For clarity, this is a general surveillance facial recognition project that was started only a few months ago (with security cameras).
The automated passport gates (for Belgian eIDs and EU biometric passports) are not included or affected and will keep working (when they're not broken or closed).
Here in the UK we have cameras that will automatically scan car number plates to check for insurance and road tax.
In principle I would see nothing wrong with having the same with facial recognition for wanted people (frankly, why would this not exist if technology is available?), though I'm guessing that people on the run would quickly learn to avoid them.
I feel we are in a phase of facial recognition development in which some people reject it as a matter of principle or because they only see the worst that might be done with it. I'm hoping for a more balanced approach (not holding my breath...).
Car number plates have a consistent, mandated font with a set of regular formats on high-contrast backgrounds that makes it almost trivially easy for computer vision to extract a number plate from an image of a car.
Cars are licenced transport. You are required to a) tax and maintain your vehicle b) identify who was driving the vehicle if you are the registered owner and requested to do so by the police and c) not obscure or obfuscate your number plate.
Faces are not licenced, are uniquely identifiable personal data - ignoring the enormous false positive detection rate from current technology - and everyone has one. Consider the implications for a supposedly free society (of which the United Kingdom no longer is one) where the state can draw arbitrary inferences from the presence of someone with a face that matches pattern X in location Y on date Z.
Your reply illustrates my comment that some people only see the worst or try to imagine only the worst.
If a facial recognition system is only linked to a database of wanted people then it won't be able to ID people not on that database. I.e. if it is no worse than a normal CCTV for everyone who is not wanted by the police.
Therefore, in principle there is no negative impacts in terms of privacy or liberty compared to existing CCTVs.
The only issue is technical: False positives (which I'm not sure the numbers would be 'enormous' at the moment). This is bound to improve over time and can be mitigated with the appropriate police procedures on how to respond to an alert by such a system.
> for a supposedly free society (of which the United Kingdom no longer is one)
Of course if that's your opinion we're not going to be able to have a discussion on the topic...
Your assumption that the system throws away everything but the images of flagged people and number plates, and expires even those after some period of time, flies completely in the face of reality (e.g. the New York State Police were keeping all their collected ANPR data indefinitely [0]). A number of US states are making legal provisions to restrict how long ANPR data is retained for [2] - are they just tech pessimists too?
> False positives (which I'm not sure the numbers would be 'enormous' at the moment)
A 98% false positive rate [1] seems quite enormous.
> The Metropolitan Police’s system has produced 104 alerts of which only two were later confirmed to be positive matches
In another deployment:
> overall only 234 alerts – fewer than 10 per cent – were correct matches.
At least in the UK you're not going to be stopped by a police officer with a gun drawn on the basis of a failed read [3].
In addition to false positives, the false negative rate must also be given. For example: if 100 "suspect cars" pass by, but only 5 are identified as suspect, then 95% of the suspects are still on the loose and the system is again much less effective.
This FOIA response [1] reckons that there was a test where the false-negative rate was 20%, but there's also this report [2] of a 30% false negative rate at the Notting Hill carnival. So to continue your example, out of 100 known wanted people walking past the system 20-30 got through without being flagged.
You're not contradicting my points, you are giving anecdotal examples of problems and bad implementations. The only assumption I made is that we can have such systems implemented in a way that is properly controlled and harmless.
>> If a facial recognition system is only linked to a database of wanted people then it won't be able to ID people not on that database. I.e. if it is no worse than a normal CCTV
Yes, and it would still be able to match other possible suspects purely because they are on the wanted list.
Moreover, you are matching against _actual_ people meaning that whenever the system makes a mistake you are still arresting someone who isn’t wanted.
As points of fact, not everyone has a face, and it has not yet been proven that they are uniquely identifiable.
Never mind the possibility of Hollywood-quality fakes.
I have occasionally considered starting a religion with a tenet of wearing uniform clothing (subject to weather) and identical masks whenever out in public. There's ample precedent from other religions. I figure it might even be profitable, from selling the "official" uniforms and masks, and from the overpriced cafe in the meetings hall.
99.99% of people have a face. You don't make laws or social rules for the tiny minority, you make them for the vast majority. Exceptions and caveats to the laws/rules to protect the interests of the minority are fine. But the general structure should suit the needs of the general populace.
Sending people fines in the mail the nanosecond their stuff expires sounds pretty darn dystopian to me. I know most here are well off enough that they've never had to <clutches pearls> drive without all their papers in order but it happens.
> Sending people fines in the mail the nanosecond their stuff expires sounds pretty darn dystopian to me.
That's the law with or without special cameras that are only there to enforce the law.
If you get caught by police instead you will also get a fine and they will seize your car as a bonus. If you get caught by those camera vans while parked they will clamp the car.
Likewise, if you are wanted you are liable to be detained the instant police recognises you. It's just that the process is very inefficient at the moment because it relies on being recognised by police. In practice it means that they have to specifically be looking for you.
If you get caught by the police you have a human to reason with. "Listen officer, I have all the paperwork right here I simply haven't gotten it to the DMV yet because today is a weekend and they are only open weekdays".
I don't live in western Europe (thank god) so the police here don't impound people's means of personal transportation at the drop of a hat so you get the same fine the camera would have handed out with the opportunity to get out of it if you are genuinely being reasonable.
Well, if that's your argument again automated systems I think it means that these systems are quick ok...
Also, we don't need to send any paperwork. For road tax they send a reminder a month in advance and you just need to pay online. Same for insurance (there really is no excuse for driving without insurance).
