There are currently three uses of biometric facial mapping/recognition within UK Police Forces.
Static Facial Recognition – deployed from still facial images taken from body worn cameras, social media, CCTV, etc, and run through a source (reference) facial image database such as a custody suite. This is not a real time deployment.
Live Facial Recognition – deployed in the field, running live in real time, checking faces in crowds against a reference facial image database. Primarily, up to present day, reference facial image databases have been prepared specifically for the event.
Post Video Facial Recognition Analytics– deployed on video post event – not in real time. The potential exists for many sources of video to be amalgamated and analysed running biometric facial mapping/detecting/recognition technology.
Static Facial Recognition use:
UK Police have been using, nationally, biometric facial recognition technology since 2014 on the Police National Database (PND) of facial images. The PND comprises of custody suite photographs uploaded from all but a few of the 48 UK police forces. Invariably this facial dataset comprises of innocent people that have gone through a police custody suite: people arrested then charges dropped and people arrested and subsequently found not guilty in court. It also consists of guilty people that have served their dues to society and yet their faces remain on the PND.
There are over 21 million images on the PND, mostly comprising of facial images but also of tattoos and other distinguishing features such as scars. The Home Office estimates that 12.5 million of these images are searchable though the biometric facial recognition system. The Home Office does not know if there are 12.5 million images of individual people as there are duplicate images of individual persons.
|(Image courtesy of the Press Association)
|In February 2017 the Home Office published the Custody Image Review, in which it revealed that non-convicted persons can apply to have their images deleted from the PND. Potentially this applies to hundreds of thousands of people. However, a Freedom of Information request by the Press Associationrevealed that only 34 images had been removed upon request from February 2017 to February 2018, one year from the review.
The current PND facial image database facility does not have the facility to ‘weed’ out categorised images of innocent people automatically. The process is manual – therefore, according to the Home Office, a job too expensive to do due to manpower costs.
All forces have access the to PND biometric facial search facility. In addition, some forces have their own ‘in house’ static facial recognition systems – Leicestershire, South Wales, Merseyside and the Metropolitan Police – to use on their own custody suite photographs from images gained from CCTV, body worn camera, social media, etc.
Live Facial Recognition use:
Leicestershire Police were the first police force to use live facial recognition technology at the 2015 Download Music Festival at Donnington Park. This caused controversy at the time as this was done covertly, and in fact it was a ‘proof of concept’ for the company supplying the technology, NEC. Leicestershire Police had no results for the use of the facial recognition biometric technology, the results of the screening being the property of NEC. Effectively the unsuspecting festival goers were guinea pigs for the Japanese corporation NEC’s facial recognition technology.
Next to follow were the Metropolitan Police using the same technology at Notting Hill Carnival in 2016 and again at the carnival in 2017.
South Wales and Gwent Police also have access to the same technology under Programme Fusion. South Wales Police have deployed live automated facial recognition technology at various sporting and music events in Wales. Since July 28th 2017 South Wales Police have used the biometric technology to make 191 positive matches, leading to 50 charges, 12 arrests and 8 prison sentences handed down at just under 10 events. To date Gwent Police have not used the technology but have access to it.
Video cameras located at these events are mobile, mounted on police vehicles. South Wales Police are overt (in a sense) in that they state Facial Recognition technology is in operation – however, the choice to participate is not optional. At the point at which automatic facial recognition is being deployed live, and an unsuspecting member of the public has read the facial recognition notice on the police van operating the biometric technology, said member of the public has already unintentionally participated in the facial screening process. This raises serious consensual issues.
However South Wales’s communications on this is, at least, slightly more overt than the ‘informative’ process the Metropolitan Police deployed at Notting Hill Carnival 2017, where an A6 leaflet was distributed near the van using the technology.
At the 2017 Notting Hill Carnival the surveillance camera van was hidden, fully screened, behind about seven foot fencing at the one of the entrances to the carnival, with only the camera popping over the top of the screening to record the public entering and leaving the event.
None of the police forces deploying live facial recognition technology give prior warning that facial biometric screening will happen at these events.
Source image databases to be referenced are compiled before the event, comprising of wanted persons and persons of interest, not just complied from the PND custody images. The facial images which the live facial recognition referencing are from various sources, of which the author is yet unable to determine.
Post Video Facial Recognition Analytics:
Private companies can now offer an amalgamation of surveillance video from various sources, i.e private, corporate, public cameras, body worn video, mobile phone, etc, to merge into one format to be then run through various analytical processes.
Those analytical processes can be facial recognition, facial detection, tracking of soft biometrics (such as clothes, beards, glasses), identify areas at risk (low lit, persons congregating), anomalous items left in areas, footfall, gait… the long term identification possibilities are potentially endless.
At the point of writing post video analytics are not employed on any large scale by UK police however, recent early 2018 developments are:
January 2018 – SCC awarded a contract to supply of 40 of the UK Police forces with, “The Video Analytics (VA) solution will provide UK forces with advanced post-event video analytics allowing them to view and analyse large amounts of Mobile, CCTV and Body Worn Video formats to support the prosecution process. This service, which is part of the developing portfolio of SCC Public Safety Solutions, is underpinned by the pioneering SeeQuestor platform. This will provide video ingestion, conversion, case management and an analytics capability including face, body and attribute detection, and subject re-identification.
“One of the key advantages of the acquisition, the two parties claim, will be the opportunity for NPS to integrate NEC’s facial-recognition and other biometric technologies into its software products.”
The Connect platform features an integrated police Information system using real time mobile to desktop, intelligence, investigation, custody suite case management, collaboration between partner agencies, geographic location information systems, whole data store cloud-based platform using analytics and big data. Enabling police forces to communicate with each other on the Connect platform.
The potential for profiling and assumptions made by post video analytics are unprecedented. The amalgamation of video, stored and repeatedly analysed has the unparalleled potential for agencies (and private companies) to severely intrude into the public’s privacy and civil liberties on many levels. Whilst it could be argued that this level of ‘post event’ scrutiny could assist crime solving, the opportunities that exist to mine for information/data is hugely tempting and lines need to be drawn up to absolutely ensure unnecessary ‘fishing expeditions’ do not occur on post event video analytics.
Society is blindly wandering into a ‘Minority Report’ scenario of surveillance and data, a like of which has never been experienced before. It is not a question of stopping this before it is implemented as the technical abilities are being implemented – quietly, behind the existing street furniture, police worn tech and hand-held devices. Essentially, visibly, nothing has outwardly changed to the hardware, but the capabilities lying behind the hardware is expanding into a Brave New World of data, surveillance and scrutiny.
By Pippa King