This episode of Black Mirror sucks: London cops boast that facial-recog creepycams will be on the streets this year London's Metropolitan Police is to start routinely deploying facial-recognition tech across the English capital despite legal and public opinion challenges, the force declared this morning. The so-called AFR (Automated Facial Recognition) technology, supplied by NEC Global, will be deployed on London's streets to monitor …
Police simply can't be trusted with technology. Go back 12 years and there were complaints about how the DNA database contained something like 500,000 errors.
And the rates are worse than you've made it out to be too: The BBC were saying the Met were claiming that only 1 in 1000 people scanned would generate a false alarm. That's going from a known and proven failure rate of 98% back in May 2018 to one of 0.1% today. That sort of progress in only roughly a year and a half is questionable to say the least.
Interesting to note too the ICO's continued obsession with public opinion in their press releases. It's almost as if they've forgotten that their job is to enforce the law, not try and make it popular.
They're playing fast and loose with the statistics. They say 70% of wanted people who walked past the camera were matched. How do they know that? That would presume that they know exactly how many wanted people walked by and were able to identify them all by some other means. It doesn't take account of how many people who were on the watchlist sauntered past completely unnoticed by computer and by plod (the "unknown unknowns" in Rumsfeld-speak). What they actually mean is that *at least* 30% of wanted people were not spotted.
As for the 1 in a 1000 false positives, we're meant to take that as meaning it's right 99.9% of the time but it doesn't consider the number of false negatives (wanted people who are not identified) and as I said above that presumes that we know how many wanted people are actually in the crowd to start with.
Then there's the fact that they are talking as if all people are uniformly likely to be picked out in error. As regular Reg readers will know, facial recognition is notoriously bad at identifying non-white people so while the overall false positive rate might be 1 in 1000 that could be something like 1 in 100,000 white people but 1 in 100 or even 1 in 10 black people (depending on the makeup of the crowd). Being stopped every tenth time you step out the house could get really annoying really quickly.
Finally, we are meant to just assume that everyone on the list is there because there is some genuine need for the police to stop them. They don't tell us anything about how accurate and up to date their data is. Sure, they may identify 7 out of 10 people they're looking for but if those people aren't actually wanted by police then the efficacy of the facial recognition system is greatly diminished. GIGO.
Addendum: finally finally, there's no comparison given for this facial recognition system against other methods such as, you know, giving coppers a bunch of mugshots or even just randomly stopping people and fingerprinting them.
...but alarmingly this has only been accepted in court relatively recently.
There's no doubt a fair number of unsafe convictions* based on flawed fingerprinting evidence, on the assumption that a fingerprint match is infallible, and that experts cannot make mistakes.
*Some high profile ones, if you can be bother to google it. I believe one case involved a claim that a police officer was in a property when she was not, based on a single smudged fingerprint, which resulted in her conviction for perjury, and which ended her career.
Well, I suppose they could have hired a hundred people as stand-in "wanteds", and told them to walk through the surveilled zone at some random time (or a time randomly chosen but unknown to the operators), and what fraction were "spotted" by the system at the time they were actually there.
Whether they did this, or some comparable test, is open to question. If they had, they could have said so, and made their claimed "70%" figure a good deal more credible.
As Mark Twain wrote "There are lies, damned lies and statistics"
I agree with your scepticism of the apparently stunning advance in the recognition capabilities (read: we pulled this number from our propaganda files so I would add a 'citation needed' tag on that).
Facial recognition also struggles a bit with sunglasses (I wear photochromatic glasses and even in winter there is sufficient UV to make them go dark) and beards (which I also have).
Until the plod become technology literate (I won't hold my breath on that) they will not understand the limitations of the technology; all they see is a shiny new toy that removes the need for proper policing.
Slightly related.
In the 1970s, a UK university ran a study of the effect of calculators so it was a technology study in that sense.
They enticed people in with a couple of quid and were then sent to one of two rooms with desks, a question paper and calculators.
In one room, the calculators operated normally; in the other they were rigged to give ridiculous incorrect answers - think 7 x 11 giving a result in millions - over 70% of the answers from the group with rigged calculators ticked the clearly incorrect answers.
I would expect the plod to blindly accept the (incorrect) results and cuff someone because 'the computer says so'.
Idiots.
This post has been deleted by its author
"As Mark Twain wrote "There are lies, damned lies and statistics""
Let's not tarnish the good name of statistics by suggesting that they are the issue here - we have been told that this system is 70% accurate for something but we have no idea what that something is.