> If you get caught by the police you have a human to reason with. "Listen officer, I have all the paperwork right here I simply haven't gotten it to the DMV yet because today is a weekend and they are only open weekdays".
These sort of stuff tends to discriminate against certain segment of societies, like you are let go if you are white and fined if you aren't
OK, I'll bite and have a balanced discussion with you. Could you explain more about what you would like from such a system? Where do you see it deployed? Airports, public areas, streets, shops, everywhere?
The article at the Flemish tv [0] has a little more background (also in Dutch, but often they create specific English versions for this kind of articles, which can take some time [1]). They mention 4 reasons for the forced stop:
- There was no prior study about possible the privacy implications.
- No approval was asked before the project started although such an approval is mandatory under GDPR.
- Some changes in law are needed to allow to keep such a database. Current laws only permit licence plate databases.
The article is about the airport, not about the town. The airport is often called Zaventem because of its location. International travelers will know it by the name Brussels airport.
30 comments
[ 4.8 ms ] story [ 71.8 ms ] threadEdit: also don't bother translating or reading the comments on that website. They are youtube-comment quality.
edit: the cameras aren't being banned, the facial recognition software must be disabled
The automated passport gates (for Belgian eIDs and EU biometric passports) are not included or affected and will keep working (when they're not broken or closed).
In principle I would see nothing wrong with having the same with facial recognition for wanted people (frankly, why would this not exist if technology is available?), though I'm guessing that people on the run would quickly learn to avoid them.
I feel we are in a phase of facial recognition development in which some people reject it as a matter of principle or because they only see the worst that might be done with it. I'm hoping for a more balanced approach (not holding my breath...).
Cars are licenced transport. You are required to a) tax and maintain your vehicle b) identify who was driving the vehicle if you are the registered owner and requested to do so by the police and c) not obscure or obfuscate your number plate.
Faces are not licenced, are uniquely identifiable personal data - ignoring the enormous false positive detection rate from current technology - and everyone has one. Consider the implications for a supposedly free society (of which the United Kingdom no longer is one) where the state can draw arbitrary inferences from the presence of someone with a face that matches pattern X in location Y on date Z.
If a facial recognition system is only linked to a database of wanted people then it won't be able to ID people not on that database. I.e. if it is no worse than a normal CCTV for everyone who is not wanted by the police.
Therefore, in principle there is no negative impacts in terms of privacy or liberty compared to existing CCTVs.
The only issue is technical: False positives (which I'm not sure the numbers would be 'enormous' at the moment). This is bound to improve over time and can be mitigated with the appropriate police procedures on how to respond to an alert by such a system.
> for a supposedly free society (of which the United Kingdom no longer is one)
Of course if that's your opinion we're not going to be able to have a discussion on the topic...
> False positives (which I'm not sure the numbers would be 'enormous' at the moment)
A 98% false positive rate [1] seems quite enormous.
> The Metropolitan Police’s system has produced 104 alerts of which only two were later confirmed to be positive matches
In another deployment:
> overall only 234 alerts – fewer than 10 per cent – were correct matches.
At least in the UK you're not going to be stopped by a police officer with a gun drawn on the basis of a failed read [3].
0: https://arstechnica.com/tech-policy/2012/09/your-car-tracked...
1: https://www.independent.co.uk/news/uk/home-news/met-police-f...
2: https://arstechnica.com/tech-policy/2015/03/virginia-passes-...
3: https://arstechnica.com/tech-policy/2014/04/due-to-license-p...
1: https://www.whatdotheyknow.com/cy/request/use_of_automated_f...
2: https://theconversation.com/why-regulating-facial-recognitio...
Yes, and it would still be able to match other possible suspects purely because they are on the wanted list.
Moreover, you are matching against _actual_ people meaning that whenever the system makes a mistake you are still arresting someone who isn’t wanted.
Never mind the possibility of Hollywood-quality fakes.
I have occasionally considered starting a religion with a tenet of wearing uniform clothing (subject to weather) and identical masks whenever out in public. There's ample precedent from other religions. I figure it might even be profitable, from selling the "official" uniforms and masks, and from the overpriced cafe in the meetings hall.
That's the law with or without special cameras that are only there to enforce the law.
If you get caught by police instead you will also get a fine and they will seize your car as a bonus. If you get caught by those camera vans while parked they will clamp the car.
Likewise, if you are wanted you are liable to be detained the instant police recognises you. It's just that the process is very inefficient at the moment because it relies on being recognised by police. In practice it means that they have to specifically be looking for you.
I don't live in western Europe (thank god) so the police here don't impound people's means of personal transportation at the drop of a hat so you get the same fine the camera would have handed out with the opportunity to get out of it if you are genuinely being reasonable.
Also, we don't need to send any paperwork. For road tax they send a reminder a month in advance and you just need to pay online. Same for insurance (there really is no excuse for driving without insurance).
These sort of stuff tends to discriminate against certain segment of societies, like you are let go if you are white and fined if you aren't
I don't understand how you could have interpreted my comment as suggesting otherwise...
- There was no prior study about possible the privacy implications.
- No approval was asked before the project started although such an approval is mandatory under GDPR.
- Some changes in law are needed to allow to keep such a database. Current laws only permit licence plate databases.
- There were too many false positives.
[0] https://www.vrt.be/vrtnws/nl/2019/09/20/politie-mag-geen-aut...
[1] https://www.vrt.be/vrtnws/en/
https://hastebin.com/raw/wiwerobayi