Schneier on Security
A blog covering security and security technology.
« Airport Credentials Manipulated to Commit Crime |
| Find Out if You're on the "No Fly List" »
March 13, 2007
The Difficulty of Profiling Terrorists
A recently completed Dutch study of 242 Islamic radicals convicted or accused of planning terrorist attacks in Europe from 2001 to 2006 found that most were men of Arab descent who had been born and raised in Europe and came from lower or middle-class backgrounds. They ranged in age from 16 to 59 at the time of their arrests; the average was 27. About one in four had a criminal record.
The author of the study, Edwin Bakker, a researcher at the Clingendael Institute in The Hague, tried to examine almost 20 variables concerning the suspects' social and economic backgrounds. In general, he determined that no reliable profile existed -- their traits were merely an accurate reflection of the overall Muslim immigrant population in Europe. "There is no standard jihadi terrorist in Europe," the study concluded.
In an interview, Bakker said that many local police agencies have been slow to abandon profiling, but that most European intelligence agencies have concluded it is an unreliable tool for spotting potential terrorists. "How can you single them out? You can't," he said. "For the secret services, it doesn't give them a clue. We should focus more on suspicious behavior and not profiling."
Posted on March 13, 2007 at 5:42 PM
• 43 Comments
To receive these entries once a month by e-mail, sign up for the Crypto-Gram Newsletter.
" In general, he determined that no reliable profile existed -- their traits were merely an accurate reflection of the overall Muslim immigrant population in Europe. "
Well, thats your profile right there.
Um... OK who is wearing the captain obvious costume here?
If the bad guys take a guess that they are being profiled it is likely they will change their profiles to appear as someone that shouldn't be profiled.
Here we go 'round the mulberry bush...
Ditto, that's your profile right there - Muslim Immigrant.
"A recently completed Dutch study of 242 ISLAMIC RADICALS convicted or accused of planning terrorist attacks IN EUROPE..."
So European Muslim is not the profile, it is the _entire population_ from which we are trying to isolate the terrorists. If the study had included all those convicted or accused of planning terrorist acts in Europe, we'd be including native groups (e.g. ETA) and the "Muslim" part would drop out partially or completely. Or, to put it another way, the profile is "human being".
@ all who write "that's your profile right there"
Note a phrase from the article: "study of 242 Islamic radicals". Profile of being a muslim is a predetermined feature of the study. Or are you just friendly neighbourhood trolls?
Muslim immigrant would not be accurate. The report says "Arab descent who had been born and raised in Europe". Thats not an immigrant. Thats a citizen.
Dan, not necessarily; in some European countries, citizenship isn't tied to birth. Until 2000, children of Turkish guest workers, even if born in Germany, could not obtain German citizenship. Even now it isn't automatic.
Profiling is an art just as acurate as generalization. Money spent on profiling is money well spilled as it could have been put to better use elsewhere. Instead of making a profile of the terrorists and than say "he could have fitted the profile" why not do some actual counter terrorism measures.
Yeah, thats right, they havn't found any effective counter terrorism measures yet...
If you are born and raised in a country, you are *not* an immigrant.
It's just a theory, but if you are born, raised, and come to adulthood, perhaps for generations of a family, and you aren't a citizen, maybe you might feel ostracized. If there is a substantial population of such people, I'm not surprised if they are more likely to be radicalized than otherwise similarly situated "citizens" of the same country.
Being born there might not make you a citizen, but it does make you a non-immigrant. Anyway, the important thing is that they ended up with no common thread except the one they started with.
oops, I meant Dan, not Bob.
Profiling based on race or religion is bound to run into the sort of problems described in the article. However, the "look for a Muslim" advice that some are giving is the logical conclusion of the same incorrect premise that motivated the Dutch study: the belief that Muslims are more likely to be terrorists than the rest of the population.