If, in 5 years, when millions have been spent on this system, we find out that "70% accurate" describing the installers ability to point them at an area of interest that Scotland Yard had provided, I won't be entirely surprised.
Or is 70% the rate at which Chinese authorities have been able to extract organs from "criminals" identified by the system?
I have a few T-shirts with the hyperface pattern on them - looks quite geeky but cool as I'm a privacy wonk. That said (as I mentioned in another post somewhere here) - just round up the people with hyperface shirts.
Although as this article... https://qz.com/878820/new-camouflage-promises-to-make-you-unrecognizable-to-facial-recognition-technology/ says the actual point is not to elude the facial recognition software but to reduce the confidence level to the point where the system doesn't flag you.
To that end... sunglasses, hats, new beards, brollies etc. likely do help... unless as above it makes you the one standout idiot that Mr. Plod would zero in on anyway.
I read some years ago that when the DNA database was run against itself (i.e. looking for matches within its own database), it was assumed that such matches were duplicates. But further investigation revealed that several such matches were of obviously different people - e.g. one of a person described as 1.7m tall and 58 years old against matching DNA from a 2m tall 25 year old (with nowhere near that time difference between when the samples were taken).
DNA profiling does not compare the entire genome. It focuses on specific small areas where there is enough variability for a profile to almost uniquely identify an individual. At least, that's as much as I can recall, I'm sure there are others reading this that can explain better.
This post has been deleted by its author
@Vimes
The Met has been taking the piss for a lot longer than 12 years, you need to go back to the '90s for that.
Re oversight and auditing, it will be the same as the record with all the illegally held photos they have, far from feeling the need to uphold the law and to prevent crime, the average phuzz is of the opinion that if they look hard enough they can nick anyone. British police make me more nervous than the criminals.
It might seem that way (and the Met aren't helping the perception). The main reason crimes go uninvestigated is that the government has cut the funding to the police significantly since 2010, and as a result there are fewer officers around.
At the same time, the government has significantly cut funding to social welfare; not just things like benefits, but also to councils, which can no longer keep things like youth clubs running, or provide enough social housing. As a result of this, there are more people in desperate situations; more people on the street, turning to drink and drugs as their only escape, and as a result of this, crimes go up.
What happens when the number of police goes down and the number of crimes goes up? Well, addressing the root cause would be helpful, rather than blaming the police. It's probably worth considering that they are also being pressured into "producing more results with fewer resources", which necessarily leads to more invasive surveillance-state stuff like this.
Taking the piss was the Met's strongest suit even in the 1950's*. and I am sure that anyone who was alive in London in the 1930's will fall over laughing at the idea that things were better in earlier years. Just how difficult it to stop mail bags from leaving trains without due care and attention?
* Since the Kray gang featured regularly in the Daily Mail, you would have thought the idea of arresting them could have come to an ossifer or two.
"the average phuzz is of the opinion that if they look hard enough they can nick anyone. British police make me more nervous than the criminals."
If the laws of the land are such that everyone is a criminal depending on the mercy of their betters to be allowed to get on with their lives in peace, then the police are not the worst problem - politicians are.
"1 in 1000 people scanned would generate a false alarm. That's going from a known and proven failure rate of 98% back in May 2018 to one of 0.1% today."
No it's not. 98% of people flagged were not on the watchlist. That's not the same as 98% of people being flagged.
Say 100,000 people are caught on camera. 98 are flagged innocently, 2 are on the list. Here you have 98% false positives as reported before, and 0.1% false positives as reported here - they are very different metrics.
It grates me if numbers are so grossly misused. Just because you agree with the criticism (which I do BTW) you shouldn't just blindly and uncritically believe and repeat every argument that "supports" your case.
"An important thing to remember, I think, anyway from my perspective, is live facial recognition makes no decisions. All [it] does is suggest to an officer that a person there might be wanted for GBH or attempted murder or whatever."
No but then it give the office preconceived opinion on the 'suspect' and their response to anything this 'suspect' doesn't will be representative of that. Meaning, this is a murderer that has been known to do anything to get away. So beat the shit of of him before he can try anything. Oh wait, wasn't him, lets fit him up for something else.
"live facial recognition makes no decisions. All [it] does is suggest to an officer that a person there might be wanted for GBH or attempted murder"
Does this person really think that if AFR thinks an innocent person bears a substantially close resemblance to someone wanted for attempted murder, that the local cop is going to look at them, say "Nah!" and send them on their way with a cheery "G'day"?
