Featured

Police defy human rights outcry to press ahead with facial scanning technology

Police defy human rights outcry to press ahead with facial scanning technology
Written by Publishing Team

Shoppers hoping to catch a post-Christmas bargain on London’s Oxford Street encountered an unusual additional element to their retail experience last month. As they navigated between the shops, a large dot matrix sign announced: “Police Facial Recognition in Progress.”

By the time the day-long deployment by the Metropolitan Police Service had ended, its van-mounted mobile cameras had scanned the faces of 12,100 people wandering the famous shopping street to find matches with a database or “watchlist” of wanted or vulnerable individuals. Four men, including an alleged fugitive drug dealer clearly not in the habit of reading street signs, were intercepted by officers and arrested.

The operation using live facial recognition or LFR was the first of its type by Britain’s largest force for nearly two years and represented part of a flurry of activity using a controversial technology touted by its proponents as a powerful tool for snaring criminals and deterring violence and by its detractors as the vanguard of an Orwellian, industrial-scale breach of privacy that threatens core freedoms.

Some 150 miles from the capital, 70 officers from the neighbor South Wales and Gwent forces are currently trialling a different version of facial recognition technology which uses a smart phone app to instantly check the face of a potential suspect stopped on the street to see if they are wanted.

Officers are sent a selection of photographs to verify the identity of a suspect, though commanders insisted the system will only be used in circumstances such as when there is a suspicion that false or misleading details have been provided.

In the meantime, two forces, Nottinghamshire Police and the Met, are making use of yet another form of the technology – known as retrospective facial recognition (RFR) – which scans existing images taken from sources as diverse as CCTV of a pub brawl to social media postings to match against a “mugshot” data.

More on Facial Recognition

Like all facial recognition technologies, the systems work by capturing multiple measurements of the human face, such as the distance between eyes or from forehead to chin, and creating a “facial signature” which is then compared with a database. If the system finds no match, the signature is instantly discarded.

Scotland Yard quietly announced last year that it had signed a £3m contract with a subsidiary of Japanese tech giant NEC, the same provider of its LFR software, to upgrade its RFR capabilities while the Nottinghamshire force said last month that it had begun use of its newly-acquired system with the aim of speeding up the arrest of repeat offenders.

The result is what appears to be a concerted push by some of Britain’s police forces to overtly deploy a technology which is already used or proposed in scenarios from biometric passport gates to submitting tax returns in America – but which remains hugely contentious when it comes to being used to swift law-abiding citizens from those seeking to evade the long arm – or rather, advanced sensors – of the law.

It is understood by i that up to five UK forces have acquired live facial recognition equipment and six are operating or testing the latest retrospective software.

Nottinghamshire Police said its trial was in effect an accelerated version of the tried and tested tactic of comparing visual evidence with a database of known offenders. Assistant Chief Constable Steven Cooper said: “Schemes like this are a vital way for officers to quickly take repeat offenders off our streets.

But it is a digital policing revolution that is happening against a backdrop of deep concern that the legal basis for deployment of facial recognition technology is at best unproven enforcement and its accuracy as a law tool open to question.

Some 31 human rights and privacy organisations, recently published an open letter calling for an outright ban on LFR, saying both police and the Government had bypassed Parliament in setting rules for its use and asserting that the technology “fundamentally alters the relationship and balance of power between citizens and the State.”

They are far from being alone in feeling queasy. Among those who have raised concerns are Elizabeth Denham, who while Information Commissioner last year said she is deeply troubled by the potential for LFR to be used “inappropriately, excessively or even recklessly” by both public and private bodies, and a Surveillance Camera Commissioner who has warned that existing guidelines do not go far enough in providing “effective scrutiny”.

The EU’s two most powerful data protection bodies last year warned that LFR “interferes with fundamental rights and freedoms”, while a number of countries including New Zealand and Belgium have imposed a ban or moratorium.

Professor Peter Fussey, a surveillance and human rights expert at the University of Essex, who carried out an independent study of trials of the Met’s LFR system, told i that UK police forces are pressing ahead with the technology at speed despite the absence of specific – or “explicit” – governing legislation its use.

He told i: “There has been this alacrity within the Met and elsewhere to push forward with facial recognition technology really, really quickly. The issue is that intrusive technology of this nature needs an explicit legal basis and it really isn’t clear that this explicit legal basis exists.

“In other jurisdictions we are seeing countries pause with introducing this capability while legislation is put in place. That has not been the approach in England and Wales.”

