Schneier on Security
A blog covering security and security technology.
« Friday Squid Blogging: Squid Police |
| Matthew Weigman »
August 31, 2009
On London's Surveillance Cameras
A recent report has concluded that the London's surveillance cameras have solved one crime per thousand cameras per year.
David Davis MP, the former shadow home secretary, said: "It should provoke a long overdue rethink on where the crime prevention budget is being spent."
He added: "CCTV leads to massive expense and minimum effectiveness.
"It creates a huge intrusion on privacy, yet provides little or no improvement in security.
Earlier this year separate research commissioned by the Home Office suggested that the cameras had done virtually nothing to cut crime, but were most effective in preventing vehicle crimes in car parks.
A report by a House of Lords committee also said that £500 million was spent on new cameras in the 10 years to 2006, money which could have been spent on street lighting or neighbourhood crime prevention initiatives.
A large proportion of the cash has been In London, where an estimated £200 million so far has been spent on the cameras. This suggests that each crime has cost £20,000 to detect.
I haven't seen the report, but I know it's hard to figure out when a crime has been "solved" by a surveillance camera. To me, the crime has to have been unsolvable without the cameras. Repeatedly I see pro-camera lobbyists pointing to the surveillance-camera images that identified the 7/7 London Transport bombers, but it is obvious that they would have been identified even without the cameras.
And it would really help my understanding of that £20,000 figure (I assume it is calculated from £200 million for the cameras times 1 in 1000 cameras used to solve a crime per year divided by ten years) if I knew what sorts of crimes the cameras "solved." If the £200 million solved 10,000 murders, it might very well be a good security trade-off. But my guess is that most of the crimes were of a much lower level.
Cameras are largely security theater:
A Home Office spokeswoman said CCTVs "help communities feel safer".
Posted on August 31, 2009 at 5:59 AM
• 88 Comments
To receive these entries once a month by e-mail, sign up for the Crypto-Gram Newsletter.
For me there is one convincing argument against all those cameras. You can find it at http://en.wikipedia.org/wiki/...
The picture, taken by a surveillance camera, shows Mohammed Atta boarding his 9/11 flight. The camera prevented nothing.
1 crime:1000 cameras / year (solved, not prevented)? That's a lot higher than I would have guessed, maybe they do have some value after all.
What do the X-10 people say about this?
This reminds me of the TV series "Homeland Security USA". I like watching it; it presents some things I never thought about, and it shows civil servants working in security while treating people with respect. Its just a shame that with 50,000+ people working in the area 24x7 they can only find 1 hour or so over the course of a week of the govt treating people with respect.
> To me, the crime has to have been
> unsolvable without the cameras.
I'm no CCTV advocate (quite the opposite in fact) but I think that cameras would be justified even if they helped the police in a meaningful and time-saving fashion.
Then we could argue that the money spent were useful in freeing policemen's time, so that they could concentrate in other work.
But my guess is that the cameras (at least in the way they're currently used) are not even close to achieving this.
Later in the same article, the police spokesperson says that 70% of solved murders are solved with CCTV footage. Based on the population of London and the UK murder rate (assuming it is the same for London, although in fact it is probably a little higher), this would be about 1100 murders.
For obvious reasons, it makes sense that CCTV would contribute disproportionately to the solution of murders, and for much the same reason that it is particularly effective in deterring crime in car parks.
"I'm no CCTV advocate (quite the opposite in fact) but I think that cameras would be justified even if they helped the police in a meaningful and time-saving fashion. Then we could argue that the money spent were useful in freeing policemen's time, so that they could concentrate in other work."
It depends on how much time is saved, and what the freedoms sacraficed for saving that time is worth.
How much does a policeman's time cost? £200 an hour? If the cameras save at least 100 hours of police work per crime solved -- and waste zero hours of police work for all those crimes not solved -- then the only cost is the freedom and liberty. But I'll bet the cameras aren't worth it purely on a financial analysis.
Additionally the claims seem difficult to substantiate given that we don't actually know how many cameras are in use in London today:
"Yet it has emerged that one of the figures most frequently used to justify the claim — that there are 4.2 million cameras — is an estimate based on a survey of only two streets in London seven years ago."
From the article:
"...most serious crime investigations have a CCTV investigation strategy."
Yeah, I have a strategy all mapped out for when I win the lottery!
The statistic "70% of solved murders are solved with CCTV footage" is meaningless, since it fails to give the details on how many of these murders wouldn't have been solved without the use of the CCTV material?
In the same fashion, they could proclaim that 99% of all murders were solved with ball pens and/or pencils.
Another observation: when people want to emphasise (or exaggerate?) the size of government expenditure, they tend to aggregate it over a decade for no apparent reason other than that it makes the numbers ten times bigger.
£500 million over ten years averages £50 million per annum, which in fact is a very small sum for a country of 61 million people. The national police budget is £10,500 million per annum, or more than 200 times what is spent on CCTV.
"...money which could have been spent on street lighting ..."
(Which is known to be even less effective than cameras, and in some cases actually increases crime.)
"... or neighbourhood crime prevention initiatives. "
Initiatives, huh? i.e. "erm, ah, stuff, thingamajigs, we'll think of the details later, but it will be about neighbourhoods, so it will be good, m'kay?" I think this is what is known as a "glittering generality."
Yeah, no-one in the UK is really that bothered by CCTV for obvious reasons. None of the footage is really watched and it is usally very poor quality.
I think the real problem will come when large networks (e.g. the police CCTV network - you couldn't really do it with in-store CCTV) are connected with people-tracking software.
Ironically enough, there's little evidence that communities "feel safer", and quite a bit that they hate the things - CCTV is seen as being nosy, which British people don't like.
Unfortunately it's the sort of "don't like" that translates to "don't invite the nosy git round for lunch", rather than "do something to get rid of them".
"For every 1,000 cameras in London, less than one crime is solved per year"
I know it's harder to measure, but the article focuses on solving crime and pretty much discounts the deterrent effect — and no matter how inefficient the use of CCTVs might be, it _has_ to be a deterrent. Hasn't it?