No, getting flagged by the AFR system will essentially decide the situation, because no cop wants to explain to his superiors how he got an alert about an attempted murderer, let the person go, and then they walked down the street and murdered someone else.
I understand the 98% false positive comment in the article but how good is this technology with actual wanted suspects. If 100 wanted people were scanned how many would this system expect to identify? Is it any more effective than a police officer looking at photos (excluding the scaling aspect)?
If your aim is to look for particular wanted people what you need is a low false negative result. The false positive result goes up as a result. Even with a 98% false positive result it's still effective.
The low false negative rate removes some of the clutter and allows a human operator to look at the potential match.
"If your aim is to look for particular wanted people what you need is a low false negative result. The false positive result goes up as a result. Even with a 98% false positive result it's still effective.
The low false negative rate removes some of the clutter and allows a human operator to look at the potential match."
Actually, I think you want a tiny false positive result and are quite happy with a large false negative result.
Almost everyone on the street isn't wanted (let's say, 1 in 10000), so if you have a false positive rate of 1% and a false negative rate of 0%, for every 100 people you stop one of them is wanted. That's a big problem. Pull the false positive rate down to 0.1% and bump the false negative rate up to 50%, and for every 20 people you stop, one is wanted.
The point is, a false negative result for a wanted person just means they will carry on walking, like they would have done without the technology. So anyone stopped that is wanted is a bonus. Furthermore, if this technology is deployed more widely, you will catch the wanted people eventually, so you can afford to miss them on the first ten traps if they get caught on the eleventh. Most wanted people are not on their way to a crime, so not stopping them immediately is not really a problem.
(I'm not saying I agree with the technology, just thinking about the statistical implications.)
The problem is the same as in disease screening. If you have a 98% effective test for a disease that only 1 in a million people have then you get that one person, but also 2% * 1 million = 20,000 false positives.
Finding one suspect in 65million Britains, or 500million if they could be anywhere in Europe, is the ultimate 'rare case'
We only have the previous trials to go on with the results analysed by the University of Essex.
Basically, for every suspect identified, 80% will be false positives... Hope you don't look like a criminal? I mean, you've probably done something wrong...
The current EU plan is privacy friendly for the next few years but establishes EU jurisdiction over the use of the tech. So the laws will be framed by the EU, not your national government. And you can't vote for anyone who will change those laws because the people who write EU legislation are not elected.
What do you think the massively centralising bureaucracy which cannot be voted out, is going to do when the "privacy-friendly" period is up? For a few years of better privacy, you'd have surrendered control forever.
I'm going to buy a hat, with a really wide brim and those dangly corks the Australians use. Maybe I'll invent a panama hat with a small semi-mirrored screen around the front. You know, because global warming.
Don't rely on the government for things you can fix for yourself.
"And you can't vote for anyone who will change those laws because the people who write EU legislation are not elected."
You mean the European Council? It consists of the 28 heads of government, each of which is democratically elected. Or you mean the European Commission? That's 28 (now 27) commissioners, each of which is appointed by their democratically elected government, in much the same way as ministers are appointed.
"You mean the European Council? It consists of the 28 heads of government, each of which is democratically elected. Or you mean the European Commission? That's 28 (now 27) commissioners, each of which is appointed by their democratically elected government, in much the same way as ministers are appointed."
And to oppose legislation, you need to demonstrate opposition across the majority of those 28 countries.
Getting significant opposition to government policies is hard in a single country - it often fails and you rely on the next election to try and enact change, assuming sufficient support. Doing that across multiple countries over a 5 year election cycle (across EU states that you are targeting) is basically impossible.
Hmmm, no, it's the European Parliament that ultimately votes legislation. It's also elected, directly.
That all sounds legitimately democratic.
If you mean that a single country's opposition should be able to block all others, then you mean that Scotland or NI should be able to stop the UK from leaving the EU.
And don't forget that there's also the EUCJ, checking the two other branches so they don't overstep. And for this case, the non-EU ECHR also could have its say. Plenty of checks and balances, all that.
"And to oppose legislation [in the UK], you need to demonstrate opposition across the majority of those 28 countries. 650 constituencies".
Congratulations, you've just found the downside of democracy, in order to get what you want, you need a majority of other people to want it too. (Although, depending on the 'democracy' we're discussing, large sums of money will do just as well).
As Churchill called it "Democracy is [..] the worst form of Government, except for all those others".
"And to oppose legislation [in the UK], you need to demonstrate opposition across the majority of those 28 countries. 650 constituencies"
My point is that democracy gets less representative and more difficult to change as you scale it up. I will happily agree that adequately representing 650 constituencies is difficult via a democracy as you will get winners and losers within that system. Just as you get winners and losers within the EU system.