Those who believe this are only slightly inconvenienced that the study didn't come up with "look for left-handed cross-eyed Muslims who attended university in Rotterdam". If the study did find that such Muslims were more likely than other Muslims to terrorize, even that conclusion would have been useless from a security perspective without the initial assumption that Muslims terrorize at a higher rate than the rest of the population. That is, this study could not possibly have produced usable data. So did the researcher running the study make that assumption? Maybe not, but he probably got funding from someone who did.
The article as a whole didn't make this particular misstep, but it did cite the "study". Incidentally, even though the study didn't give us any reason to believe that profiling based on religion works, it actually did give some indication that profiling within a religion could work on a limited basis, unless we are to believe that 25% of European Muslims have a criminal record. Even so, it was a bad study and a waste of everyone's time.
No, it wasn't "a bad study and a waste of everyone's time". Proving that profiling doesn't work is worthwhile as long as people continue to believe that profiling _does_ work. We need more studies like this to drive the point home.
...unless we are to believe that 25% of European Muslims have a criminal record.
Why should that be hard to believe?
Blacks in the US have a much higher percentage of criminal records than the muslims, so it's entirely possible, and probable.
"Proving that profiling doesn't work is worthwhile as long as people continue to believe that profiling _does_ work."
Umm... sorry, profiling does work, as well as any other system that is able to detect statistical differences in a given input data set.
What does not work is the way people are interpreting the data set they get out of the process, and tempering it against the limitations of the data set they put into the process.
As Jess pointed out the artical indicated that the data set used in the study showed that 25% had criminal records. But we do not have any norms for the population the very small sample set was taken from.
Unfortunatly there is no link to the study so we are arguing on very insufficient data (kind of like trying to see into the loft of a house from outside the front door).
So is a criminal record a good indicator or not? First ask is the input data set good? Then is the output set when put into the correct more general context good? Then ask is the indicator sufficient to use on it's own or does it need to be further combined with other indicators...
One indicator that keeps on comming up is education level, which is not mentioned at all in the artical. There is (percieved to be) a high degree of correlation between graduate level education and current European / U.S. Terorism (I first noticed this some time ago and did a little digging and commented on it in this blog).
Education level as an indicator has come to sufficient prominance that the U.K. Police / Intelegence services amongst others are starting to target U.K. Unis as areas to put resources into.
Now what if you combine the two indicators "Criminal record" and "Graduate education" and ran it against the data set used in the study and against various data sets that represent the population in genral.
If it gives you a high degree of confidence what do you do,
1, Do you arrest everybody that fits the profile
2, Do you ignore the data and carry on spreading your scarce recources thinly across the whole populance (app 50million in the U.K.).
3, Do you use the information to help you focus scarce resources in what may be more productive areas of enquiry.
Israel Torres made a point in that,
"If the bad guys take a guess that they are being profiled it is likely they will change their profiles to appear as someone that shouldn't be profiled."
But failed to mention that this is in itself an indicator. If you fit a given profile and then change it then this can given the right data set be seen. Which will then leave you with the unanswered question of "why have they changed" not a statment that "they are a potential terorist".
Or more importantly if you can recognise a group of people who are likley to be "radicalised" and potentialy turned into "terorists" or part of the "terorist support structure" then it can if used correctly help you focus your resources in recruitment prevention.
There are effectivly two basic types of terorist, those that are "self made" and those that are "recruited and made".
Self Made terorists are always going to be difficult to spot if they are reasonably well educated or naturaly wary/wily.
On the other hand those that are recruited and made are more easily spotted if you look in the right place at the right time, otherwise they would not get recruited in the first place.
So if they can be spotted they can be further monitored to show if it is likley they are being recruited (an indicator might be droping out of an academic course unexpectedly).
Or more appropriatly guided to sources of information / help / support which will reduce the likleyhood of their "radicalisation"
As part of the "making" process the potential recruit then gets effectivly taken into an organisation that has a support infrestructure. This offeres an increased number of other indicators that can be used, such as "association" with others who are on a watch list etc.
So having described ways in which profiling can / does work do I belive in it's use / usefulness.
It depends on several things,
First and foremost is the issue of the input data set. Aggrigation of the level of personal data required to make it work well scares me for all sorts of reasons.
Secondly invariably the input data set is bad for a variety of reasons and the Garbage In Garbage Out (GIGO) principle applies.
Thirdly the type of inferences people try to prove with these systems is often inapropriate irrespective of the input data set.
Forthly the type of inferences being investigated often are applied against inapropriate or insufficent input data sets, resulting in vary muddled reasoning and outputs that are overly sensitive to minute changes in input data set. Worse the data is often poluted well beyond the point of usefullness so GIGO applies.
Fithly the use to which this type of profiling is put is sometimes inappropriate (and these occasions are the ones that tend to make the headlines).
Therefor as a process for anti-terorist activity it produces sometimes (very) questionable probabilities in a group, and definatly not certainties in individuals.
Therefore its best and probably only effective use is to focus resources, either for more detailed intelligence gathering or proactive prevention activities such as providing appropriate resources to relativly large groups of people.
" ...unless we are to believe that 25% of European Muslims have a criminal record.
Why should that be hard to believe?
Blacks in the US have a much higher percentage of criminal records than the muslims, so it's entirely possible, and probable."
Yes, well, it's neither probable nor true.
The terrorists groups are also probably profiling those they use in the acts. Those that would be too obvious would be a risk to the operation and are kept out. The groups even have access to good filters thanks to the screening done at airports. Those bothered too much while innocently flying are left out of any operation.
This is not a random sample, but a selection that is intended to be close to the general population.
The problem with the "25% had a criminal record" statistic is that using it would still result in such high false-positive _and_ false-negative rates as to be completely useless.
> "A recently completed Dutch study of 242 ISLAMIC RADICALS convicted or
> accused of planning terrorist attacks IN EUROPE..."
Convicted or accused. HUGE difference!
For example, the number of suspects arrested for murder each year is several times higher than the number of actual murders committed. Law enforcement is quick in giving out numbers of arrests made (because it suits their agenda). Most arrests for serious crimes are reported in the media, but when people are released because the cops got the wrong one you rarely hear about it.
Now take into account the current hysteria regarding "terrah", the number of people who are being detained on questionable charges and who end up being convicted for unrelated charges (like immigration violations etc.) and I wonder how many of those 242 supposedly Osama loving islamic radicals were really up to no good.
> Until 2000, children of Turkish guest workers, even if born in Germany, could
> not obtain German citizenship.
Wrong. Of course they could obtain citizenship. But the number of those who actually tried and got it was rather small.
> Even now it isn't automatic.
Of course it isn't automatic. They have to fulfill a number of prerequisites and actually ASK for citizenship. Which many of them don't do, for a multitude of reasons.
Nothing gives me a buzz like shooting down ignorant political correctness with my breakfast.
Quick google search seems to suggest that >25% of US Blacks have a criminal record buy saying that a third of african americans are under the supervision of the US justice system.
I had the same thought that you did when I saw the figure, and said to myself, that can't be right, but I was suspicious when you didn't site any source.
With the internet and Google, there's no reason to stick your head in the sand like that.
Nothing is perfect. Looking for 'suspicious' behavior is also prone to failure...for some, everything, or the least likely thing looks suspicious.
We know it is Muslims that have blown up passenger and subway trains. That is why the study focused on Muslim terrorism...because they are the ones doing all the terrorism. Since the Muslim terrorists haven't been able to recruit other groups of people to perform their terror, the profile of the Muslim terrorist stands valid.
"So is a criminal record a good indicator or not? "
In this case I think not.
This was a study of people arrested and convicted of terrorism related offences who also happened to be Muslim.
So we have a four factors which will skew these statistic:-
1. The possesion of a previous criminal record probably contributed to thier detection and arrest. i.e. The police already had fingerprints, known associates etc.. So these people were easier to find than terrorists with no previous police record.
2. As members of a minority community they are sadly much more likely to be arrested for trivial crimes which would not result in an arrest for a member of the host (majority) population.
3. Such arrests could be a contributing factor in thier turning to terrorism.
4. According to the article the majority were from "lower or middle" class backgrounds. i.e. the same background as the majority of convicted criminals. (.. a statistical anomily due to the rich being able to afford better lawers?)
So all we can really conclude is that a Muslim is more likly to be a Muslim terrorist, and, much less likely to be an IRA or ETA terrorist.
>>So all we can really conclude is that a Muslim is more likely to be a Muslim terrorist, and, much less likely to be an IRA or ETA terrorist.
We can conclude much more than that. All of the ultra-violent terrorism in Europe in recent memory has been the result of Muslim terrorists, not the IRA or ETA. These terrorists have not recruited out of their group. That group, the European Muslim horde, should be the target of police surveillance.
Reading down the comments sequentially, I first thought that maybe you were joking. Then I realized that you're not.
Not all terrorists are (or were) Moslem: Eric Rudolph, Timothy McVeigh, Terry Nicholes, Eric Harris, Dylan Klebold, David Koresh, Michael Collins, Gerry Adams, Dail Eireann, Velupillai Pirapaharan, Mujika Garmendia, Jose Maria Arregi Erostarbe, Manuel Marulanda. Look them up if you're not familiar with any of them. They're just some of those who have committed relatively minor violent acts in an attempt to further a cause.
Then there's the group of people who did it on a larger scale: Benito Mussolini, Adolf Hitler, Augusto Pinochet, for example. They terrorized nations, often causing wars. Then we have the ones who were successful, known to us now as heroes, "Founding Fathers," martyrs, or freedom fighters.
To say that all terrorists are Moslem, or the converse, that all Moslems are terrorists is incredibly shortsighted, isolationist and xenophobic. Xenophobia and isolationism, on the other hand, have a much higher correlation to any crime, from simple racism, which is illegal in some circumstances, to violence and terrorism. If we should profile anyone, we should be profiling people who make blanket statements accusing large unrelated groups of people without any evidence.
Since 9/11, terrorism in the world has escalated dramatically and all from one group in society...Muslims. All the terrorist plots and actions one hears of TODAY come from this one group of people.
It is true that only .00001% of Muslims are terrorists, yet 99.9999% of terror TODAY is committed by them. For that reason the profile of the Muslim terrorist is valid.
Also, don't confuse the Founding Fathers with Al-Qaeda. Need I say, Al-Qaeda is a band of sickos that love to see death and misery and nothing beyond that.
@Bob -- Re: 99.99999% of terror TODAY being committed by Muslims. Are you aware of the Tamil Tigers? Hindu more-or-less terrorist group, currently very active.
Does 'terror' include near-random killings designed to intimidate and persuade people away from a course of action? If so, then inner-city gangs in the U.S. are definitely 'terrorists', and few of them are Muslim.
I mean, it's great to feel like you're part of something important, and a tremendoud world-wide struggle for the fate of civilization, but it's also nice not to have to worry about random flying bullets, too, isn't it?
FTA: "How can you single them out? You can't..."
What the study failed to find, is all of the detainees DID fit the following profile:
- All were left handed,
- All preferred pasta over the chicken for in-flight meals,
- None of the men cried during 'Love Actually'
- They all had odd socks
The "25% have a criminal record" statistic is useless even if valid because if you used it in profiling, you'd miss 3/4 of the terrorists. And that's before they figure out that you're looking for criminal records and ensure that only the ones with a clean record are carrying out the actual attacks, while the ones profiling would pick up keep security distracted.
Even if 99% had a criminal record, "Muslim with a criminal record" still wouldn't be very useful, as it would tag far too many people to really investigate.
I hereby formally declare that the investigative technique most commonly referred to as “profiling��?, hereafter and forever more be referred to as “set theory��?. There. Stop arguing.
I'm on the list! Yay!
It's clearly worthwhile to track the movements of every David Thomas. There's so few of them, and virtually all of them but me are terrorists, so I don't mind the extra scrutiny.
"Since 9/11, ..."
You're already off track. The very use of an arbitrary date that happens to support your statistics is weak. How about the USS Cole bombing? Sudan (the country) was just convicted today in US federal court of assisting in that. Why don't we start there? For that matter, why not after 9/11/2001? If you don't have a good answer to those questions, your whole line of reasoning is wrong.
"... terrorism in the world has escalated dramatically ..."
Compared to when? Escalation implies growth of growth, which needs two periods of time. It's true that terrorism *on* 9/11/2001 "escalated dramatically" from terrorism *on* 9/10/2001. It's not true that terrorism has escalated between the beginning of recorded history and 9/10/2001 compared to the period between 9/11/2001 and now. Once you define some time periods, you'll realize that either you're being arbitrary, or your statistics are wrong.
"... and all from one group in society...Muslims."
That's completely incorrect. Not all terrorism since 9/11/2001 has been committed by Muslims. Your statement is not a defendable opinion but an incorrect fact.
"All the terrorist plots and actions _one_hears_of_TODAY_ come from this one group of people." [emphasis added]
Finally, a correct statement for a person with a limited view that's restricted to common Western media. (Not to say that you necessarily are and there are other narrow viewpoints for which this is true.) Just because you haven't heard about any other terrorism, that doesn't mean that it doesn't exist.
The sad fact is that there's lots of terrorism in the world. CNN likes to blame Al-Qaeda for all of it, but they're sadly mistaken. There isn't even a common thread that all Muslim terrorism is targeted towards American citizens or even the west -- take a look at the newspapers in India.
We don't even have a clear, universally accepted definition of terrorism yet. What is a terrorist? How or why is a terrorist different from an organized crime boss? Does a terrorist float when bound and thrown in a body of water? It's really hard to use statistics to prove a point when you haven't definited your point. That's what you're doing and that's what anyone who profiles Muslim terrorists is doing.
Computation of statistics is a complex and precise science. For each person who knows how to use them correctly, there are many hacks. The hacks just throw around flurries of numbers to make their opinions seem real. The respected scientific community would ostracize anyone who did the profiling that the US government is using here.
@Lawrence, "Proving that profiling doesn't work":
This study proved no such thing. From the information in the article, I can discern no testable hypothesis, but the only thing it seemed to be looking for were other categories into which Muslim terrorists might fall, preferably which aren't strongly correlated with being Muslim but which are strongly correlated with being a terrorist. Of course, its entire data set was both Muslim and terrorist, so this was destined to be a wasted effort without reference to known facts about the larger population, which didn't seem to be considered.
You state, "don't confuse the Founding Fathers with Al-Qaeda." I too am proud of our country. But our history is littered with acts of terrorism, ask any Native American or any descendent of a slave kidnapped in Africa and brought over on a slave ship.
Those easy examples of terrorist acts are small change compared with our Fore-Fathers use of biological weapons of mass destruction to decimate entire populations of American Indians. Smallpox, tuberculosis, plague, and cholera were just some of the diseases deployed on Native Americans by European warriors and later immigrants. Estimates run as high as 95% of all Native Americans were killed in the century or two following Columbus's arrival. [Jared Diamond. Guns, Germs, and Steel. Norton, 1997. Page 211. And Chapter 11. Really great book. Highly recommended!]
In summary, your debating opponent has a valid point. Winners rewrite history to mask the evil they committed in achieving victory. Indeed, losers do too. Japan's erasing of its enslavement of Korean and Chinese "Comfort women" is an easy current example.