If cameras do simply "help communities feel safer" that might be a sufficient good in itself, of course. Crime in the UK has fallen massively over the last 15 years or so, with the risk of someone being a victim of crime at an all-time low since the British Crime Survey began in, I think, the 1970s. But people still feel unsafe.
Obviously that suggests that cameras aren't doing their job, if it's to "help communities feel safer", but it also means that that is a reasonable aim - the disparity between actual and perceived risk isn't good for society.
I remember a while back reading a post by Bruce about a guy rounding a corner, snatching a woman's bag and all being caught clearly on CCTV. The comment was: no amount of argumentation, statistics and logic can lessen the effect such a video has on people trusting CCTV.
The same still applies. Humans are what humans will be, with all our instinctive fears and trust. Men sometimes attempt to over-ride them using logic. Women often not only fail to understand logic; they rely strongly on feelings and experiences, giving much lesser importance to 'what should be' according to logic. Interestingly, when it comes to head-to-head discussions between the two groups, the emotional based arguments tend to be aired stronger, eventually getting the upper hand.
I do not intend this comment to be negative on either sex; logic does at times drag you to the wrong conclusions, especially when dealing with human behaviour, at which women are therefore naturally better at. I say naturally because I believe these differences in the way men and women think buds from differing mind structures and compositions.
NB - my comments are not based on fact, but purely on my perceptions.
Hey Bruce, you should create some kind of a web form for telling you about interesting news. I was gonna mail you a heads up about this a week ago but got lazy in the process :)
"...the only cost is the freedom and liberty. But I'll bet the cameras aren't worth it purely on a financial analysis."
See, that's because you put too high a value on freedom and liberty.
As Bruce says, for cameras to "solve a crime", the crime has to have been unsolvable without the cameras. The number of such crimes is difficult to estimate.
Pro-surveillance camera lobbyists will also argue that cameras have a deterrent effect, and that they help *prevent* some crimes. That too, is difficult to estimate because you cannot measure something that doesn't happen. Not to mention, for a street camera to be an effective deterrent, it has to prevent a crime altogether, not just push it to another area.
"If cameras do simply 'help communities feel safe' that might be a sufficient good in itself, of course."
It might. Depends on the cost.
Personally, I like my security theater to be as cheap as possible. Do non-working cameras help communities feel just as safe? Can we make a community feel safer at the same cost by installing 3x non-functional cameras? Or by hiring x/10 actors in police uniforms to walk up and down the streets?
@Andre LePlume "See, that's because you put too high a value on freedom and liberty."
Right. While we're at it, democracy and peace are overrated, too.
My thoughts exactly--the deterrent benefits are absent from the discussion. I also wonder about the usefulness of such cameras in non-routine crisis situations that don't really fall neatly into normal statistics.
I question the deterrent effect. If the cameras are only reviewed after the crime, why would a criminal alter his behavior? There have been cameras at ATMs and banks for years, and they still get robbed. I live in a country with the death penalty and still crimes are committed. The only deterrent to crime is to have better options. (Of course, there are some who view the crime itself as the goal, so I don't address them.)
The average bag snatcher knows that with a little disguise, he can easily grab a purse and get away. The camera doesn't prevent this kind of crime, no one is there waiting to make an arrest, and he will be long gone by the time any response can take place.
Criminals are not on average known for performing cost/benefit analysis, but at least for crimes that have been planned in advance, I have to believe that the likelihood of being caught is considered. If a crime is "solved" using CCTV, that means a crime was committed. Is the intent of CCTV to satisfy the public's desire for retribution, or to prevent crimes?
The percentage of crimes "solved" using CCTV is irrelevant if criminals do not perceive an increase in the likelihood of capture. What should be of interest is the effect of cameras as a deterrent, i.e. the crime rate. Such a measure is proven to be strongly tied to factors such as employment rates and other socioeconomic status. Maybe the money should be spent in providing alternatives to crime instead of a tool to be used after the fact?
"Do non-working cameras help communities feel just as safe? Can we make a community feel safer at the same cost by installing 3x non-functional cameras?"
As most of the cameras are not monitored then effectivly they "are not working".
I suspect that the easiest way to meausre CCTV effectivness is not on the streets but in corner shops or your local MacD's.
Find two stores in similar locations without CCTV and similar rates of theft etc over a reasonable period of time.
Put CCTV in one of them and see what happens to the crime rate in store.
It is interesting to note that a lot of stores actually trust their "till staff" less than the general public judging by the way every till has a downwards facing CCTV camera which as it only shows the tops of peoples heads is not likley to be of use identifing a "stick up artist" but would show staff taking money or incorectly ringing up goods etc.
A Home Office spokeswoman said CCTVs "help communities feel safer".
For me, it doesn't. Every safety measure is a reminder that it is necessary. It may *be* safer with it in place, but it doesn't make me *feel* safer.
CCTV does not prevent crime in my experience. Most shoplifters at our store do not even realize we have over a dozen cameras in that tiny liqour store. When someone is watching them it helps catch them, and it helps to show the rest of the staff who it was, but it does not seem to deter anyone. Also I have shown several shoplifters to the police using surveillance camera footage and they have NEVER arrested even one for theft from our store. The only shoplifter I have ever seen arrested was one who the owners nephew beat senseless and held until the police arrived. And he had enough sense to press charges the camera would have mysteriously not been working that day.
>> Women often not only fail to understand logic; they rely strongly on feelings and experiences,
Oh, we do, do we? Perhaps you are also under the fact-free perception that women fail to read Schneier's comment threads. Your intent, non-negative though it may be, is irrelevant: Your comments are hostile. Why go there? I mean, there I am, minding my own business, reading about the banal realities of CCTV cameras, and Wham! Gender-Left-Hook!
And don't think that lame-o excuse about how no-facts-were-involved-in-the-making-of-this-comment gets you off the hook. You invoked _logic_. Logic tastes better with facts. You'll get more mileage out of your logic, and far fewer complaints from the peanut gallery if you deign to include them.
>> For me, it doesn't. Every safety measure is a reminder that it is necessary.
That's a really good point, and probably has a lot to do with how communities react to these cameras. I'd also add that an aspect of CCTV cameras that I find disquieting is this idea that the criminals couldn't possibly be the ones on the other side of the camera. Who Watches the Watchmen? etc. People in authority are in a much greater position to do unchecked harm. If we give them more authority (access to data about people), how do we protect against this increased risk?
``A Home Office spokeswoman said CCTVs "help communities feel safer".''
In that case install fake cameras. They will have the same effect and are a lot cheaper to run.
In general I have no interest in the police solving a crime using CCTV. I want the police to PREVENT crime. It is of no interest to me at all for the police to find out who planted to bomb that blew me to smithereens.
I forgot to add - if the police can't find the time to look at the CCTV images why not outsource the job to China or India?
Alternatively, make all CCTV footage available to anybody who wants it. If Galaxy Zoo can find people to look at images of galaxies, surely we can find people who will trawl through CCTV images.
The question that you have to ask of any video security system is "would you want to show that picture to a jury?" Normally, you can get only the vaguest description out of a CCTV system.
A cool trick that can prevent more shoplifting. Make the all cctv feeds visible in the store.
This worked well in our case. Shrinkage when down 70% in the next financial year...and no turnover did not go down (it was up just a little).
"Maybe the money should be spent in providing alternatives to crime instead of a tool to be used after the fact?"
Most of he studies I have seen have said that the best use for resources in lowering crime in the long term is in providing decent education / childcare with strong role models.
However this appears to be disliked by politicos because it takes 10-20years for the benifit to be seen.
The next best use for resources is keeping the streets clean, tidy and maintained.
Although apparently not reducing street crime decent lighting does increase peoples sense of safety (and I do have my doubts about the ineffectivness of improved lighting for a number of reasons).
For CCTV to be of use to joe public in general it must produce either a deterant effect or rapid response.
There are trials in progress where the CCTV has speakers attached and the operators can actually say things to people.
A lot of "CCTV Success" stories have one thing in common which is high initial "buy in" from the Police, ensuring rapid response, however long term studies show that due to a number of factors this does not last and the crime rate tends to rise back towards pre CCTV levels.
There are four basic types of street crime commiters,
1, boozed/druged up individuals who have no idea or care that CCTV is recording their actions.
2, Oportunistic crime, where an individual is actually not going out to commit a crime but sees an oportunity and avails themselves of it.
3, Those planning on commiting a crime who will wait for a non specific target to appear. They often take precautions to either avoid CCTV or disguise themselves in some manner.
4, Specific target premeditated crime. Often the targets are high value in some way or another and the criminals know there is (or assume there is) CCTV and take specific action to limit the effects CCTV has on their actions.
Each level/type of criminal would be detered in different ways by CCTV and it's associated responses, some not at all. Which significantly adds to the difficulty of assessing if CCTV is actually working against crime.
What however has become clear is that unless used correctly CCTV does not stop crime or aid in solving / prosecution crime.
Further that the on going cost of using CCTV correctly needs to be evaluated properly prior to implementing a system which currently does not appear to happen.
Also CCTV should not be used as is currently the case in some parts of the UK for "fines" for very minor offences (children droping small items of food, workmen smoking in their vehicals etc).
If cameras serve as an effective deterrent then would more cameras be even more effective? Therefore could we theoretically deter crime almost to nothing with more cameras? Then we could look for people who look like they might commit a crime by the way they are acting. We could call it "though crime"
My car was broken into in 1997, and the incident was caught on camera. Before the incident, a co-worker saw the perpetrators snooping around my car and he wrote down the license plate number of their blue hyndai. An hour later, my car was broken into. The camera quality was so crappy that we could tell it was a dark hatchback, but could not positively identify the vehicle. The camera, by itself, would have been useless at identiying the perpetrators. However, without it, the only thing was had was someone seen by my car an hour before. By using both, the police knew that:
*Two people in a blue hyndai were caught snooping at my car.
*An hour later, my car was broken into and the perpetrators left in a dark hatchback.
This would have been quite a coincidence, so taking the information, the police were honestly able to locate the blue hyndai and tell the driver that a witness got his license plate number and the incident was recorded by a security camera. The perpetrator then confessed.
The camera alone was useless, but it was useful in corraborating other evidence.
How much do additional Police Officers cost in London?
What is the effectiveness of London Police Officers in solving or preventing crimes?
If nothing else, these cameras will prove very helpful in case of an alien invasion. Well, maybe not:
CCTV is not just for "serious" crimes. We also have "talking" camera's to tell you off if you've dropped some litter on the streets, or any other activity deemed "politically incorrect".
Lulu made me laugh. Responding to a comment about how women only react to feelings by reacting to her feelings being slightly brushed. Amusing.
It is of course true, women react to ... specific things in general. Women are typically easier to push towards socialism because, well, they're basically geared to raise children ... they have to keep everyone on level ground and solve problems. That's often done by taking stuff away from some greedy kid and handing it off to some poor kid. Men pretty much beat things up and take their crap.
More importantly, both inherent and artificial drivers can be used in the same way on various types of people. We can create an artificial racism-based driver and get "people of an inconvenienced race" following a political party line. We can get certain classes of men-- lower working men, homosexual men, etc-- to believe they're "entitled" for some random reason (work's dangerous, people hate you for being gay, etc).
Just play on these and convince people your initiatives are good. Cameras? Cameras make everyone play nice. Cameras will show that black people aren't all drug dealers. Cameras will prevent the football team from beating up gays in the streets.
I'm sure someone, somewhere, thinks this shit out when they push this stuff on the public.
Lulu's response was not overly emotional. She called a guy for using a weak argument. In fact, most replies have some level of emotion or the user would not have replied at all.
Since 'we' can create a racism based driver, do you conclude that there is no racism? If you really think that all motivators are contrived, then you must be at ease with any outcome.
I think it's pretty unfair to call the cameras security theater. The cameras are primarily a crime fighting tool not a means of stopping terrorism and unlike terrorists potential criminals can be effectively deterred by the prospect of legal or social punishment.
The real question is how many crimes don't happen at all because of the cameras. Wheter or not the cameras really make the crime easier to solve they likely make people think twice about breaking the law. Even being watched by crude drawings of faces has been shown to encourage better behavior in studies.
I mean if this is security theater then having uniformed police do nightly patrols is as well. I mean they would catch far more criminals if they wore plain clothes but the visible presence of police in the neighborhood discourages crime.
