Army Rumour Service

Register a free account today to become a member! Once signed in, you'll be able to participate on this site by adding your own topics and posts, as well as connect with other members through your own private inbox!

Is Facial Recognition & Public Surveillance A Good Thing?

Chef

LE
If it happened once every 10 years, yes I would.
Do you think all ATMs should be banned on the off chance that occasionally one of them doesn't work correctly?

Fair enough, of course if the £20 is the start of a run of mistakes then at the least of it you'd be £20 down before you started taking action.

As to banning ATMs a bit of an extreme reaction. The banks are quietly removing them anyway leaving only the ones which charge for their use. I expect you're happy with that.

Given the way that cash is, quite literally, going the question will soon be redundant and we'll be relying on the integrity of the banks not to rip off the consumer more than they do now.

That aside I'm sure that facial recognition will be fine and that if it ever gets to 99% accuracy the 1%ers who get minorly inconvenienced will also write it off to experience.
 
Fair enough, of course if the £20 is the start of a run of mistakes then at the least of it you'd be £20 down before you started taking action.

As to banning ATMs a bit of an extreme reaction. The banks are quietly removing them anyway leaving only the ones which charge for their use. I expect you're happy with that.

Given the way that cash is, quite literally, going the question will soon be redundant and we'll be relying on the integrity of the banks not to rip off the consumer more than they do now.

That aside I'm sure that facial recognition will be fine and that if it ever gets to 99% accuracy the 1%ers who get minorly inconvenienced will also write it off to experience.
Im sure they will eventully, like most technology, from contactless cards to paying over the internet eventully the luddites decide that it benefits them more than it inconviences them and accept it.
 
Fair enough, of course if the £20 is the start of a run of mistakes then at the least of it you'd be £20 down before you started taking action.

As to banning ATMs a bit of an extreme reaction. The banks are quietly removing them anyway leaving only the ones which charge for their use. I expect you're happy with that.

Given the way that cash is, quite literally, going the question will soon be redundant and we'll be relying on the integrity of the banks not to rip off the consumer more than they do now.

That aside I'm sure that facial recognition will be fine and that if it ever gets to 99% accuracy the 1%ers who get minorly inconvenienced will also write it off to experience.
You might want to have a look at The UK's ATM network
and https://www.psr.org.uk/sites/default/files/media/PDF/Protecting_free_to_use_ATMs.pdf
 

Chef

LE
Im sure they will eventully, like most technology, from contactless cards to paying over the internet eventually the luddites decide that it benefits them more than it inconviences them and accept it.

I've no doubt you're right on the matter. Scant comfort when one of the banks/building societies had a computer glitch that stopped their customers accessing their accounts for a few days.

To quote the Jam.

'The public wants what the public gets.'
 
Im sure they will eventully, like most technology, from contactless cards to paying over the internet eventully the luddites decide that it benefits them more than it inconviences them and accept it.
That was a very short lived, but I’m sure significant revenue stream for the banks. I wonder what else they will sting us with to compensate.
 

Cutaway

LE
Kit Reviewer
'The public wants what the public gets.'
Pubs and Wormwood Scrubs and too many right wing meetings or Eton Rifles ?
 

Cutaway

LE
Kit Reviewer
An interesting development for anyone who thinks that FR is still unreliable:

Copied from Quora

In the last several weeks there have been peaceful mass protests on the streets and squares of Belorus against President Alexander Lukashenko. A large number of security forces have been deployed. And they have not been so peaceful. They have been beating up demonstrators, shooting at them with tear gas, pulling people out of crowds, hauling into vans and taking them away to jail.


With their faces covered with masks, members of the security forces have done their best to remain incognito. Belarus is a small country, and a person you are dragging across the street might be your neighbour.
And then Artificial Intelligence came to help demonstrators. A group of Belarusian programmers said they have created a computer program that can recognize faces of the security forces even when they are wearing masks.


The programmers made a special video showing how this technology works.

“You will pay for your illegal actions, no matter how many socks you wrap around your head,” said the author of the video, Andrei Maksimov. “This is your last chance to quit,” he warned.

The video has already received over half a million views on YouTube.
According to various sources, the level of the program’s correct face recognition is between 70% and 90%.




 
An interesting development for anyone who thinks that FR is still unreliable:

Copied from Quora

In the last several weeks there have been peaceful mass protests on the streets and squares of Belorus against President Alexander Lukashenko. A large number of security forces have been deployed. And they have not been so peaceful. They have been beating up demonstrators, shooting at them with tear gas, pulling people out of crowds, hauling into vans and taking them away to jail.


With their faces covered with masks, members of the security forces have done their best to remain incognito. Belarus is a small country, and a person you are dragging across the street might be your neighbour.
And then Artificial Intelligence came to help demonstrators. A group of Belarusian programmers said they have created a computer program that can recognize faces of the security forces even when they are wearing masks.


The programmers made a special video showing how this technology works.

“You will pay for your illegal actions, no matter how many socks you wrap around your head,” said the author of the video, Andrei Maksimov. “This is your last chance to quit,” he warned.

The video has already received over half a million views on YouTube.
According to various sources, the level of the program’s correct face recognition is between 70% and 90%.




Seventy to 90 percent is unreliable. Let's put it this way, if you had a 30% chance of being identified as a wanted rapist every time you walked past a camera, you would be arrested multiple times per day in a place like London.

I'm assuming of course that you're not a wanted rapist, if you are, then pick some other crime that you're not wanted for (assuming there are any).
 
Seventy to 90 percent is unreliable. Let's put it this way, if you had a 30% chance of being identified as a wanted rapist every time you walked past a camera, you would be arrested multiple times per day in a place like London.

I'm assuming of course that you're not a wanted rapist, if you are, then pick some other crime that you're not wanted for (assuming there are any).
Much as I agree with your take on the numbers I believe that those figures are based upon recognition of someone wearing a balaclava .

In the wider world:

In ideal conditions, facial recognition systems can have near-perfect accuracy. Verification algorithms used to match subjects to clear reference images (like a passport photo or mugshot) can achieve accuracy scores as high as 99.97% on standard assessments like NIST’s Facial Recognition Vendor Test (FRVT).

.

The GaussianFace algorithm developed in 2014 by researchers at The Chinese University of Hong Kong achieved facial identification scores of 98.52% compared with the 97.53% achieved by humans. An excellent rating, despite weaknesses regarding memory capacity required and calculation times.

 
Much as I agree with your take on the numbers I believe that those figures are based upon recognition of someone wearing a balaclava .
Some facial recognition systems look mainly at the area around the eyes anyway, and that example had a lot of bare skin showing around the eyes. If they used a system that worked that way, then we shouldn't be surprised that it worked at least so-so when used in the manner intended.

In the wider world:

In ideal conditions, facial recognition systems can have near-perfect accuracy. Verification algorithms used to match subjects to clear reference images (like a passport photo or mugshot) can achieve accuracy scores as high as 99.97% on standard assessments like NIST’s Facial Recognition Vendor Test (FRVT).

.

The GaussianFace algorithm developed in 2014 by researchers at The Chinese University of Hong Kong achieved facial identification scores of 98.52% compared with the 97.53% achieved by humans. An excellent rating, despite weaknesses regarding memory capacity required and calculation times.

Nobody has said that facial recognition systems can't recognise faces. The problem comes in how you try to use them.

For example, suppose you had your front door set up to recognise your face and greet you upon your arrival home. It only has to do a very limited number of compare operations, that of one reference face to a limited number of people arriving at the door. A 98.52% recognition rate would be more than satisfactory.

Now suppose you had every surveillance camera in the UK comparing every face it saw to a database of every person living in the UK. That would be many millions of compare operations per day. A 98.52% success rate would be an abysmal failure.

That humans would have only a 97.53% success rate is irrelevant. There aren't armies of people monitoring each and every camera and comparing each face that passes before the camera to a photo of everyone in the UK.

So it all comes back to "what are you trying to do with it?" The proposals for mass surveillance of society quickly run up against the basic math of very large numbers - high success rates are still poor results if you are talking about large numbers of samples.

Fingerprint readers have the same basic limitations, as do many, if not all, biometric sensors. They can be useful when used within very narrow constraints, but fail when used outside of their limitations. The key is to identify their limitations and ensure that your use case is carefully constrained to operate within them.
 
Some facial recognition systems look mainly at the area around the eyes anyway, and that example had a lot of bare skin showing around the eyes. If they used a system that worked that way, then we shouldn't be surprised that it worked at least so-so when used in the manner intended.


Nobody has said that facial recognition systems can't recognise faces. The problem comes in how you try to use them.

For example, suppose you had your front door set up to recognise your face and greet you upon your arrival home. It only has to do a very limited number of compare operations, that of one reference face to a limited number of people arriving at the door. A 98.52% recognition rate would be more than satisfactory.

Now suppose you had every surveillance camera in the UK comparing every face it saw to a database of every person living in the UK. That would be many millions of compare operations per day. A 98.52% success rate would be an abysmal failure.

That humans would have only a 97.53% success rate is irrelevant. There aren't armies of people monitoring each and every camera and comparing each face that passes before the camera to a photo of everyone in the UK.

So it all comes back to "what are you trying to do with it?" The proposals for mass surveillance of society quickly run up against the basic math of very large numbers - high success rates are still poor results if you are talking about large numbers of samples.

Fingerprint readers have the same basic limitations, as do many, if not all, biometric sensors. They can be useful when used within very narrow constraints, but fail when used outside of their limitations. The key is to identify their limitations and ensure that your use case is carefully constrained to operate within them.
The essence of the argument is, are you willing to be serially inconvenienced while the authorities take however much time they like to find out you are not their person of interest.
 
The essence of the argument is, are you willing to be serially inconvenienced while the authorities take however much time they like to find out you are not their person of interest.
You get to spend time in jail, lose time and income from work (if you don't just lose your job), and spend large wads of money you don't have on lawyers. And once you get out you get to do it all over again the next time you pass in front of a camera. Yes, that sort of inconvenience.

As I've said the issue isn't facial recognition itself but rather where it will fit into a policing system that has become obsessed with metrics, cost savings, and micro-management of policemen and women.
 
An interesting development for anyone who thinks that FR is still unreliable:

Copied from Quora

In the last several weeks there have been peaceful mass protests on the streets and squares of Belorus against President Alexander Lukashenko. A large number of security forces have been deployed. And they have not been so peaceful. They have been beating up demonstrators, shooting at them with tear gas, pulling people out of crowds, hauling into vans and taking them away to jail.


With their faces covered with masks, members of the security forces have done their best to remain incognito. Belarus is a small country, and a person you are dragging across the street might be your neighbour.
And then Artificial Intelligence came to help demonstrators. A group of Belarusian programmers said they have created a computer program that can recognize faces of the security forces even when they are wearing masks.


The programmers made a special video showing how this technology works.

“You will pay for your illegal actions, no matter how many socks you wrap around your head,” said the author of the video, Andrei Maksimov. “This is your last chance to quit,” he warned.

The video has already received over half a million views on YouTube.
According to various sources, the level of the program’s correct face recognition is between 70% and 90%.






Its a long time since I have been at university, but the people I was sharing research with back in the late 90's and early 00's were the ones developing this kind of stuff. Back in the late 90's we were already looking at facial recognition and AI based fingerprint recognition systems that researchers were putting out there FOC into the community for interested parties - a bit like LINUX.

What oiky sociopaths like Lukashenko, and others, need to remember is that the tech used by their NKVD para-military police and intelligence services has always grown out of research in a university. Start shitting on people in universities and they will use that tech against you - and their tech is probably 2 generations newer than the govt. tech. You should see some of the micro drone tech that was coming out of MIT 10 to 20 years ago, what you see available nowadays is still not where the research state of the art is quietly sitting. Wait till they start using that against oppressive state actors, quietly tag senior riot police officers and let the drone follow them home......scary, and possible.

I remember doing a presentation to senior officers in a Police Force in the UK back in around '98 to develop an AI based predictive modelling system for crime overlaid onto a mapping display. An absolute piece of piss for me to build and populate as long as I had access to the data and I could link into their db to download daily crime information. I was more or less received as if I was trying to sell them a cauldron, magic wand and a ouija board. Nowadays they pay a fortune for the same stuff whereas they could of had it developed FOC eons ago.
 
Last edited:
I agree. The tech is far better than most people realise.
 

New Posts

Latest Threads

Top