Schneier on Security
A blog covering security and security technology.
« Passlogix Misquotes Me in Their PR Material |
| More on Kish's Classical Security Scheme »
February 7, 2006
Gladwell on Profiling
This article by Malcolm Gladwell on profiling and generalizations is excellent.
Posted on February 7, 2006 at 12:09 PM
• 44 Comments
To receive these entries once a month by e-mail, sign up for the Crypto-Gram Newsletter.
Yes, I like the article too, but found several concerns. For example:
"do we really know what mafiosi look like? In 'The Godfather,' where most of us get our knowledge of the Mafia, the male members of the Corleone family were played by Marlon Brando, who was of Irish and French ancestry, James Caan, who is Jewish, and two Italian-Americans, Al Pacino and John Cazale."
Truth be told, the mafioso in America were in fact Italian, but the Irish (Charles Dion "Deanie" O'Banion) and Jews (Meyer "The Brain" Lansky) were also prevalent in organized crime and gangs.
Ironically, a common trait of these group s actually had much to do with dislocation from normal social and economic opportunities, due to stereotypes and discrimination against them (e.g. Catholics/Jews were heavily disenfranchised in the early 1900s).
What's the root cause of high-risk pitbull behavior? While they might be bred for severity, the likelihood of attack seems to be a direct result of training/mistreatment of the dog.
I like the concept of the "American Temperament Test Society" as it seems to takle the likelihood issue head-on.
As I tried to suggest on the log entry from a few days ago, a similar program for people might have identified someone with high likelihood of attack and helped prevent the CA post office tragedy...even if the severity couldn't have been reduced (e.g. prevented her from acquiring a gun with more effective control laws)
Funny. You must have read a different article.
Contrary to your claim, I read that there was a "known and meaningful connection between dangerous dogs and negligent owners" - which I would have expected.
I also read:
"Whereas guard dogs like German shepherds usually attempt to restrain those they perceive to be threats by biting and holding, pit bulls try to inflict the maximum amount of damage on an opponent. They bite, hold, shake, and tear. They don’t growl or assume an aggressive facial expression as warning."
... which you seemed to have missed and which also seems to be of importance.
But, of course, you *hadn't* read a different article. What you did was ignorantly seize on one part of what was said in order to extract an interpretation that suited your ideological bias.
Ah! Let me see ...
"stereotypes and discrimination".
*That* was what you were really after - a false parallel to human life: the dog is a "victim"; the criminal/terrorist likewise.
In point of fact, many people are subject to [unfair] discrimination without their doing anything other than leading blameless lives. But your attempt, of course, was not to *explain* crime but to excuse it. It's known as decadence.
Not clear if you are referring to my post, but since Bruce didn't post anything I assume you are?
"known and meaningful connection between dangerous dogs and negligent owners"
Right. The severity of attack is known to be high for pitbulls in particular, but the likelihood of attack is increased significantly based on how the dog is treated. In fact, it is such a critical factor that it can make the pitbull's severity insignificant to actual risk v. other dogs.
"which you seemed to have missed and which also seems to be of importance"
No, that's severity of the attack not the likelihood. But as the article explains, a pitbull that doesn't attack would pose less risk than any other animal that has been mistreated/trained to kill, etc.
"a false parallel to human life: the dog is a 'victim'; the criminal/terrorist likewise"
What's the false parallel? A risk equation allows you to review likelihood and severity of threat regardless of who or what is the agent. It is simple proof of fact that if you abuse people or dogs they are unlikley to behave in a manner consistent with those who have not been abused. So if you base your prediction on those who have not been abused, and ignore the prevalence of abuse, you are very likely to be surprised.
"your attempt, of course, was not to *explain* crime but to excuse it."
On the contrary, if you do not understand true threat you then what chance do you have of adressing it effectively? By integrating with the locals and figuring out the source of resistance and IEDs in Iraq the Special Force Alpha teams had a good chance of reducing the likelihood of the explosions. They weren't trying to *excuse* the resistance by understanding it better. Instead they were in search of the most cost-effective route to permanently eliminate it as a threat.
"a false parallel to human life"
Well, in some cases it is a false parallel. Pit Bulls don't hide their identity by disguising themselves as a different breed of dog that is perceived to be safe (see reference to Takfir W'al Hijra above).
Another issue is knowledge. How do you know that a person or dog has been or is being abused in some way? More information is better but in many cases the information you need to build a better risk profiles is unavailable.
"On the contrary, if you do not understand true threat you then what chance do you have of adressing it effectively?"
I think this falls into the 'Lies, damn lies, and statistics' category. Statistically, yes, you can make many points about the violent nature of pit-bulls (or muslims). But is it a valid set of measurements/results? Is it even meaningfull after you add more variables, and discover where the actual causal variables are.