Questions remain over accuracy and bias in facial recognition software

One area of ​​concern is that facial recognition software was at its inception developed and tested using the biometric details of white males and as a result risks discriminating when it comes to assessing different genders and ethnicities.

An American study in 2018 found that while error rates were just 0.8 per cent for pale skinned males, they are rocketed to up to 34 per cent for darker skinned women.

Professor Peter Fussey, of the University of Essex, who has studied the use of live facial recognition, said: “Racial bias is a scientific fact in these systems. A problem faced by these systems is that they are inherently discriminatory because of the way they were designed.

“Their effectiveness can be improved but you still have to put in place measures to mitigate these inherent biases. Even then, this doesn’t resolve the question of the explicit legal basis on which these systems are being established.”

Both police forces and software makers insist that “red boxing” – the moment when the software places a red box around an on-screen face to take measurements – technology has improved to the point where the rate of “false positives” based on ethnicity or gender is not statistically significant.

Campaigners remain unconvinced. Ksenia Bakina, legal officer at Privacy International, one of the organizations calling for LFR to be banned, said: “We are concerned that [LFR] may be used in a broad range of public gatherings such as sporting events, music concerts, and protests, threatening protected rights. Deployments of this surveillance technology could mirror and exacerbate existing disproportionate policing practices towards the minority communities.”

It is a debate that is set to continue to rage. As one law enforcement source put it: “In the end, society will have to decide whether being ‘red boxed’ is a price worth paying for an enhanced policing capability. It’s a really difficult judgment.”

Critics have pointed that no fewer that 30 separate organizations play a role in deciding how Britain’s police use new technologies such as facial recognition, with no single entity exercising overall control or scrutiny. Crime and policing minister Kit Malthouse said last month that a new National Policing Digital Strategy will provide “the right governance to sort all this stuff out”.

The National Police Chiefs Council declined to confirm the number of forces using facial recognition but said the technology had proved its usefulness in tracing wanted criminals and the decision on whether or not to intervene always remained with officers.

A spokesperson said: “We carefully consider our decisions to deploy facial recognition technology to ensure we do so in an effective, ethical and proportionate way, with the right safeguards in place to protect people’s privacy and human rights.”

Supporters of facial recognition technology, in particular LFR, argue that technical shortcomings with the systems on offer, in particular their accuracy in correctly matching passers-by with watchlists, have been ironed out. In its deployment last month on Oxford Street, Scotland Yard said the false alert rate – the number of times the system had wrongly identified an individual – was 0.008 per cent.

The Yard said it was confident in the “operational effectiveness” of its facial recognition equipment and indicated it is ready to use LFR routinely on London’s streets. A spokesperson said: “The Met may deploy LFR where this is an intelligence case to support that deployment and there is a policing need to do so. This technology helps support our commitment to tackling violence.”

In order to comply with existing rules, LFR deployments require members of the public to be notified that they are walking through a zone where their faces will be scanned and compared to a database.

The Court of Appeal last year emphasized that the watchlists checked by the software must only feature individuals who might be reasonably expected to be in the area of ​​the deployment, for example because they live nearby. In other words, overtly cameras cannot be linked to a system like the Police National Database which contains millions of records identifying individuals including terror suspects and organized crime groups.

But campaigners insist there are still significant question marks, in particular with the efficiency of facial recognition systems and whether they discriminate on grounds of gender or ethnicity.

Some experts argue that given the amount of resource required to use LFR in terms of personnel to monitor cameras and respond to alerts, they represent poor value for money. It is understood that one deployment lasting six days resulted in a total of eight correct identifications of wanted people, the majority of whom had already had their cases dealt with by courts because watchlists were out of date.

A separate analysis by i shows that a total of 68 LFR deployments by South Wales Police between 2017 and the start of the pandemic – at locations ranging from an Elvis Festival to a royal visit – resulted in just 59 arrests from a watchlist total of nearly 36,000.

South Wales Police did not respond to a request to comment on its use of LFR or its app trial but it has insisted it only uses facial recognition when it is “necessary and proportionate” to do so.

Officers nonetheless recognises that many remain to be convinced. One law enforcement source said: “Facial recognition has the potential to be a powerful weapon [in fighting crime] and we need to give it time to get it right. But policing in this country is by consent and putting a van on a high street with signs saying ‘we’re scanning you’ while you walk to the supermarket or pub is a big deal. We have to persuade people that this really works.”

About the author

Publishing Team