to Clearview for fear of misuse. Then-CEO Eric Schmidt said it was a perfect tool for dictators to use. So far, other than for law enforcement use, we do not know into whose hands this software has fallen. Since it is against their "rules", Facebook, in a futile attempt to stem the flow of photos, have sent cease-and desist letters to Mr Hoan. Kind of ironic when you think about the things Mr Zuckerberg gets up to.

So now that you know about Clearview, will you regain control of your privacy by jumping into action and making your Facebook account private? Certainly from that point onward but here's the rub. Whatever was out there previously for all to see, Clearview will already have it and won't be wiping it from their database any time soon. Mr Hoan has released this product into the market perhaps not realising or caring about the negative consequences. When finally tracked down he did admit that his product needed some form of government regulation but at this juncture, the genie is truly out of the bottle.

For further background on this feature, here are a selection of recent press articles for your reference: