Schneier on Security
A blog covering security and security technology.
« Shaping the Obama Administration's Counterterrorism Strategy |
| Stupid Security Tricks: Key Management »
January 13, 2009
Two Security Camera Studies
From San Francisco:
San Francisco's Community Safety Camera Program was launched in late 2005 with the dual goals of fighting crime and providing police investigators with a retroactive investigatory tool. The program placed more than 70 non-monitored cameras in mainly high-crime areas throughout the city. This report released today (January 9, 2009) consists of a multi-disciplinary collaboration examining the program's technical aspects, management and goals, and policy components, as well as a quasi-experimental statistical evaluation of crime reports in order to provide a comprehensive evaluation of the program's effectiveness. The results find that while the program did result in a 20% reduction in property crime within the view of the cameras, other forms of crime were not affected, including violent crime, one of the primary targets of the program.
From the UK:
The first study of its kind into the effectiveness of surveillance cameras revealed that almost every Scotland Yard murder inquiry uses their footage as evidence.
In 90 murder cases over a one year period, CCTV was used in 86 investigations, and senior officers said it helped to solve 65 cases by capturing the murder itself on film, or tracking the movements of the suspects before or after an attack.
In a third of the cases a good quality still image was taken from the footage from which witnesses identified the killer.
My own writing on security cameras is here. The question isn't whether they're useful or not, but whether their benefits are worth the costs.
Posted on January 13, 2009 at 6:58 AM
• 33 Comments
To receive these entries once a month by e-mail, sign up for the Crypto-Gram Newsletter.
"quasi-experimental statistical evaluation... ...the program did result in a 20% reduction in property crime within the view of the cameras"
What about property crime outside of the view of the cameras?
Did it go down up or stayed the same proportionate to the general trend of adjacent areas without CCTV coverage.
There have been findings in the UK that although there are short term gains in the covered area for aprehension and conviction in the long term rate tends to be slightly worse than befor.
Various reasons have been fielded primarily that the criminals either take precautions to hid their identity or they move to adjacent areas. And it has been sugested that a combination of both has actually made the figures worse than would otherwise be the case.
The problem is that these things are very difficult to assess in a truly independent and statisticaly sound way.
What the first study does show (and is confirmed by many UK findings) is that "spur of the moment" crime (usually violent) is unchanged by CCTV.
And to be quite honest what would I rather see a reduction in "property crime" or "violent crime" in it's various forms.
"One of the most infamous cases is that of Richard Whelan, who was stabbed to death on a bus in 2005 as he attempted to defend his girlfriend. The horrific 33-second attack was all caught on camera. Anthony Joseph, a paranoid schizophrenic, baited the victim by throwing chips at his girlfriend and the killer grinned at a CCTV camera as he left the bus."
Are we supposed to believe that a clinically insane murderer, covered in blood and with a live witness would not have been caught without CCTV?
I find that hard to swallow.
Oh and the UK report has the most telling figure of all,
"In 90 murder cases over a one year period, CCTV was used in 86 investigations, and senior officers said it helped to solve 65 cases by capturing the murder itself on film, or tracking the movements of the suspects before or after an attack."
So CCTV is an "after the fact" evidence gathering device not a crime prevention device.
A question that unfortunatly cannot be answered is how many of the murders would have been prevented by "officers on the beat".
Devestating as property crime is viloence ranging through threatening behaviour through to murder is likley to be more so not just to the victim but their friends family and others who know them.
As Bruce notes these systems cost money (a great deal of it) and a over and above the question of if the CCTV system is worth the money is, what else would the money fund and would it be more effective...
So, CCTVs makes it easier for the police to catch crooks. The victims are still screwed. I say get the cops out of their cars and start foot patrols - lots of foot patrols. Bike patrols. Segway patrols. Point patrols. Anything that puts cops between me and evil doers. Here in Toronto their motto is "to Serve and Protect". Patrols would help make that a reality.
Foot patrols are classic security theatre. Criminals, even the least bright, are generally smart enough to check both ways before committing a crime.
This is a well-known dichotomy for police: the more they do to reassure, the less effective they are at preventing crime. The more they do to prevent and solve crime, the less reassurance (because visibility drops). One reason why elected police chiefs are a potentially bad idea.
The problem with CCTVs are vastly udnerestiamted. When they are networked and connected with phase or other biometrics recognition, CCTV turns into realtime or delayed attack weapons.
Terrorist can attach bombs triggered by the targets to CCTV.
Commercial entities can continue to undermine markets by weakning demand power to pull the value chains through auto-id, profilling auto-response.
The technocrats will continue to make government more inefficient through command & control economics because they believe to be able to make descision better than the citizen.
And that is of course before we deal with the deeper political, criminal and social related problems.
It is a twisted mind that favour a world where people are scared to comply with desired to put them in small boxes of social control.
In some areas CCTV is effective after the fact for many reasons. It does not prevent crime at all in my experience. I have worked at 2 liqour store with many cameras and plenty of signs telling people their are cameras. Most people are too stupid to really care. I have caught shoplifters directly in front of a camera being incredibly obvious about what they were doing, after they looked at the cash registers to see if anyone was watching (which is what got my attention). They are most useful in helping us identify people after the fact and share that knowlege. Without them we would be telling each other "It was a young guy with dark hair". No way you can use that to spot them if they come back. But we are different than a city. We are a small building where 8 cameras is enough to cover ~90% of the building. Cameras make sense in small areas like stores, larger ones like airports, but not all over the city. That money could have gone to so many more useful things. Its sad.
Cameras will only deter non-determined criminals. Anyone planning a crime will just take the cameras into account during the planning. Anyone committing a crime out of passion or insanity won't care if they're recorded. So basically, they might stop some graffiti.
Meanwhile, the monitored cameras are being pointed into girls' apartments.
Interesting background info on the Richard Whelan / Anthony Joseph case:
His killer, Anthony Joseph [...] was a serial bail-jumper with a string of convictions and a record of missed court appearances. He had been wrongly freed from police custody hours earlier.
Joseph, now 23, pleaded guilty to manslaughter and is in Broadmoor secure hospital.
The blunders leading up to Joseph’s freedom on the day he killed Mr Whelan sparked a comprehensive investigation by Solicitor General Vera Baird, which reported on Monday.
Although the report makes clear no one could have predicted that Joseph would become a killer, it paints a damning picture of the attitude shown by courts and police to accused people released on bail.
So cameras to put them in, and sloppiness to get them out. Well ... But Trust The System(tm)!
It helped to solve 65 out of 90 murder cases? It's worth it. Period.
Convicting criminals gets criminals off the streets, and so reduces crime.
"It helped to solve 65 out of 90 murder cases? It's worth it. Period."
Not if the cases would have been solved without them. Period.
"It helped to solve 65 out of 90 murder cases? It's worth it. Period."
Err it is not clear what "helped" means so I would be a little sceptical on this one.
At a guess I would say that at best the footage would have been supporting evidence in court. The conviction would (/should) have been obtained on other primary evidence.
We would need to see the total "street murder" numbers and make a comparison to start making an informed decision. Then check if the footage played a key role or not. There have been several high profile cases where CCTV footage has shown persons but suspects where only found after the old fashioned detective work...
Also I guess you missed the bit in the second artical about the Home Office (UK Gov Dept responsable for Policing) spending 78% of the budget or approx 1billion USD on CCTV...
So no I'm not at all certain the price is worth it.
"Convicting criminals gets criminals off the streets, and so reduces crime."
Err no not in the UK, most are usually bailed after their initial arrest, and often they will not get a custodial sentance due to the fact that the UK prisons 70,000 places are all taken. Even then they are usually out within a relativly short period of time. It is only when the press take an interest do jail terms appear to be adheared to...
Again there have been a couple of high profile cases where criminals where released from custody and committed a murder within just a few hours of being released.
Not sure what BS wrote on the subject before, but -like every other "security device" - Cameras are only part of the system.
Where I live, I was recently attacked by 2 motorcyclists in daylight, on a state highway. Over about 15 minutes, while I was trying to dial 911 and get help, I travelled about 2 miles and wound up running over one of the motorcycles. (not manned at the time.)
When local police tried to sort this out, I found that we had passed 14 CCTV cameras - 10 Gov't run, and 4 privately owned store / bank cameras.
All 14 were not functioning that day..
or so I was told.
In my case, I wasn't hurt , and that's not the point.
The security system didn't help here at all. Maybe the cameras weren't maintained - maybe they weren't watched - maybe they didn't get recorded 'cuz something else was wrong.. I can only speculate.
Cameras - if they are strikingly obvious will deter the casual criminal. Otherwise, they won't stop crimes of passion (or road rage) or terror. They may - if durable enough - record something useful to be used later as evidence.
IMHO, cameras are mainly evidence collectors. They don't provide any security against a determined adversary. Fake cameras will prevent almost as much crime as real cameras, and then only if very visible.
Neither human police nor electronic surveillance can possibly prevent or protect you from violent crime. Incredible luck not withstanding, all a human police officer can do is administer punishment after the event. All a camera can do is record your demise. Only would-be victims and bystanders can stop such crimes.
The money would be better spent educating and preparing individuals to protect themselves at the scene and moment of the crime. Such real, distributed security is the surest and most compelling deterrent that money can buy.
I'm getting accustomed to flagrant abuse of statistics in UK newspapers, but this still got my attention.
"The first study of its kind into the effectiveness of surveillance cameras revealed that almost every Scotland Yard murder inquiry uses their footage as evidence."
This, by itself, means only that the cameras are ubiquitous, not that they're useful.
"In 90 murder cases over a one year period, CCTV was used in 86 investigations, and senior officers said it helped to solve 65 cases..."
As Clive Robinson has pointed out, "helped" doesn't mean a thing, but now we can test it: if I'm reading this right, during "a one year period" (why don't they just say which year?) the solution rate was about 70%. So if the cameras made all the difference, the rate in, say, 1990 must have been about 0%, and if they made no difference the 1990 rate must have been about 70% (after correcting for advances in, e.g., forensic science). Can anyone supply a source of raw numbers (which should have been cited in the article)?
The vast majority of murders are trivial to solve. The "perp" is usually obvious and the only real problem is a) finding them and b) geting evidence for conviction.
As a potential victim, I would much prefer not to be murdered in the first place. CCTV does nothing to prevent murders.
CCTV has its uses. For example to prevent shoplifting or "shrinkage" in shops (or cheating in casinos.) As a general crime deterrent it is useless .
Police foot patrols may be mostly theatre, but if the only time you see the police is in a speeding car going to some incident there is a huge disconnect between the police and the population. For example, if you know the local beat bobby, you are far more likely to talk to him/her and provide valuable intelligence. Information can also flow in the other direction.
(It suddenly occurs to me that the police should be on all the social web sites and twitter passing info back to the population.)
We can have a lot better and safer "evidence-collecters" than CCTVs.
Really vulnurable locations can merely setup a "login"-framework - they key is to make it conditionally identifiable only
there is a more insidious problem with ubiquitous surveillance. most people believe that seeing is believing, and visual identification is usually sufficient proof of guilt. when crimes are "solved" by identifying the perp from a grainy video, nobody presses for details. the fact that the footage exists is good enough.
this strikes me as similar to police forces who use psychics to generate clues in difficult cases. it always sounds better in court to say that the psychic identified the killer or the murder location than to say "well, we were pretty sure who the guy was, but we had no proof and couldn't get a warrant. so we broke in while he was at work, and looked around until we found some evidence." when cops and departments are under pressure to solve a case, they have to get someone quick; maybe not the right guy, but close enough for jazz. and the public is much more concerned with WHO the evidence implicates than HOW it might have been acquired.
the abuse of technology is an inevitability of human nature. this is not Orwellian in itself. the real problem is public ignorance of how ANY information that is collected on someone can be molded into a desired behavior pattern, and the fact that so much data is now being collected automatically. it is no longer necessary to spy on people, if police, politicians, and any private person who has political connections can search archives of private info [credit card transactions, phone records, license plates recorded by city cameras in london] whenever the need arises, and construct a portrait of unsavory or criminal activity based on arbitrarily selected or omitted data points.
Police foot patrols - "Bobbies on the beat" - are not in themselves a good use of resources, except in a few specific situations. However, when I was a city councillor in Cambridge I found that Community Beat Officers were a very valuable part of the police force. The CBO doesn't just plod around hoping to catch criminals in the act: he or she develops links with the community, gets to know the area and key individuals within it, knows where the problem areas are and which individuals are responsible for most of the trouble. They're the old-fashioned end of intelligence led policing. No, they don't catch many burglars hanging out of windows with bags of swag. But if there's been a couple of recent burglaries, and they know that Wee Bob McBurglar has just come out of a spell in pokey for the very same offence, then they'll be straight round to Chez McBurglar to make a few polite inquiries.
The bug in this system is that CBO is a low-ranking job, and anyone who's any good gets promoted out of it just as he or she is really starting to become effective. What we need is a proper career structure within community policing, so that it becomes an area where experienced cops can make a real contribution that is valued by the system, rather than it just being a stepping stone for junior cops on their way to higher things.
>> The program placed more than 70 non-monitored cameras in mainly high-crime areas throughout the city [San Francisco]
I noticed the psuedo-scary signs. Yawn. San Francisco is known for scary useless signs. One campaign using posters on the sides of buses offers a $1000 reward for an illegal handgun reported to police. Total number of rewards paid to date, over a year into the program: two.
Non-monitored cameras fit into two categories: evidence-quality, and time stampers. The former can be and are used for criminal prosecutions, although this doesn't seem to deter criminals all that much. The latter tells you that you were victimized by a dark blur and his four door car of unknown color at 03:35 AM.
The SF implementation was extremely poor as the cameras had very low frame rates (1-2 fps!). Not only not prosecution quality, but useless even as a time stamp.
Monitored cameras are a whole different matter. These are used extensively by security, but rarely by police (and even then, only to secure their own buildings and in custody environments.) Security can and does use cameras to observe and report suspicious activity and potential crimes in progress to law enforcement. The best extreme example is football stadiums.
I have heard that 'police patrolling in police cars to stop crimes makes as much sense as firefighters patrolling in fire trucks to stop fires.' If the police drive around with windows rolled up, A/C or heater blasting, and/or find convenient alleys to "rest their eyes" in, so be it.
The effectiveness of targeted police and security patrols is well documented. Mode of patrol depends on environment: foot, car, Segway, bicycle, motorcycle, boat, etc.
Random driving interspersed with enjoyment of the USA nationwide 50% Denny's (24 hour restaurant) discount for uniformed peace officers, not so much.
The Telegraph article is even more of a masterpiece of spin than you think. Even ignoring the cherrypicking of roughly 10% of all murders, using something for evidence doesn't mean the evidence it provides is actually meaningful. For example, CCTV use includes tracing the movements of suspects at times before or after a murder, which might exclude someone if the CCTV is far enough from the scene of the crime, but is otherwise nothing you couldn't have done as easily with interviews (and of course requires you to have suspects first, something that requires actual police work again.) Similarly, the 1/3 of cases where witnesses ostensibly identified a criminal by CCTV footage appears to mean that they had a suspect, combed a city's worth of CCTV until they got a good picture, and then showed it to witnesses. If I were a defense lawyer, I'd be thinking seriously about the prejudicial effects of showing a witness a surveillance photo and asking "is that the killer you saw?" rather than doing proper ID.
So we don't know to what extent the british CCTV plague has actually made a difference in solving crimes, and we don't know what the cost is compared to that of other, possibly more cost-effective methods. But we do know that that some of the police love it.
Is the underlying problem here that 97.423% of statistics (in a sample I choose not to disclose for exactly this reason) are made up on the spot and the remaining $percentage are interpreted by people who don't understand them?
How many journalists know the difference between correlation and causation? How many of those know how to spot a misleading interpretation of statistics? How many can follow the logic of a formal proof? Is journalism a sensible career choice for someone with these skills or are there other careers that offer better money for less risk?
The same questions apply to politicians.
It seems to me that a better career choice for anyone with both statistical, journalistic and political skills would be selling a product that is /statistically proven/ to achieve some holy grail.
...a product like CCTV systems?
Having said that, there are many places that seem to be run by people who understand statistics and formal logic, but I'm sure there's a reason why academia is academia and not the real world.
Perhaps watching the unwatched cameras could be a task given to the increasing numbers of unemployed. Wait, no, that's bad and wrong on so many levels, the least of which being the quality of daytime TV may increase.
"So we don't know to what extent the british CCTV plague has actually made a difference in solving crimes,"
The general crime figures in the UK (excluding "spur of the moment" type crimes) show upwards trends that appear to be mirrored or worse in areas adjacent to CCTV covered areas, and the same or slightly lower in CCTV covered areas. However it is difficult to the point of imposability of getting hard data to eliminate "adjustments" made by those "in charge".
"we don't know what the cost is compared to that of other, possibly more cost-effective methods."
You need to ask two questions, the first is who is spending the money? and the second is who else benifits?
Which brings me round to the reason for,
"But we do know that that some of the police love it."
There are three main reasons for this, the first and formost is that in the UK the Police don't pay for the systems, central government, local government and private companies pay for CCTV systems. From the article we know that the Home Office spent the equivalent of 1 Billion USD on CCTV systems. Local government has spent probably as much again, and private companies who knows but I suspect it might double or tripple what has been spent by government. In total this dwarfs the Police spend on fixed assets...
Secondly is the so called "CSI effect" which works on the fact that the general population are not sufficiently knowledgable to tell when they are being "sold a line".
This works for the Police in two ways.
The highest risk/reward is with juries as there is a chance the defence will have an expert to say why the CCTV evidence is not actualy evidence, or there is a small chance there is somebody on the jury who is knowledgable.
The second use of the "CSI effect" is showing a less than knowledgable criminal the CCTV footage and saying "we know this is you" and when the suspect says "it could be anybody" the police say "wait until the lab has cleaned it up". If the criminal still declines to "fess up" the police say "of course we can you've seen CSI" and if that still fails they remind the criminal about the picture a man posted to the Internet with his face swirled out and how they got his face back...
The third use is nearer to old fashioned policing, the Police have two methods of using it.
The first is to tell the suspect they have an eye witness who saw you do the following and read out they have seen on the CCTV tape. The chances are there is an unrelated person on the CCTV who the police can say "did you see the bloke walking his dog on the other side of the street well he saw you do..."
The second is as a method of catching the suspect out, which is to withhold certain key events and over a period of questioning see it the suspect reveals some of the details the police have not told them.
Re: ..while the program did result in a 20% reduction in property crime..
My theory is that 20% of reported property crime consists of fraudulent reports and no real crime ever happened. I expect we all know people who've lied in an insurance claim and you often have to make a police report just to get paid. Only a retard would report that a fake car accident or robbery happened directly in front of a security camera. Normal people would claim it happened around the corner in an unobserved spot.
This theory explains why reported property crime is lower within view of cameras, but real crime doesn't drop.
Has anyone seen the actual study?
Nah, I can believe that actual property crime in areas under surveillance could go down. Why would you shake someone down in front of the cameras or break into a building knowing that your actions are being recorded, when you could just as well do it somewhere else?
What's disappointing is the 20% number -- that means 4 out of 5 criminals decided to commit their crimes on tape despite the opportunity to do otherwise.
As one of the co-authors of the San Francisco study, let me address a few of the comments.
First, the study is available at: http://www.citris-uc.org/news/SFcamerastudy . I apologize in advance for the length; the executive summary contains the primary findings.
Next, the 24% drop in property crime occurred only within a 100 foot range of the cameras (and @ Pete Austin - these are felony property crimes, not just minor infractions. We have no reason to doubt that they are legitimate reports). Felony property crime outside of the 100 foot viewshed did not change in any significant amount. As we note, there was no effect on violent crime, which was the primary motivating factor in why the cameras were installed.
While there are flaws in how this system was implemented, it is important to note that merely moving to a monitored system is not an instant panacea. Effectiveness reviews of monitored systems (almost exclusively in Europe) note mixed results - just because you add people to the mix does not guarantee effectiveness. Systems that successfully deter crime depend on many contextual factors, including, for instance, mundane issues such as training and administration (which is why our study examined these factors).
So as long as we can afford to place a camera every 100 feet or so, we can expect a drop in property crime? What about the corresponding social discontent created by the monitoring itself? Any studies of the effects of having government minders watching you and judging your life on a constant basis?
@paul. Hi. So if you don't buy my first theory, why do you think there's a drop for property crime and not other types of crime? What do you think is different about these criminals which means that cameras matter for them?
To answer my own question, I wonder whether most property crime is committed by relatively few criminals who need to steal several times a day to buy drugs. If that's the case, the sheer numbers of their crimes means that even if cameras are *very* ineffective, "property criminals" who commit crimes in front of them will soon get unlucky.
Yet more evaluations into the effectiveness of "CCTV", and yet more confusion as a result.
There is a fundamental issue that relates to almost all academic assessments of video surveillance, and it's one which is rarely if ever explored in anything more than a passing reference.
The application of CCTV is as much about technique, as it is about technology, so saying that cameras work or don't work, actually means little unless the way in which the systems have been structured and operated, are viewed in relation to what might have been achieved if the systems had been correctly optimised from the outset.
To use a simple analogy, saying that CCTV tackles crime, is like saying vehicles are ideal for moving large numbers of people around a town.
If the transit company uses buses or coaches appropriately, it may work well, but if they decide to use two seater sports cars instead, the statistical outcome will of course be significantly different (yet in both situations, vehicles are being used to provide public transport).
Certain categories of crime can potentially be supressed through the effective use of video surveillance, a technique which I've previously referred to as "Deterrence through Detection" (DtD); in other words, the deterrent effect is not enhanced merely by the presence of equipment, but rather a belief that the technology is so efficient at producing tangible results, that capture on camera and subsequent identification would be almost inevitable.
Experience suggests that for some categories of crime, this may well result in displacement, and for others, the deterrent effect is somewhat mitigated by habituation.
Whatever the school of thought, in my experience almost all CCTV systems can be significantly improved (often at little if any cost), and their effectiveness as a crime reduction tool (in a general sense) certainly improved above what is frequently, an embarrassingly poor level of performance.
As for addressing quite justified concerns over civil liberties in relation to video surveillance, much could be done, and much still needs to be done.
One thing to think about. A camera may not prevent a violent crime but what the camera captures can be very helpful. First it can made the ensuing investigation go much quicker, thus saving money on a possible lengthy investigation. If all you have is a description of a suspect, it could take weeks, months or years to catch someone. Meanwhile, the suspect is still out there possibly committing more violent crimes. Second, once the suspect is caught, the trial could go much quicker, the suspect will have less of a defense or the suspect may even plead out due to the video evidence, thus saving money on a lengthy trial. So there are more benefits to having a camera than just trying to prevent criminal activity. When it doesn't prevent the activity, but catches it, you have indisputable proof of what occurred. If you were to become the victim of a crime, especially a serious or violent crime, you would want all the evidence you could get to catch the suspect as quickly as possible and to convict the suspect once he/she is caught. I think the positive benifits outway any negatives here.
You forgot to mention the significant costs of these systems and their environmental impact.
Also you forgot to mention that the output of most CCTV systrms even when a crime is commited right in the central field of view portion are effectivly usless for identification as only the least thoughtfull of oportunistic criminals do not take precautions to either hide their identity or carry out their activites out of the effective field of view of CCTV.
You also forgot to mention that (in the UK atleast) CCTV systems (from certain aproved suppliers) are a method by which the current incumbrent political party has sort funding into it's political coffers...
On balance I would say that the arguments in favour of CCTV systems are tenuous at best and the overall costs to society are considerably greater than the meager benifits they provide.
Schneier.com is a personal website. Opinions expressed are not necessarily those of Co3 Systems, Inc.