My argument is whether you can democratically oppose an organisation the size of the EU. Based on the voter base and spreading elections over a number of years, I believe it to be difficult, bordering on impossible. And if you cannot enact change, is it still a democracy or just using maths to remove the outlying opinions?
"the people who write EU legislation are not elected." That is true!
But the people who ask for those laws to be written are elected representatives from the states. Then when the law has been written up, by the commissioners. It has to be approved by the elected representatives before it actually becomes an EU. law.
"A spokesman from tech supplier NEC told the briefing that its own internal trials showed a 70 per cent accuracy rate without giving details of how this was established. Independent trials, as well as the Met's own test deployments, showed an inaccuracy rate closer to 98 per cent – not that this is stopping PC Plod."
Numbers ... Are they not great ? You can send any of them without explanation and make up anything. I can probably show up a 100 % of blind guessing cards ... if the person handling the cards hints me what it is ...
"Technology moves at a frightening pace. We all know that."
Yes, mate, and some of us also know you tiny brain is now WAY behind so you cannot comprehend the implications ...
The Met are quoted as claiming that it falsely flags someone as a suspect "only one in 1000" times.
Let's say they stick one of these on each end of Westminster Bridge, for example. That apparently has foot traffic of around 40,000 people a day. That's 40 false positives each day on one bridge alone. I would imagine that's enough to keep a couple of officers busy full time. I would imagine there are places a lot busier than Westminster Bridge where they would like to stick these things as well, such as busy tube stations, public places like Leicester Square, etc. etc.
Not only useless, but costly, and invasive. Sounds like it's right up the Met's street...
It's a pity the Met don't have any jurisdiction inside parliament.
Putting these in the House of commons chamber facing the Govt. & Opposition benches with an armed response squad ready to pounce on anyone flagged would really improve viewing figures for PMQs.
Depends on how good it is at recognising people from the backs of their heads. I don't know about you, but I don't spend a lot of time looking behind me when walking through busy places. It's bad enough avoiding the people who walk straight at you as it is.
Soon to be tied to the DWP systems - you were seen on the street (and soon to be any form of public transport - can't leave that gap open) when you claimed SSP/PIP/ESA/DLA/Universal Credit - "obviously" fraud m'lud, court, convicted, jail - while your ignored for highlighting that you were taking the bus to get to the hospital for an X-ray or checking in on your sick grandmother as the judge decides that anyone well enough to leave the house is clearly a scrounger and a ne'er do well....
You should probably visit still, I reckon they will deploy these initially away from tourist areas and more in the east end and places where the stabbing hotspots are.
After a few years of bedding them in there, they will move them to tourist places.
I would also say, these cameras wont stop anyone who has yet to commit a crime but plans on committing one, aka most lone wolfs
Apparently, London buses nowadays are bristling with video cameras. Can't remember the exact number but I think it's around eighteen. There are some on the top deck and lower deck looking at the passengers and driver. All the rest face outwards and are used if the bus is involved in an accident with a pedestrian/cyclist/motorist.
If plod want to look at a disturbence, or whatever, in an area not covered by their static cameras they just grab all of the footage from passing buses.
Which means that the central premise of "The Capture" on the BBC last year was all a load of old tut.
Shame really, I thought it was going to be a drama about the dangers of audio PTS drift.
Although I agree with your sentiment, because like most readers here I have at least a basic understanding of statistics and a reasonable understanding of the state of facial recognition technology, I suspect the majority of the population have little idea about either and thing what they see in TV crime shows is all true and therefore this is a "good" thing "because it'll help catch criminals, innit"
> He continued: "An important thing to remember, I think, anyway from my perspective, is live facial recognition makes no decisions. All [it] does is suggest to an officer that a person there might be wanted for GBH or attempted murder or whatever."
How many people, whose picture is already on record, are currently wanted for GBH or murder and are at large? The CrimeStoppers website currently lists 7 wanted for murder; 2 for attempted murder and 6 for GBH/wounding.
So, to make it slightly easier for the police to catch just 15 people, the entire UK population needs to be put under surveillance.
Wishy Washy "consultation" that only asked how much further people thought they should go, SPA rubbished Openrightsgroups concerns, SNP offers a "biometrics commissioner" who like the rest will be a party lackey and happily rubber stamp anything or face being replaced with someone more willing.....