From personal experience in London, the crime most likely to be "solved" by a CCTV camera is a traffic offence such as parking illegally or straying into a bus lane. Resulting fines go to pay for more cameras
Erm, Not a lot of discussion about how easily CCTV cameras can be hacked.
At night just use a few high-intensity IR LED's and the camera is blinded.
Or, wear a balaclava, bike helmet etc,
Or, use the low quality images against the camera and wear a simple disguise. The make-up the movies use is easy, cheap and very google-able.
CCTV is good for policing the innocent, while helping those who want to commit crime. IMHO
@ Nomen Publicus
"...if the police can't find the time to look at the CCTV images why not outsource the job to China or India?"
Or why not develop a distributed search program.... We'll call it: Search for Everyday Intelligence at Home or SEDI@Home.
Thanks for the idea Greg. I will mention it to the management tommorow.
If the purpose of the cameras is to prevent crime then they are a failure. If their purpose is to convince the population they are being watched and best not get out of line, that is another matter.
If civil liberties people really wanted to make a point about the omnipresent CCTV cameras, they ought to get some black spraypaint and spray the lenses one by one.
Bonus points if you can do it to more than a handful of cameras before armed cops swarm in and take you down!
The Sydney rail network has something like 1600 cameras, but they are monitored by staff who monitor approx. 80 cameras each AND are actually allowed to watch TV (TV sets are provided for their leisure). How much crime is THAT going to solve? Their slogan is "1600 cameras to make every station a safe station." I don't think that my family receiving a blurry photo of a man in a mask after my death is safety.
I think Frank's comment is the most dead-on accurate so far. A "blurry photo of a man in a mask" after a crime has been committed is of questionable value. How much has CCTV footage actually been instrumental (rather than merely just present) in investigations of serious crimes? All the evidence points to a very minor effect. It's just not that helpful. And Britain's crime rate trends would seem to put the lie to the notion that heavy CCTV use has a significant deterrent effect on crime. At best it has a minor effect.
As many people point out above, the real question about cost is the opportunity cost of not spending that money on police. How many more crimes could have been solved by having spent the money spent on CCTVs on hiring new detectives? How many more crimes could have been deterred by having spent the money spent on CCTVs on hiring new police officers to walk beats and be on call? It's very, very difficult to imagine that such changes in spending would not have had a much larger reduction effect on crime.
This message is being typed unemotionally by a woman.
"Sexism is a social disease"
"The statistic "70% of solved murders are solved with CCTV footage" is meaningless, since it fails to give the details on how many of these murders wouldn't have been solved without the use of the CCTV material? "
First, I would say that this is not just an overly strict criterion, but absurdly so. It is patently not true to say that something is useless unless it does something that cannot be achieved by any other means.
Secondly, even if this line of argument be thought valid, it is rather unhelpful as it is impossible to answer either way. To give a true answer we would need to investigate every murder twice, once with and once without CCTV footage. Every other possible analysis (geographical, time series etc.) is riddled with confounding factors which, particularly in the case of murder, cannot be eliminated by any morally acceptable experiment.
So we must conclude that if you insist on irrefutable evidence then there can be no debate, just an argument.
But there is indicative evidence, even if it can't be claimed to be wholly reliable -- and all of that indicative evidence suggests that while CCTV is no silver bullet, when used thoughtfully it is one of the most effective anti-crime measures yet developed. In certain types of crimes (notably those that naturally tend not to have human witnesses) it is the most cost effective anti-crime measure yet devised.
"As many people point out above, the real question about cost is the opportunity cost of not spending that money on police. How many more crimes could have been solved by having spent the money spent on CCTVs on hiring new detectives? How many more crimes could have been deterred by having spent the money spent on CCTVs on hiring new police officers to walk beats and be on call?"
A perfectly reasonable form of analysis. But here you go astray:
"It's very, very difficult to imagine that such changes in spending would not have had a much larger reduction effect on crime."
Actually, that's not true, because the relative cost of hiring police officers is much higher than you seem to imagine, and the cost of installing cameras is much lower. Never having installed even a single one of those cameras would have released funds to hire just 690 more officers, to cover the entire country of 61 million people. It is not a good trade-off, it is a very poor one.
That, in fact, is much of the point of using cameras: by using cheap cameras in some areas (especially those that need to be monitored for a very long time to achieve small pay-offs) you free up expensive, highly trained police officers for more effective patrolling duties.
I see a problem with the "feel safer" Aspect:
If people think (Even if thats not the case; you dont now it for sure after all) that someone is monitoring the place and calls for help in an emergency (thats why my Mother is pro-CCTV) they may act less cautious. They may not hide their money well, don't keep their wallets on a save place, don't keep their purses close to the body and so on, leading to MORE Opportunistic crime.
@Rob Oxspring: "one of the figures most frequently used to justify the claim — that there are 4.2 million cameras — is an estimate based on a survey of only two streets in London seven years ago."
I can imagine them not knowing how many cameras are installed or regularly monitored. But surely they're embarrassed not to know how many they've purchased? Do they know how many vehicles they have? How many employees? How many firearms? Most organizations would fail their audits without some sort of asset tracking in place.
As an American living in Birmingham for the past year, I have had several different reactions to the cameras here. First, of course, came surprise and suspicion. Then, as I realized they were in places where I had no expectation of privacy, apathy. Later, as I began to notice the sometimes strategic placement, I actually felt comforted. Not so much for prevention, because I don't think that was ever their purpose, but if I *were* mugged *here*, there would be some record of it and that might help ID the perp.
If I believed criminals were smart, which I don't, I might buy the argument about prevention. But I think cameras mainly exist to record an actual crime, therefore, by definition, they are not about prevention since the crime is presumed to eventually take place. They are an investigative tool. Maybe they reduce crime in their specific location a bit. How you would go about proving that is beyond me. And even if it turned out to be true, it may be that the bad guys merely wait for you to round a corner and change the location of the attack. If measurable rates go down by the camera but go up in the surrounding area, what have you accomplished?
CCTV is a tool, nothing more or less. We can use it properly or we can misuse it. But it is an expensive tool, and if they actually only solve one crime per thousand, then it seems to me this is more a question of economics even than civil liberties or security theater. The question simply is: are they worth the cost? If the answer is no, then the civil liberties debates are purely academic.