Lots of things are coincidental. Far fewer are causal.
Causality isn't really important if your goal is to locate existing violent dogs or people. Corrolation is probably enough.
If your goal is instead to prevent the creation of new violent dogs or people, causality becomes more relevant.
There is an economic aspect to this. It's like newborn screening programs that states run to identify babies with inherited and other diseases so they can get early treatment. You don't necessarily run the best test you have as it would be too costly. You run a screen that strikes some balance between false positives, false negatives and cost.
"Causality isn't really important if your goal is to locate existing violent dogs or people. Corrolation is probably enough."
That depends on what you do with matches. If you're talking incarceration of people, or the banning/putting down of dogs, then just matching on corrolation isn't nearly enough. Too many false positives.
In the case of vicious dogs, selecting for minimal false positives means that you'll likely miss some, and attacks will still happen, but then the reaction to that attack is very important. Does the animal get put down (not necessarilly a requirement)? Does the owner get banned from owning in the future (loss of rights), pay reparations, and be put on parole (to ensure compliance)?
In the case of terrorists, a minimal false-positive situation that allows some false negatives can be very damaging. But then what's the cost/risk analysis on that? How much are you likely to miss, and how much damage will those you miss cause?
I'd rather aim for the situation of minimal false positives, and minimal false negatives, knowing and accepting that some will slip through the cracks, and in those cases, the best reaction is not to wring hands, but deal with the aftermath, and move on. Blowing up politically just is the kind of reaction that they want.
"I'd rather aim for the situation of minimal false positives, and minimal false negatives, knowing and accepting that some will slip through the cracks, and in those cases, the best reaction is not to wring hands, but deal with the aftermath, and move on. Blowing up politically just is the kind of reaction that they want."
Well, exactly but I think what one would gather from reading this blog is that a lot of the activities the government undertakes in the name of security and preventing terrorism doesn't accomplish anything like this. Huge cost (finacial and otherwise), huge false positive rate, and few if any positive hits.
False negatives rate? Who knows. What we do know is that there is no obvious profile because the terrorists are tailoring their behavior to avoid fitting a terrorist profile or will fit a profile that encompasses so many people that it is worthless.
"One of the most dangerous aspects of Al-Takfir W’al Hijra is that in some cases, they utilize a method of integration and assimilation in order to preserve secrecy and stealth. The Al-Takfir network is considered to have some of the most dangerous terrorist groups because their ideology allows them to adopt infidel practices in order to blend in to society and plot attacks more efficiently. These Al-Takfir groups are nearly undetectable because their members act and live in the norm."
A very well done essay however I do have an issue with his reference to pitbulls are safer than other dogs because they pass the ATTS test.
I know he is not saying this but referencing it to question the assumption that pitbulls are unsafe is somewhat misleading. The president of ATTS is quoted as saying: "Eighty-four per cent of the pit bulls that have been given the test have passed, which ranks pit bulls ahead of beagles, Airedales, bearded collies, and all but one variety of dachshund." All this does is state that they passed a battery of tests but doesn't state how the other breeds failed. Are we supposed to think beagles aren't as safe as pitbulls? I doubt the beagles failed because of aggression. I looked at the breed statistics for this test and would have like to seen the scores broken down into the five main categories of the test instead of pass/failure statistics.
Failure of this test includes: Unprovoked aggression, Panic without recovery, Strong avoidance. This test for example: "A handler takes a dog on a six-foot lead and judges its reaction to stimuli such as gunshots, an umbrella opening, and a weirdly dressed stranger approaching in a threatening way.", would cause many dogs to panic or avoid the situation. Pitbulls were breed for bull baiting so I would expect them to deal with the more threatening aspects of this test then a Basenji or a Beagle both of which were breed with entirely different purposes in mind. Also the backgrounds of the dogs would affect the results of this test. Aunt Mae's lap dog toy poodle is going to have radically different reactions than Uncle Joe's standard poodle he uses for hunting. Also the temperament of different classes of dogs i.e.: hounds, sporting or working are going to have different results to this test.
I don't think the ATTS test is a valid test to prove pitbulls have been falsely generalized as a killer breed. I don't disagree that pitbulls have been falsely profiled I just feel that using this test to prove his point weakens his argument.
I agree. The test itself is not perfect, but at least it is trying to find a way to measure the likelihood of attack to help narrow down the root causes. How would you propose identifying whether a dog is likely to attack?