Just to back up Stephan, it seems the correspondent has been looking the wrong places. Non-Islamic terrorism continues, it just doesn't get the same press:
In the UK we've had a recent case of mailbombing connected to transport policy(!) and a series of bank robberies and shootings related to extremists in Northern Ireland
ETA's bombing campaign flickers in and out of life. Looks like it might be calming down again.
In Greece the ultraleft recently fired an anti-tank missile at the US embassy (it missed.)
The Tamil Tiger tactics (largely copied by Al Qaeda in Iraq) continue. Most recently 21 people died in an attack on a bus.
Everything I've heard about terrorists/suicide bombers is that they're a great deal like normal people. It's certainly possible to talk a normal person into joining the military, and imho, part of the hook is doing something important (i.e. something that is likely get someone killed) for their country. Dying for sure for a religion is only a medium-sized step away from that.
I suspect that those who recruit terrorists are *not* normal people, but perhaps in ways which are hard to profile. In any case, I haven't heard of anyone trying to profile terrorist recruiters.
Another point to consider is you say "_one_hears_of_TODAY_" and I'm not sure there isn't a greater agenda on the part of politicians, law enforcement and the media to discuss every single "potential plot" in great detail (if for no other reason than because people seem to be clamouring for information, and maybe for more self-serving reasons of their own). In contrast, I don't recall very much discussion of terrorism that hadn't actually succeeded in the case of, eg, the IRA. (We heard lots of press releases from Sinn Fein but precisely because they were claiming to be not involved with the military wing, they weren't talking about plots.) The only act of terrorism I've been involved with was an evacuation of a film cinema about twenty-odd years ago after an "IRA" bomb threat. Given I was in a nowhere town it was probably more likely some local idiot phoning in a fake threat, but the event didn't even make the local papers. I wonder if it would be considered newsworthy if it were to happen today?
Hmmm - European, Muslim immigrants? And yet the TSA continues to feel up toddlers, grannies, and any females that catch their fancy. Tax dollars at work.
Derf, are you paying attention here? They're European and Muslim because the study was looking at European Muslims convicted of terrorism in Europe. If you weren't either of those you weren't studied. And most of them aren't immigrants: they're European born. To be an immigrant you usually have to come from somewhere else?
@dave, indeed it was British policy to deny our local terrorists the "oxygen of publicity". IRA spectaculars therefore only got mentioned if they were spectacular and I fail to remember a single instance where a large group of Irish men were arrested, followed by extensive secret briefings about their alleged plot. (The former was common, the latter not.) The situation in Northern Ireland was much more like present day Baghdad, where the authorities are seeking to play down the extent of terrorism in an effort to normalise the situation, than in the US where imagined threats are create to suit political agendas.
I do not understand -- what is meant
with profiling does not work. In my understanding profiling is basically search
for something common (characteristic feature(s)) to some group -- so while saying that profiling does not work
what is meant with it:
There are no such common things in the group or the group is so large that it is
not useful? If it is the first then I think it is not true, in this case they were
all male European Moslems, 16-59 and
these are just 3 features, how about
the way they spent leisure time, media
sources prefered, area they lived in, etc.? If it is the second one--
it is not useful because it is too large--
the question is then: it is not useful for what, for whom, in what situation and how usefulness is defined.
Surely you're not suggesting that there's a valid reason for the TSA to be sexually assaulting elderly ladies?
Of course there is. What better way to understand criminal behavior than to exhibit it institutionally, preferably as flagrantly as possible? Especially when they're recruiting bottom-feeders who get up every day thinking "Man, how did I luck into getting PAID to do this?"
The first step to removing (or at least diminishing) pandemic crime from society is to remove criminal behavior from those in authority - or to remove authority from those who exhibit criminal behavior. That this has not happened since December 2000 implies rather strongly that civilized concepts of "criminal" are either ignored or successfully spun via propaganda by those in power. The extent of and delay to fixing that problem will determine at least as much as anything else when, or whether, the America we knew as children will ever be restored.
Schneier.com is a personal website. Opinions expressed are not necessarily those of BT.