Regarding crime rates and feelings of safety - From one of the papers on http://john-adams.co.uk/papers-reports/ (I think the latest one) I noticed the stunning statistic that one third as many children die on roads now as in 1922 (in the UK). Couple that with the other statistic that, in 1970, 80% of 7 to 8 year olds traveled to and from school unaccompanied by an adult and you infer that the crime rate is probably lower because people are too afraid to leave their homes/cars.
There are many reasons why it is not probative to compare statistics from 1922 to those from 2009. Even 1970 is really pushing it.
"If cameras do simply 'help communities feel safe' that might be a sufficient good in itself, of course."
I agree that sometimes there is a gap between how safe people feel, and how safe they are. I agree that aligning the two would be a Good Thing.
I don't think that 'security theatre' is a good way of doing this though.
If A, B and C actually MAKE you safe, but D makes you FEEL safe, which of the above do you think will get cut in the next budget?
I fear that real safety spending (A, B and C) will be cut back (partially to pay for D) until eventually you actually are as unsafe as you feel!
Arguably D has now done it's job by aligning perception and reality.
Is Michel Foucault spinning in his grave yet?
> > ... the claim — that there are 4.2 million cameras — is an estimate ...
> But surely they're embarrassed not to know how many they've purchased? ... Most organizations would fail their audits without some sort of asset tracking in place.
Ah, you seem to have been deceived by the number one myth -- or should we say propaganda story -- of CCTV: the idea that most CCTV cameras are part of a government operated security network. In fact to get to the huge numbers commonly cited today, you have to include all CCTV cameras of all types, most of which are not government owned, most of which are not networked, and many of which are low resolution cameras intended to serve a safety function rather than a security function.
The Home Office does know how many they have purchased -- and it is a LOT less than 4.2 million. What they don't know is the number of privately installed ones that can observe public places, which some people have claimed (apparently on scant evidence) now numbers in the millions.
How it seems to go is this: some people are very concerned about the civil liberties implications of CCTV cameras. They are not alone there, it certainly is an area that could easily be badly abused and so protections must be put in place and publically monitored. However, some people are not satisfied with such limited nervousness and so seek to create greater fear by promulgating many myths, including such ideas as:
* most CCTV footage is fed into networks
-- In fact so few cameras are networked that the main way of obtaining footage for a criminal investigation is to send police officers from door to door asking for it. The Home Office is attempting to compile a voluntary registry in order to speed up this process.
* all street cameras are high resolution security cameras
-- I don't know the exact breakdown, but certainly many if not most are traffic surveillance cameras intended to alert traffic controllers to accidents and traffic jams on major roads. The resolution is an order of magnitude too poor to record faces or registration numbers. And most of the high resolution cameras are not video cameras but still cameras activated by detection of a traffic offence.
* coverage in British cities is so intensive that there are few urban areas that are not under surveillance
-- The urban area of the UK is somewhere around 13,000 km^2. If 4.2 million is the correct number of cameras, then even if there is no overlap at all (a highly conservative assumption), they need to average 3,100 m^2 each, or, say, a 30° arc out to a range of 110 metres. High quality cameras can do this but the great majority cannot resolve a face at this distance, or see anything at all in poor light. Furthermore for every camera that covers just a few square metres by focussing on a garage forecourt, the back wall of a shop or the other side of the street -- or covers thousands of square metres, but does it on a "black spot" somewhere out on the M1 -- some other camera needs to cover an additional 3000 square metres to keep up the average. It becomes clear that 4.2 million cameras is nothing like enough to provide total coverage of British cities; it wouldn't even be enough to cover a featureless plain of the same area.
* CCTV footage is commonly retained for very long periods of time
-- In most systems of which I am personally aware, the operators planned to rotate through a week's worth of tapes, but found it a lot harder than it sounds, not to mention largely pointless, and gradually slipped back to a couple of days. With the advent of digital recording, longer periods of retention are becoming more common but anything more than ~ 1 TB is still very unusual because once you go over 1 disk, maintenance effort increases greatly for negligible benefit. At typical MPEG data rates, 1 TB corresponds to approximately (2 months) / (number of cameras).
* facial recognition technology enables CCTV cameras to track individuals
-- After 2 decades of biometrics companies claiming they would have the problem cracked within a couple of years, facial recognition error rates are still so high that they are completely useless for this sort of application. And at current rates of improvement, it will be a century before they are good enough.
I mentioned earlier that with (allegedly) 4.2 million security cameras, the claim that "coverage in British cities is so intensive that there are few urban areas that are not under surveillance" cannot be true. A further observation on this point:
A typical reasonably good quality digital camera might have a resolution of 1280 x 1024 (remember, we're talking about averages here.) By fiddling with some digital photos, I'd say a face is kind-of recognisable at 32 pixels, but not at much less. A face is around 20 cm wide. Hence, the maximum scene width at which a face is recognisable for such a camera is about 1280 / 32 x 20 cm = 8 metres.
Now that scene width can be constructed from any combination of short range / wide angle through to telephoto, and the surveyed area is maximised at the greatest possible range. But to achieve 3,100 m^2 would require a range of 775 m and a field of view of just 35 minutes of arc. Clearly very few cameras are fitted with such enormous telephoto lenses (in fact more of a telescope!) A more typical field of view is the 30° I mentioned earlier (and actually, 45° and 70° lenses are also quite common.) At 30°, the total ground surface area surveyed out to the point where the field of view is 8 metres wide, is a mere 60 square metres. *SIXTY*, not three thousand! The camera may of course be able to view objects outside this area but it cannot do so with sufficient resolution to identify them, and hence no more has them under surveillance than if you glance at the windows of a distant tower block.
So it would seem that even under unrealistically optimistic assumptions, the level of surveillance in British cities cannot be much above 2%, and more realistically must be much lower. You might well consider even 1% to be too much already, but it does go to show that the alarmist claims at present are filled with hyperbole.
"... hiring x/10 actors in police uniforms to walk up and down the streets?"
In the UK we call them PCSO's (Police Community Support Officers).
Roger, your calculations are way off.