Reminds me of the study of great white sharks in their natural habitat, which apparently led scientists to find remarkable differences in how they behave:
"There were aggressors and there were clowns; there were mellow sharks and peevish sharks and sharks that meant absolute bloody business"
ah, another distorted apologia for pit bulls, the author would have us believe that they're safer than beagles. then an insulting comparison of the rights of various dog breeds to the rights of people suspected of terrorism on account of their ethnicity. hellll-ooo, people used to have constitutional rights. the equal protection clause does not require me to treat a pit bull equal to a labrador. in the case of the pit bull, it's the combination of the aggressive nature and the unusually powerful jaws which have led to the death of one innocent child after another after another until finally, state and local legislators finally had enough and rebuffed the lobby of pit bull owners, people who train dogs to be aggressive and the methamphetamine cooks who hold this breed so dear. the reward is an unknown number of children who won't be savagely mauled by a pit bull, maimed or killed, and the price for this is you can't own a pit bull in enlightened places like denver and ottawa. that's a helluva bargain.
Interesting point, but I find it odd that you left of the closing sentence, since it supports your suggestion that there is no profile that fits..."These Al-Takfir groups are nearly undetectable because their members act and live in the norm. On the other hand, some Al-Takfir groups believe in a more conservative ideology and in turn, completely isolate themselves from society with the objective of launching their war from a distance."
Secret, yet open. Embedded, yet distant. While contradictory at first glance, it actually seems similar to the definition of a sophisticated military organization that assigns different roles based on strategy.
"contradictory at first glance".
No kidding. Read this from SF Chronicle October 4, 2001. Drinking, lap dances and more in Las Vegas. Not the profile I think. But their extreme fundamentalist beliefs allow this if it secures their ultimate goal. That's why they are so dangerous.
"However, locals and investigators are still exploring two main theories for why the radical Islamic killers gathered in a city they undoubtedly saw as a den of infidels wallowing in Western wretchedness...."They all listed Florida driver's licenses and very much fit in, pretty clean-cut guys," said Gerfy. "It's a shame. It would have been better if they were more criminal looking and easier to spot for something wrong.".
I was nearly attacked by pit bulls once. I lived in a very economically disadvantaged neighborhood, and had a neighbor who intentionally raised his pit bulls to be mean. Without a doubt, this person could have easily been profiled by the criteria that Gladwell suggests.
I take exception with "the reward is an unknown number of children who won't be savagely mauled by a pit bull, maimed or killed". Politicians love using these arguments because the premise is that a lack of data supports their argument.
I don't see this as any different than the following: I can guarantee your safety for only $2,000 per month. Send me the checks and if anything bad happens to you, I'll refund your money. If enough people want to be safe, I'll make a lot of money. If they are part of the statistical majority of people who don't have problems, they can attribute that to my "service".
I'm semi-neutral on pit bulls. I'm not going to support breed bans, but I'm also not going to own one. I have three dogs (70, 110 and 180 pounds). I've seen some proposals calling for bans on all large dogs, large being defined as bigger than a cat. The two larger ones are St Bernards. This article is one of the few places where I've seen references to people being killed by St Bernards. No doubt, they are large, powerful dogs. I like them because they have a good temperment. We have three small children and all of the dogs are very good with them. On the other hand, I certainly wouldn't want to be someone or another dog trying to harm any member of my family.
I liked the general idea of the article. I think the basic problem is differentiating between a problem and what people want to be a problem. I agree with the example that people want to think pit bulls are the problem. I think the data suggests that the people are a more significant contribution to the problem. That's certainly consistent with my experience with dogs and dog owners.
The conclusion of the article with regard to Islamic terrorists is that there is very little to be learned from profiling dangerous dogs. The New York City’s police commissioner, who according to the article has been pretty smart about targeting crime and, in his previous position as head of the U.S. Customs Service, appropriate profiling of drug smugglers, is doing random searches for terrorists in the subway.
"“You think that terrorists aren’t aware of how easy it is to be characterized by ethnicity?��? Kelly went on. “Look at the 9/11 hijackers. They came here. They shaved. They went to topless bars. They wanted to blend in. They wanted to look like they were part of the American dream. These are not dumb people. Could a terrorist dress up as a Hasidic Jew and walk into the subway, and not be profiled? Yes. I think profiling is just nuts.��?"
"politicians love using these arguments because the premise is that a lack of data supports their argument."
no sir. there is no lack of data from the past instructing us that pit bulls are an exceptionally hazardous breed. the only data we lack is data from the future, how many kids will be saved by a pit bull ban and which ones they are. this absence of future data is not peculiar to the pit bull issue, it's a general issue common to all legislation intended to increase safety, as yogi berra observed, "forecasting anything is difficult because it involves the future."
"it's a general issue common to all legislation intended to increase safety"
I think what's actually at play here is the following question: "Are bans against any given breed of dog designed to increase ACTUAL safety, PERCEIVED safety or both?"
I think that we all understand that actions taken that increase actual safety, but that don't lead to a perception of increased safety while quite effective, tend not to be big hits with the public - after all, since they can't perceive how they're any safer, the public will have legitimate questions as to what the program is doing.
To opponents of breed bans (and other sorts of profiling), they represent an easy way out - a method that increases the public perception of safety, but that doesn't increase actual safety. (Of course, there's nothing that says that a policy enacted with a cynical eye towards increasing only perceived safety won't also increase actual safety - which is where things get confusing.) That opposition grows when there are demostrated methods that (it is thought) will increase actual safety, but have costs that the mere preception of safety do not.
To the degree that it is possible to have perfect actual safety and security, the costs are going to be astronomical, and may very well (if not by definition) involve prices that will be considered to be not worth paying - especially if the perceptions of unsafety and insecurity linger.
For my part, I don't believe that banning dog breeds is the most efficient way of increasing actual safety, when compared to other means of detecting, tracking and controlling/disposing of dangerous and aggressive dogs. Therefore, I'd rather see other policies enacted.
It's not an "excellent" article. It's a rambling compendium of anecdotes designed to conclude that nobody knows anything. An "excellent" article would not attempt to pass off such logical monstrosities as refuting the observation that "nearly every jihadi ... is a young Arab or Pakistani man" with the observation that "Islam is a religion that spans the globe." Aristotle laid down some important principles of logic 2350 years ago, and an "excellent" article would use some of them.
In general, when reading an article that asserts that everything is confusing and nobody knows anything, it's best to bear in mind that the confusion and ignorance do not necessarily extend beyond the mind of the author.
It's not the devil you know, it's the devil you don't know.
Any mistreated animal will act aggressively, even the human ones. Although the pit bull isn't to blame, he is the tool used by the human to commit bad acts. Just like a gun. But removing the weapon isn't the solution to prevent crime, it simply removes one (of many) tools to commit crime. To quote a Mafia saying, you don't quiet a dog by cutting off its tail. We need to go after the big fish, the guy on top.
Peter Pearson: "Aristotle laid down some important principles of logic 2350 years ago."
Aristotle's syllogisms won't help you much. For one thing, they don't deal with probabilities and uncertainties. They don't go beyond "all humans are mortal, Socrates is human, so Socrates is mortal", and as the logician Bertrand Russell has pointed out, even this deduction is actually based more on inductive inference than on pure logic. Your statement "nearly every jihadi ... is a young Arab or Pakistani man" is already out of bounds for binary logic.
But let's consider the issue. The argument against profiling, presented not by Gladwell but by New York City’s police commissioner Kelly, is based on the correct observation that terrorists are able to adapt to profiling, by adapting clothes, appearance and behavior, or by recruiting people who don't match the profile. Even a probabilistic strategy cannot address this fundamental problem because whatever is known about terrorist patterns ("X% have been men and Y% Arabs") need not apply in the future. The article's conclusion is indeed uncomfortable as it shows that simple solutions are not likely to be effective. I find this very plausible and not at all confusing.
Concerning the pitbulls, I think the ATTS test statistic cannot be generalized as long as we don't know who selected the dogs. If only nice dog owners have submitted their dogs to the test, then the results are worthless. To get significant numbers, all dogs, or a sufficient random sample of all dogs, would have to be tested, which appears not to be the case.
Gladwell: "Figuring out what an Islamic terrorist looks like isn’t any easier."
Most of the recent ones in Europe and the US look and act just like everyone else. They blend in. However, we do know quite a lot about Islamic terrorists in Europe and the United States. There are patterns. See link to "Understanding Terror Networks" above. The problem lies in the fact that the patterns were built up from detailed biographies. They are only any use if you can match them against similiarly detailed biographies of all the people you'd like to screen. It isn't going to happen in the New York subway so the NYPD has to do random searches. And based on what we hear about Do Not Fly Lists etc. they aren't using that type of sophisticated patten matching in the airports either. And if they were I'm sure the terrorists would work around it.
As Sagemen concludes (Understanding Terror Networks): "The whole network is held together by the vision of creating the Salafi state. A fuzzy, idea-based network really requires an idea-based solution. The war of ideas is very important and this is one we haven’t really started to engage yet."
Disclaimer: I like dogs.
One defense of "pit bull bans" has nothing to do with the stereotypical profiling of the breed as aggressive.
Whether or not pit bulls are more or less likely to attack someone is only part of the issue. As the author of the book points out, the behavior of the dog owner is more relevant to the probability of a dog being aggressive than the breed of the dog itself.
However (regarding the justification of a ban) the other important factor is the behavior of the dog during an attack scenario -> the disaster recovery aspect. If a cocker spaniel is involved in an attack incident, it is unlikely to be fatal or cause egregious injury. If a shepherd type dog is involved in an attack incident, it may cause injury but the dog itself is usually attempting restraining maneuvers, so the likelihood of fatality is more related to the size of the dog vs. the size of the target than the attack pattern of the dog.
All terrier types aren't "pet" dogs. They are bred to be killers (albeit of different prey animals) - http://www.akc.org/breeds/terrier_group.cfm. A cute little fuzzy Westie will latch onto a mouse, lock jaws, and shake it to death just like a pit bull will. Terrier type dogs aren't for casual dog owners -> they require a lot of attention and activity to be happy, and a lot of active training to be well behaved socially. They do make great pets for owners who are suitable psychological matches for terrier-type personalities. Of all dog breeds, the "working" and "terrier" type dogs are the ones bred for behaviors that are the most likely to cause severe harm to a human *if* the dog is involved in an attack scenario.
Bull/Staffordshire terriers represent a bad disaster recovery scenario, because the dogs are physically capable of bringing down and killing much larger animals (unlike some of their smaller terrier cousins), and their inbred attack pattern of "latch and shake".
I disagree with "pit bull bans" on the whole, because I believe that someone who is a good dog owner who matches well to a terrier type should be able to have the dog (s)he wants to have, just like someone who wants to own a firearm should be permitted to do so if they demonstrate that they can be responsible gun owners. However, along those same lines, I do think it is reasonable for there to be restrictions for those who want a working or terrier type bred dog.
We license people to drive, and we have different levels of license for different types of vehicles. We do this because driving is dangerous, and certain types of vehicles are more dangerous than others. There is no reason why we can't extend that same philosophy to dog ownership, especially given that we already have a licensing mechanism in place for dog owners.
I don't think licensing pet owners differently would help. The owners are already liable for the actions of their pets. I generally think of government as a tool that can only be used for evil, never for good. Even measures that are implemented with the best intentions tend to get twisted in time to promote someone's agenda.
We do license classes of vehicles separately. Once someone demonstrates a minimum level of knowledge to operate a Yugo during a driving test, we give them a priviledge that immediately extends to a Corvette. If someone's actions are careless, those actions have their own consequences. Only in extreme cases is a license revoked.
Licensing enforces a lowest common denominator. It is not a credential that reflects ability. The problem with more government intradiction is that it replaces judgement with procedure. The consequences for bad judgement are already in place.
Should I be allowed to own a dog by default or only if I can prove some level of education? As with all things, there will be fees that are associated with the education and licensing. From your suggestion, it sounds like the requirements would vary based on certain characteristics of the dogs. It sounds like no matter how good the intentions are, the end result would have an adverse affect on poor people who are perfectly good dog owners.
How would my license be affected if I've had dogs that have bitten people in the past? What if, in those cases, I was the one who was "bitten"? What if it was as a result of playing where the dog hits my skin with an open mouth, resulting in broken skin? All of these things have happened, but I don't consider the dogs bad or myself to be a bad dog owner.
I equate minor injuries in playing with large dogs(70-180lbs) to injuries from riding a bicycle or skateboard. It's a risk that is accepted by participating in the activity. Yet, I've heard some people say that any time a dog harms anyone, it should be put down. I don't see a net benefit to society by giving those people yet another avenue by which they can enforce their opinions on everyone.
The fact that "nearly every jihadi ... is a young Arab or Pakistani man" is kind of beside the point, as far as profiling is concerned. Profiling bets on the fact nearly every jihadi is going to continue to be a young Arab or Pakistani man, and publically profiling such people bets on the fact that people who are not young Arab and Pakistani men cannot be induced to be jihadis, or to act on their behalf.
To the degree that such a simple statement about jihadis is true, profiling does make us safer, and even an extremely high rate of false positives has very low costs for the vast majority of us who don't fit the profile. The risk is that it can be rendered useless by finding an older asian woman, say, who has enough of a grudge that she's willing to help blow up a hotel or something.
Mike: "If someone's actions are careless, those actions have their own consequences... The consequences for bad judgement are already in place." It would be preferable to prevent those careless actions. It doesn't help much that the dog owner suffers negative consequences of bad dog behavior. For this to be effective, the owner must be aware that his actions may have those consequences (which might require some educational effort), and of course, he must care about it, which isn't always the case. Moreover, even if the owner is liable for any damage, he probably hasn't enough money to indemnify the victim if something grave happens. A solution would be to require dog owners to buy insurance. Premiums for pit bulls would be very high, so this might hurt your poor dog owner even more than a licensing requirement.
"Only in extreme cases is a license revoked." Driver's licenses are quite frequently revoked, and I think it is a good thing that reckless drivers face losing their license, and not only when it's too late.
"Licensing enforces a lowest common denominator. It is not a credential that reflects ability." Maybe in US, but this needn't be so. "The problem with more government intradiction is that it replaces judgement with procedure." Yes, if it's done badly. But if governments are forced to act by public opinion and the alternative is an outright ban, maybe some considerate licensing system would be not too bad.
when you scale down from the community to the individual, actual safety and perceived safety merge, it's an entirely subjective judgment influenced by experience. i suggest to you that just one bad experience with a pit bull will influence your analysis. i agree that breed bans may not be the most efficient remedy, but i don't look at it as either/or. the arguments on display against breed bans apply with equal force to species bans, why should my liberty to raise tigers next to a daycare center be curtailed in the name of hypothetical risk to children i don't even know?
your mafia analogy fails. when you remove the mafioso, his gun will be safe for knowledgeable people to handle. a pit bull is still potentially dangerous all by itself, even to knowledgeable dog lovers.
"profiling does make us safer, and even an extremely high rate of false positives has very low costs for the vast majority of us who don't fit the profile."
Like the federal government's profiling of Japanese-Americans in WWII?
So you inflict costs on a large bunch of "others", in this case the law-abiding members of the Muslim community, whose help and cooperation could actually greatly assist in the fight against the extremists. There's always a high cost to be being ignorant and behaving stupidly.
"A solution would be to require dog owners to buy insurance. Premiums for pit bulls would be very high, so this might hurt your poor dog owner even more than a licensing requirement."
Maybe, maybe not. In researching insurance costs for a new vehicle purchase about a year ago, I was very suprised to find that the $40K sports car carried about the same rates as my $30 4x4 pickup truck did. And I'm married/over 25, so no worries about those factors. I'm labelled a "safe driver" by my ins. co after over a decade of clean record driving. THAT mattered more than the vehicle I was purchasing.
I'm MUCH more likely to get into trouble with the car than the truck, f=mv^2 and all that. Yet my history, and how it fit in the pattern of other's histories provided the most important item here.
Which is, in a way, getting back to the profiling. It works when you have detailed histories (of the factors that matter). In a security line at the airport, they don't have that sort of time.
Now, a wild thought, would it be beneficial to actually do this like insurance does? The pay system for a "to the front of the line free" card isn't a bad idea, so long as the card can be easily (or quickly) determined to be valid, yours, and that the company that issued it will be held responsible if they're wrong.
The first two are the hard ones. How do I prove that the card is valid (cryptographic smart-card should be enough, with a db lookup, which can be done offline, and between when you make the reservations and when you get in line)? how do I prove in line that *I* am the valid card-holder. Hmm. That's the hard part.
A good thing about obligatory insurance for dogs (from a certain size upwards, say) would be to get reliable figures about damage caused by dogs. However, it would also involve quite some administrative overhead and dog owners would bitterly complain about the financial burden and the hassle. After all, dogs cause much much less damage than cars. The question might come down to how much the perceived vs. the real threat of pit bulls is exaggerated.
I like the quote you provided:
"'You think that terrorists aren’t aware of how easy it is to be characterized by ethnicity?' Kelly went on. 'Look at the 9/11 hijackers. They came here. They shaved. They went to topless bars. They wanted to blend in. They wanted to look like they were part of the American dream. These are not dumb people. Could a terrorist dress up as a Hasidic Jew and walk into the subway, and not be profiled? Yes. I think profiling is just nuts.'"
This is largely revisionist. The hijackers did in-fact stand out prior to 9/11 and their unusual behavior was noted by several authorities, but the failure to follow-through (lack of resources, unfamiliarity with procedure, and bureaucratic interference, not to mention the odd classification of Chechen rebels a non-threat) is what was found to be at fault, not the lack of information or failure to profile.
Profiling isn't something we can turn off, incidentally, and it usually works to our disadvantage if we think we can. For example, over the years I always found in pen-tests that if I could isolate a prejudice (either pro/con) in a group then I uncovered a simple attack vector. If you hate bananas, then I would hate bananas to gain trust. Take for example the nigerian fee fraud scam letters: people who think that africa is a bunch of lawless dictatorships with suitcases of money to launder, and who are motivated by the concept of "easy-money", are far more susceptible to the scheme. I did some research on this with a linguist, incidentally, and presented findings at a recent anthropology conference. I hope to publish the paper soon. In reverse, those who believe police officers always wear blue shirts with silver badges are easy to decieve...
So the quote by Kelly is well-intentioned but dangerous to believe at face-value. If we examine the particulars we find that the "blend in" aspect is incorrect. One of the issues is the concept of "different". The strippers claimed that the men appeared different and unusual to their regular clientele. But from the general perspective (far from being a regular in that environment or even familiar with the particulars) is to assume that if a fundamentalist muslim shaves and drinks they must suddenly disappear into the sea of secular club patrons.
My point is that if you *really* have an idea of what constitutes "usual" because you have an established investment in belonging to a group (and I don't mean facial hair and clothes, it's something much more significant based on a combination of meta information) then you are naturally prone (it's almost unavoidable) to spot the unusual. This cost is one of the controls against infiltration, incidentally, since it is generally non-trivial for a true outsider to assume the identity of another group. There's the famous story of the American spy who was caught in Europe because he ate his dinner with the fork in the wrong hand, for example...
Profiling is not a simple black/white exercise, and it requires care to work properly, but that's not enough of a reason to dismiss it entirely.
I don't think we are in disagreement. I'm not dismissing profiling at all. If you look above I actually argue that we know quite a lot about the patterns of Isalmic terrorists based in Europe and the US but to profile them effectively you need to have quite a lot of information. The cop in the subway station is going to be able to do this.
It is true that some hijackers did stand out but I think if we look back we'll find that the people who identified them as suspicious were looking for this type of activity and this type of person, had access to a lot of information and were skilled. They weren't your TSA people at the airline gates or your average cop on the street. And even then, as you say they didn't have enough information to make all the connections because of the bureaucracy. So, yes, maybe you are right that they don't completely blend in but they blended in enough to not be pulled before they accomplished what they wanted to accomplish.
"Profiling isn't something we can turn off". Agreed, as humans we're making generalizations all the time. You walk down a street and you are making judgements about the location and the people and how safe you are. If your information is good then those judgements will hold up most of the time. Take a look at Sally Merry's Urban Danger: Life in a Neighborhood of Strangers (1981).
Although I agree with some of the general points you make, I'm not sure I agree with the specifics. It is actually pretty easy for these groups to blend in in places like Vegas or NYC. What is usual? Forget the terrorists. There are lots of first generation immigrants, lots of language and different customs, lots of people in transit. In a small rural town they'd probably stick out like a sore thumb but in Vegas? Come on. Unusual is par for the course. They met in Vegas for a reason. They weren't stupid.
"The strippers claimed that the men appeared different and unusual to their regular clientele." Methodological point, did they really think they were different from everyone else and in what way? Did they think it at the time or only claim it with hindsight? Whatever, the case, nothing came of it. Maybe they really didn't quite fit any of the usual customer profiles but they got put in the "odd customer" profile but nothing came of it. What's an odd customer to a stripper in Vegas?
"In a small rural town they'd probably stick out like a sore thumb but in Vegas? Come on. Unusual is par for the course. They met in Vegas for a reason. They weren't stupid."
No, confusing the outsiders' view with the insiders' is fraught with pitfalls. This is the very nature of an anthropological approach to security. You and I aren't from Vegas and so we assume that everyone in Vegas is as random as we think they are. Yet, if you spend enough time talking with the locals in Vegas you find that they actually harbor a number of very tell-tale indicators to detect the locals from the visitors, the kind from the cruel, rich from poor, etc. for obvious reasons. I don't just say this lightly, I have been to several security conferences there and each time I try to spend enough time with a local to glean more information about what it really means to be a local and be accepted. The trained eye...
The attackers were smart, but it is very expensive to gain this level of understanding of how to truly disappear into a foreign environment. I would say they did not spend the money or time to achieve any significant level of actually evading profiling because they didn't need to -- they only needed to use our own shallow prejudice(s) against us, about what a terrorist would look like.
Here's a personal example: many years ago when I stepped off the plane into a small foreign country I looked around and saw a sea of people who seemed remarkably similar to me. I had a hard time profiling who/how to be secure (who I could trust with my money, my bags, etc). After a while of studying the environment very thoroughly I not only could tell what profession/caste people were from, but I also knew fairly accurately which local region people came from themselves (born and raised), their political preferences, and their level of education/income. And in more general terms, I knew who was friendly and who was a foe. None of this was obvious to the outsider. And again I found people's prejudices often held the key to their own weakness in profiling.
For me the issue is not that profiling is inaccurate as much as the fact that if we have disdain for learning how to be accurate we will continue to waste time and money on the least effective forms of profiling.
I still think we are splitting hairs here. I agree that even in a diverse, complex environment that the people who are used to living in that environment will have built up all sorts of subtle judgements (although their maybe a lot of variation among the "locals" in the sort of places we're discussing) . And probably most people have had the experience you describe of getting on a plane and going somewhere (maybe not very far away) and feeling a little ill at ease because he or she is not quite sure what's appropriate. But in places like Las Vegas and NYC and any number of US cities there are lots of tourists, travelers, people who are living in the city but are from elsewhere, maybe from another part of the US or maybe from a different county, people who don't even speak much English. So what would have made this group stand out in this situation as suspicious? I think it was very easy for them to dissappear into these types of environments because there are so many people who aren't "locals".
Even if they were suspicious what makes you think anyone would do anything? This takes us back to the earlier discussion about anonymity and accountability. One of the things that happens in these type of environments is that a lot of people mind their own business. We have the notorious bystander effect made famous by the Kitty Genovese case.