"Live Facial Recognition is a fantastic crime-fighting tool," an enthusiastic assistant commissioner Nick Ephgrave told the nation's press (and El Reg) this morning, comparing it to "showing photographs to [constables] before they go out on patrol".
More like demanding citizens present themselves to the authorities multiple times to a day to ensure they haven't done anything wrong......yet.
What's next? Crime swoops?
Oh!....no mention of whether the Met is ALSO using Clearview AI:
- https://www.nytimes.com/2020/01/18/technology/clearview-privacy-facial-recognition.html?action=click&module=Top%20Stories&pgtype=Homepage
*
Billions of identified Facebook pix....and plod only has to upload a CCTV image to see if there's a match. Secret technology, NO consent provided by the individuals whose images have been hoovered up, cheap for the plod......and not in EITHER THE UK OR THE EU, so safe from the Supreme Court and ECJ. What's not to like?
*
Also (off topic)....the GDPR legislation is a HOAX!!! How can I ask for my personal details to be deleted....when I HAVE NO IDEA WHO OWNS DATABASES WITH MY PERSONAL DETAILS????
"What about my right to privacy?"
TL;DR - in public you don't have any privacy.
The general argument is that if you are out in 'public' spaces (ie. walking the streets, in a shopping mall, at the cinema, eating out, etc) then you have not expectation of privacy, as anyone and everyone is able to see you.
While you are quite free to pull your jumper over your head, wear a mask or do other things to obscure your face in doing so you inadvertently bring attention to yourself. AFAIK it is not illegal to obscure your face unless the police tell you to identify yourself (usually that involves comparing some form of photo identification against your actual face), so you may run into issues there. Also were you to obscure your face around anywhere there are valuables (say a bank, jewelry store, etc) then it's likely, again, you would quickly bring attention to yourself. Having said all that I note that some Muslim women get around quite fine with full face veils and can say 'don't discriminate against my religion' if challenged. Not sure how they go around banks, jewelry stores, etc.
There are methods to confuse facial recognition systems using clothing, but I believe these only work on against specific algorithms, so you would have to know what you're up against and dress accordingly (if anything had even been designed to confuse the systems in your local area) and then you'd have to keep on top of any changes that might be made to said algorithm(s).
https://www.businessinsider.com/clothes-accessories-that-outsmart-facial-recognition-tech-2019-10
"An important thing to remember, I think, anyway from my perspective, is live facial recognition makes no decisions.
Yes you are correct it doesn't make decisions but decisions are made based on the information given, information that has been showed to be incorrect an alarming number of times.
If your mug is not matched by the system, according to Ephgrave, the images taken and the processing done will be deleted immediately.
Are you sure this is the case? sure you don't want to double check that first? We wouldn't want it to could back and bite you.
can someone clarify the law. is it a crime in the UK to take a (camera phone, for eg) pic of a policeman in a public place?
IANAL but as far as I am aware It is not a criminal offence to take a picture of anyone or anything if you're doing so from a public place. (With caveats about creepy stuff through bedroom windows etc.)
However I wouldn't recommend it as the police in the UK routinely arrest people and/or steal or vandalise their cameras for perfectly legal photography of policemen or what they deem to be sensitive sites.
Some determined victims occasionally manage to get a successful prosecution but it usually just results in an empty demand from the judge for the force to train its staff better and maybe a fine or a tiny bit of compensation - which is payed using our money of course - but individual officers almost never get even a reprimand, let alone dismissal or personal prosecution for theft, damage or unlawful imprisonment.
(I'm sure most coppers are reasonable people but it sometimes seems that they are more likely to be subject to dismissal or prosecution for ratting out their colleagues' illegal behaviour or for refusing to break the law when ordered to by their superiors than for actual misbehaviour.)
yeah the specific thing I might be thinking about is "do you have to give the police your details in the street" or something. I've seen footage of people refusing to give information that I'm assuming the policemanwoman they were dealing with, were "hoping" the member of the public wouldn't be aware that they didn't have to provide. eg, "what's your name" "I don't have to give you my name", or something like that
One thing strike me: a passer-by was fined in the UK for not wanting to show his face to the camera.
For all the media scares, in China, it is perfectly legal and allowed to wear a face mask, even now mandatory for health reasons.
So they actually seem much *less* hell-bent on, specifically, AFR surveillance than the UK. Fancy that.
So we've had quite a few high profile terrorist attacks in or around Parliament.
And we do need to protect that bastion of Democracy.
So deploy it there.
Allow the MPs to bask in the luxurious safety net feel of facial recognition security and that any one flagged up is most definitely a terrorist/criminal.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