1) The size of the picture in pixels is 1280 * 1024, not just 1280
2) You cannot have the size of the face as 32 pixels * 20 cm, it does not make any sense as the units are not the same
I won't even go into your angle calculations...
"hiring x/10 actors in police uniforms"
So do actors cost more or less than sworn officers and where does this gig pay on the scale. Does it count for my SAG card?
Martin: What about you? Joined the force, huh?
Terry: Oh no, I'm not a peace officer. Yeah, this badge isn't a meaninful symbol. We don't enforce the law, we just execute company policy for homeowners.
M: When are you authorized to use deadly force?
T: Oh well, you know, taxes provide your basic services, you know, police and whatnot, but our customers, they need a little bit more than that, so we catch you on the property, we do what we have to do.
M: So, if I just look suspicious on your customers' property - under those, you know, "heightened circumstances" - you have the authority to shoot me?
``A Home Office spokeswoman said CCTVs "help communities feel safer".''
Emotions are neurological shortcuts in our decision making and these can kill us. But more frequently they don't.
The problem is with the word security. I feel safe, I feel secure. 10 million years of evolutionary pressure has made us pay attention to that feeling, act on it, and seek it out; real or not.
We're trying to use the word as a technical term of an outcome for characteristics in design, implementation and measurement. Not a description of our comfort zone.
It's kinda obscure. A quality that works fine until it doesn't.
And here, day after day, story after story, discussion after discussion, we make the same semantic confusion that the Home office rep did.
> 1) The size of the picture in pixels is 1280 * 1024, not just 1280
Erm, I'm not talking about the area of the picture, only its width. The height of the image is irrelevant here as we are considering only a projection of the field of view onto the 2D surface of the earth.
> 2) You cannot have the size of the face as 32 pixels * 20 cm
Indeed, but I didn't say anything of the sort. 32 pixels and 20 cm are both widths of the face (one its actual width, one the width of a representation), and a value of "32 pixels * 20 cm" doesn't appear anywhere in what I wrote. "1/32 pixels * 20 cm" appears implicitly in "1280 / 32 x 20 cm", where it is the conversion factor between pixels and width at the distance at which a face is barely recognisable.
> I won't even go into your angle calculations...
Perhaps you should. I am comfortable enough with simple photogrammetry to be very confident that they are correct to an appropriate number of sig. fig.s, but as the rest of your comments seem to have misunderstood me, perhaps my expression is unclear?
"Fact is that in some parts of London where CCTV has been extensively deployed, the crime rate has actually increased http://securityauditor.wordpress.com/2009/08/26/...
Neither your own blog post to which you link, nor the UpMyStreet page that you use as a primary source, actually support this claim. UpMyStreet's figures show that Walthamstow has above average crime rates, but doesn't in any way indicate that this is due to, or even occurred after, the installation of CCTV cameras.
OK, i have reread your comment a few times and seem to finally grasp what you are trying to do here, it would have helped if your equations were a bit more clear.
I am not convinced that "crime cameras" decrease crime and I am positive that "speed cameras" increase revenue and have no other mission. So I believe they are a waste of tax dollars (if not actively criminal misappropriation thereof which would be the case for "speed cameras").
That said, however, I am contemplating putting cameras outside my house where I can log in from the internet and see what is taking place around my house (wired ones, with a DVR capability). But this is as much (if not more) entertainment value and a "fun toy" as it is security. I would probably use it more to keep tabs on how the city treats my garbage cans or whether the mail has arrived than to thwart any "crime".
@: bob "contemplating putting cameras outside my house "
I've heard it quoted that 95% of all home burgler alarms are false positives.
Humongous waste of police time. What I want is an alarm that starts my camera and sends me a text. Then I can use the camera to validate the alarm.
At the begining of summer a woman in DC used her cell phone to look at her home cameras and saw a group of men inside robbing her. Called the police. Robbers Busted.
Fark's headline for this story was awesome: "about 2 out of every 1984"
@Bruce: "It the £200 million solved 10,000 murders, it might very well be a good security trade-off."
I can tell you for sure they didn't. There haven't been that many murders in England and Wales in the last 10 years, let alone that many solved by CCTV, let alone just in London:
One of the people calling himself "Roger" sneered at "Neighbourhood crime initiatives".
Actually these can be extremely effective. I used to live in Britain and now live in Switzerland - where the crime rate, especially burglary, is much lower. One reason it is lower is that people look out for each other. They call the police when they see something very suspicious. In Britain, there are cases of people who've actually witnessed burglars stealing from homes, and just looked the other way.
I don't sneer at people whose community spirit makes them look out for each other, nor do I sneer at really thoughtful urban design that "designs out crime."
I sneer at the phrase "neighbourhood crime initiatives", because it is a "glittering generality." That is, the sort of phrase politicians use when they want everyone to think they agree with the politician, but which is actually content-free. Like the infamous "hardworking families"*. Pretty well everyone thinks of themselves or their friends as "hardworking families", even if they are unemployed. Thus everyone is able to take a selfish interest in initiatives in support of those good folk, whilst at the same time getting the warm fuzzy feeling that, yes, it's about time we did something for those poor, long-suffering hardworking families.
In the same way, "neighbourhood crime initiatives" has "neighbourhood" in it, so perhaps it's about community empowerment which is surely a good thing, and it also has "initiatives", which certainly sound good. But it doesn't really mean anything except that it isn't regional -- and hence is not the politician's responsibility. In practice, it typically really means that the speaker has no actual ideas and wants someone else to fill in the gaps.
*Lack of a hyphen seems usual among politicians.
@ Roger 8/31 6:02
> First, I would say that this is not just an overly
> strict criterion, but absurdly so. It is patently
> not true to say that something is useless
> unless it does something that cannot be
> achieved by any other means.
Granted, but that's not what Bruce was saying at all. Put another way, it's also an absurdly strict criterion to say that "if something provides any sort of functionality, we call it useful regardless of cost."
> So we must conclude that if you insist on
> irrefutable evidence then there can be no
> debate, just an argument.
Falsification standards are obviously required, yes. So...