"For me the issue is not that profiling is inaccurate as much as the fact that if we have disdain for learning how to be accurate we will continue to waste time and money on the least effective forms of profiling.". I agree. Profiling, generalization, discrimination, judgement or whatever you want to call it is only as good as the information you use and your willingness to learn.
"So what would have made this group stand out in this situation as suspicious?"
The sfgate article (cited anonymously) above provides some insight:
"'For me, what I remembered most was the guy with the beard. You don't forget a face in this business.' All of the women, she said, recall the men the same way: quiet, well- groomed, polite, light drinkers -- and the opposite of big spenders."
Add "cheap, clean and polite" to the profile of terrorists. Just kidding. Think of it this way, how would the people entering the club know how to fit in unless they had been there before, or spent considerable time studying it? This is probably a bad example, since some might suggest that strip clubs themselves shouldn't exist. But if you look for simple and obvious gaps in the hidden/obtuse codes of "normal" behavior then you have a chance of increasing the cost to the attacker and lowering it significantly for the defender(s).
"Even if they were suspicious what makes you think anyone would do anything?"
Agreed. There's an issue of empowerment and responsibility, which is in some part is affected by a top-down policy of unitary security. Congress has recognized this and attempted several times to find ways to move towards better delegation, but this has been largely undermined by Nixon's, oops, I mean Ford's, um, Bush's advisors who favor an overly focused control center where they can manipulate, I mean interpret the data.
"how would the people entering the club know how to fit in unless they had been there before, or spent considerable time studying it?". Agreed but this is true for any newbie doing anything. Newbies are part of the scene. So maybe the question is what makes them stand out from other newbies or even a group of people entering the club out of curiosity? You go to Vegas and maybe you do something you would never do at home and maybe you'll never do it again but it's part of the Vegas experience.
Also, going back to the SFgate article, its not clear that they were newbies. The thing that really seems to have got them noticed was that they were cheap and didn't tip. I'm sure that isn't unusual either but I'm guessing that type of behavior is a pretty good way to get youself remembered by a stripper. If you work in any industry where tipping is important, you'll remember the good tippers and the bad tippers.
Here's an instance of a person, Clancy Prevost, who actually put 2 and 2 together and acted on his suspicions. From the account it could easily have been a different flight instructor who wouldn't have been as interested in the who, why, etc. of the student and would have just passed off the oddities.
Notice that the person isn't suspicious because he's a Muslim in flight school. It's that his appearance isn't quite right, he doesn't act quite right, he has no aptitude, he's vague and defensive about his identity, and he pays in $100 bills. A lot of the triggers are quite subtle and would only be apparent to someone who's engaged in sustained interaction with the subject in a very specific context with which he has great familiarity. And given the long association of the activity with terrorism, the instructor starts asking who is this guy and if we don't know who he is, why are we training him?
The Grounded Man by Dean Staley
"Newbies are part of the scene."
Well, a good point, but that's precisely what makes profiling a potentially useful method -- "newness" represents unfamiliarity with the secrets/patterns of belonging. To put it another way, there are those who know the secrets and those who don't. It's unfortunate when people think they should rely on obvious characteristics (e.g. outward appearances) as it shows they don't realize that they could use complex patterns and secrets instead, and it tends to open them up to attack.
You also have a point about an increase in density bringing with it a higher percentage of newness, but that just means more data is available (a more complex/rich map) to use for effective profiling. Small tight-knit groups can be harder to crack if they are aware of a particular threat (e.g. outlaw gangs such as the one infiltrated in the book Under and Alone), but their isolation can also make them more susceptible than a small group that blends with a sea of others. If you are successful in capturing data about all the groups in a larger pool, then matching comes prior to profiling with a secret-challenge. Matching thus adds an additional control point.
"its not clear that they were newbies. The thing that really seems to have got them noticed was that they were cheap and didn't tip"
I guess I just don't know enough about the culture of strip clubs but it seems to me that anyone cheap would be spotted right away, whereas "regulars" spend their money or at least know when/how to spend it. The bigger question is whether someone hiding in a stripclub would have any need to fear being turned into the authorities, and from that perspective I would guess Vegas is one place where that likelihood is highest.
The recent BBC programme "Travels With My Beard" (google for it, I can't link to it in a comment) also makes some excellent points about profiling.
Sorry, I don't see how you are going to easily profile terrorists in places like Vegas, NYC, etc. without having a lot of false positives and/or false negatives, especially if the people you are trying to identify are trying to be inconspicuous. What will be the tell take signs and who is going to be looking for them, where? It isn't going to happen during passing encounters in public or semi-public places.
Schneier.com is a personal website. Opinions expressed are not necessarily those of BT.