> But there is indicative evidence, even if it
> can't be claimed to be wholly reliable --
> and all of that indicative evidence
> suggests that while CCTV is no silver
> bullet, when used thoughtfully it is one
> of the most effective anti-crime
> measures yet developed. In certain
> types of crimes (notably those that
> naturally tend not to have human
> witnesses) it is the most cost effective
> anti-crime measure yet devised.
(cough) Uh, citation needed, rather desperately, for a claim of this magnitude.
"-- The urban area of the UK is somewhere around 13,000 km^2. If 4.2 million is the correct number of cameras, then even if there is no overlap at all (a highly conservative assumption), they need to average 3,100 m^2 each"
Not sure where you got the figure for the urban area of the UK as somewhere around 13,000 km^2.
However I suspect even if it is acurate in terms of total area it is not at all acurate in terms of public area.
The avarage suburban 1930's semi-detached house is some 10m x 7m and sits in a 10m x 50m plot.
If you assume back to back properties then public side streets are 100m appart and are about 14m wide for side streets and 20m for roads. So the public area is going to be less than 1/8th of the total area.
Further CCTV is rarely put on side streets but main roads and shoping areas.
I suspect on looking at the local high res OS map (which shows individual properties) and with local knowledge of where CCTV is that the total percentage of total area covered is less than 0.5%, whilst covering more than 80% of public acess areas.
But the coverage area on main roads etc spreads out quite far.
Within 50m of where I live is the start of the CCTV covarage area that if I walked into town would have me covered 100% of the way for over 4km. It is similar for other main roads in and out of town.
There are ways I can walk down side streets and ally ways to avoid the coverage but the town center of about 1km x 1.5km is very much covered with only a few publicly accessable spots (church yard etc) that are not obviously covered.
So even though the percentage of total land area covered is very very small it feals like you are being covered in an area of 4Km radius around the town.
As regards your calculations of face and camera resolution, I can assure you that most of the high street cameras in my area are capable of zooming in so that your face fills the screen at 50m. The analoge camera's themselves realy only have a usable resolution of less than 500 lines with maybe a bandwidth sufficient for 500-1000 points on a line.
However some have primary optics of 150mm diameter (which puts most SLR lenses to shame ;). I have had the chance to play with similar lense size CCTV systems and faces are recognisable out to well beyond 100m (without problems on a sufficiently solid mount).
However the newer high speed digital camera's with their optics can read car number plates for automatic recognition at 70m on a heavily overcast day.
These are now being produced in sufficient numbers that cost/reliability/performance wise they will soon start replacing the analoge cameras.
With regard to "face recognition" it is a very thorny subject at the best of times. However it has progressed far enough for the likes of the UK Passport Office to accept it for trials on automatic user/passport validation.
Likewise you have to define what you mean by "face recognition" ie pull it out of a DB of the UK population, pull a face out of a DB of local trouble makers or follow a face through a crowd. The latter two are currently in use in the UK. The London Underground did have a pilot scheme where a recognition system would tell if individuals (beggers pick pockets etc) where repeatedly appearing or loitering at stations.
While the cost effectiveness of CCTV in terms of actual crimes solved using footage from them, what about the preventive effect?
Any numbers or estimates on the amount of trouble and mischief that does NOT happen because would-bes realaize they're on camera?
"£500 million was spent on new cameras in the 10 years to 2006"
That is only for buying those cameras, or is installation, observing, service, power, storage, observation, ... included in those costs?
"That is only for buying those cameras, or is installation, observing, service, power, storage, observation, ... included in those costs?"
I don't know. All the data about security cameras in the U.K. that I've seen -- cost, number, etc. -- are sloppy.
Hello Clive, only just noticed your reply, sorry.
> Not sure where you got the figure for the urban area of the UK as somewhere around 13,000 km^2.
I calculated it, by dividing the mean of the published population density for a couple of cities into the urban population (80% of UK total population.) I've since found an estimate on the UK National Statisitics Office web site which is even higher at 22,000 km^2, and thus suggests the coverage is even lower than my estimate. Of course the exact value doesn't matter much as I am simply arguing that the coverage is much less than 100% and it clearly is less than that by orders of magnitude.
[... calculations about ratios of house areas to roads ... ]
> Further CCTV is rarely put on side streets but main roads and shoping areas.
> I suspect on looking at the local high res OS map (which shows individual properties) and with local knowledge of where CCTV is that the total percentage of total area covered is less than 0.5%, whilst covering more than 80% of public acess areas.
Perhaps so. It doesn't invalidate my point; the myth is that coverage is ubiquitous, when in fact it cannot be. That it is more common on main roads than side streets, shops and back yards is to be expected, and supports the point, it doesn't oppose it.
> ... if I walked into town would have me covered 100% of the way for over 4km. It is similar for other main roads in and out of town.
> There are ways I can walk down side streets and ally ways to avoid the coverage but the town center of about 1km x 1.5km is very much covered with only a few publicly accessable spots (church yard etc) that are not obviously covered.
Well, unless you live in a very small village, the main roads up to 4 km out + 1.5 km^2 in the centre amounts to a pretty small fraction of the total urban area, again supporting my argument not opposing it. (I now live in a small country town / large village of well under 10,000 souls, and our urban area is just under 25 km^2 -- not including the Common, forest reserve, etc.)
> So even though the percentage of total land area covered is very very small it feals like you are being covered in an area of 4Km radius around the town.
Feels like, perhaps -- in other words, an illusion.
> As regards your calculations of face and camera resolution, I can assure you that most of the high street cameras in my area are capable of zooming in so that your face fills the screen at 50m. The analoge camera's themselves realy only have a usable resolution of less than 500 lines with maybe a bandwidth sufficient for 500-1000 points on a line.
> However some have primary optics of 150mm diameter (which puts most SLR lenses to shame ;). I have had the chance to play with similar lense size CCTV systems and faces are recognisable out to well beyond 100m (without problems on a sufficiently solid mount).
I am aware that there are plenty of cameras with that sort of performance. However the argument is about *average* performance. Remember, the 4.2 million camera figure is suspect, and is definitely far in excess of the number of Home Office funded cameras. So to get to 4.2 million, we are apparently including commercial cameras that face public areas. It is very rare indeed for these to have a 150 mm lens!
> However the newer high speed digital camera's with their optics can read car number plates for automatic recognition at 70m on a heavily overcast day.
Which is well *below* the performance it would need to support an average continuous coverage of over 3,000 m^2, so even if a very large proportion of the 4.2 million (probably including shops and garage forecourts) were replaced by these high quality cameras, they still wouldn't match the performance of the myth.
> These are now being produced in sufficient numbers that cost/reliability/performance wise they will soon start replacing the analoge cameras.
> With regard to "face recognition" it is a very thorny subject at the best of times. However it has progressed far enough for the likes of the UK Passport Office to accept it for trials on automatic user/passport validation.
Not a valid comparison, sorry. Due the the "base rate fallacy" -- much discussed on this blog -- automatic passport validation is many orders of magnitude simpler than pervasive surveillance. Even so, I am mildly sceptical as the EER is marginal even for this. I'd be interested to see their results. (The Australian government announced similar trials in 2002, allegedly to last 6 months, but only opened it to the public in 2007. Official statements have claimed a FAR of 1% with FRR of 2%, but off-the-record informers have said the true FRR is 20% and the true FAR is unknown. The operators claim that the 5 year delay was due to difficulty integrating the e-passport reader with the biometric system, a claim which is frankly unbelievable.)
> Likewise you have to define what you mean by "face recognition" ie pull it out of a DB of the UK population,
Yes, that is what people often claim is possible today, and I say is utter BS.
> pull a face out of a DB of local trouble makers
While much more plausible than the former claim, even this is highly dubious unless "local" is a very small area. It might be usable if it was just cueing a human operator, not automating the surveillance. Of course that limits your "hot list" to fewer than the number of camera operators.
> or follow a face through a crowd.
This one at least is easily possible, but doesn't have anything to do with face recognition, this is just image motion analysis. You can whip one up yourself in a couple of hours using free off the shelf image processing libraries.
> The latter two are currently in use in the UK. The London Underground did have a pilot scheme where a recognition system would tell if individuals (beggers pick pockets etc) where repeatedly appearing or loitering at stations.
Colour me extremely sceptical. Unless proven to the contrary by an independent test, I suspect that it actually works the same as the systems that casinos use to exclude card counters. They advertise it as being straight face recognition, but in fact the automatic part just cues a human operator to inspect the image, because the automatic processor is nothing like reliable enough by itself.
No worries about the delay I'm back in hospital and have been out of it a while myself 8(
"Perhaps so. It doesn't invalidate my point; the myth is that coverage is ubiquitous, when in fact it cannot be."
I agree with you CCTV coverage is absolutly not ubiquitous.
It does not have to be for people to feel like it is or be largley as effective as if it where.
By my back of a napkin calculation it only needs to cover a very very tiny fraction of the whole urban area to be largely effective for population coverage.
The reason being that the average person travels on well defined routes from their home to points where probably 90% of the local populace go (shops etc).
The fact that many people get to the end of their little road or side street and bump into a grey CCTV column within 50yards on a more major road means they see it every day.
The fact that the total ground area of their little road or side street with the adjacent properties is larger than the area they actuall travel on does not occure to them they never walk it or go there.
From the point of "street crime" it is only those tiny areas that people congragate in for everyday (shoping transport points) and social activities (pubs clubs etc) that need to be covered. For exactly the same reason Willy Sutton robbed banks.
I guess the argument about coverage ought not to be about geographic area alone but area covered by the number of individuals.
For instance a side street might have only one person an hour walk down it but the main drag in a shoping area on Sat morning could easily have 10,000 or more individuals per hour.
Thus a camera in the main drag would effectivly have 10,000 times the "people area" coverage than one in the side street.
With regards face recognition like you I'm very sceptical but as Bruce frequently points out about crypto attacks "they only get better with time".
And importantly I'm very very mindfull of the difference between real evidence and presented evidence and the way presentation effects an out come.
Frequently justice has "to be seen to be done" rather than "actualy done", and technology has an illusion of infalibility that realy is not there.
With regards the London Underground CCTV system I'll see if I can dig out some "unofficial" "first hand" perceptions as if I remember correctly the backend work is being done at a University I used to work at from time to time.
"They advertise it as being straight face recognition, but in fact the automatic part just cues a human operator to inspect the image, because the automatic processor is nothing like reliable enough by itself."
Personaly I think I would prefer this sort of system, simply because it would keep humans in the loop.
There are many reasons to do so but two spring to mind that are arguably direct positive aspects.
1, Checking for hinky behaviour.
2, Teaching hinky awarness.
In most cases I have seen of video tape of criminals they actually stand out quite noticably when you know what to look for.
It's not in major ways (stripped jumper and swag bag ;) but in little ways the way they stand about, walk around, what they look at and importantly what they don't look at. It's usualy different to other people going about their "lawfull occasions" and to a trained eye it stands out as "hinky behaviour".
Like many things spotting "hinky" is a skill and in a perverse way these face recognition systems that require human intervention are helping train the human to spot "hinky", which hopefully will decrease false positives, and thus free up resources for dealing with actual criminals etc.
In a way the systems are a "lossy matched filter" in that they help an operator by lifting out a possible signal out of the noise. The operator then applies other huristics to decide if the person is a problem or not.
Here's an interesting article about one company that is apparently trying to increase camera monitoring coverage and lower their costs by introducing a "spying game." http://www.dailymail.co.uk/news/article-1218225/...
Though I think is some ways it would have the property of increasing monitoring, and would lower cost, I also think that creating these feeds increases the risk of abuse as well... whether from private capture or viewing for personal gain (ie: extortion, revenge, snooping).
Unfortunately, such abuse would be an externality to the company in question, lowering their incentive to control use or capture of the footage.
Quite a statistics you got there, indeed cameras aid in solving crimes but aside from cameras there are a lot of ways to solve a crime and it gives much more substantial results like community awareness by conducting crime prevention seminars or perhaps fast-accessible police force action. What I think is that we should not rely on cameras only.
Quite a statistics you got there, indeed cameras aid in solving crimes but aside from cameras there are a lot of ways to solve a crime and it gives much more substantial results like community awareness by conducting crime prevention seminars or perhaps fast-accessible police force action. What I think is that we should not rely on cameras only.
Schneier.com is a personal website. Opinions expressed are not necessarily those of BT.