The Problem with Profiling

August 18, 2006

Comments

With the recent U.K. terror plot to blow up airliners with liquid bombs, there has been lots of discussion on the radio talk shows (at least the ones I’ve listened too) that make a big deal about profiling and how it would be an effective screen against terrorists. There is only two problems with this and those problems were born with the names Don Steward-Whyte and Oliver Savant, two white converts to Islam.

So far in all the bloviations I’ve heard on the radio, nobody seems to mention that these two gentlemen are white and converts to Islam and could very well be Islamic extremist. Everybody focuses on the names and from there assumes that all the people arrested in connection to the plot are Pakistani (and to be sure, most of the people arrested are of Pakistani descent).

The problem with profiling is the Carnival Booth effect. The strategy of terrorist could use to circumvent a profiling system such as CAPS (see the link for a description)¹ is to send people through the airline screening system and see which one’s get caught. Eventually, the terrorist organization will learn enough about the profile to circumvent it, possibly with white converts to Islamic extremism such as Stewart-Whyte, Savant and also Johnny Walker Lindh. Similarly terrorists groups might look to females and also female converts.

Further, people get upset when they see that grandma has been selected for additional scrutiny or that the mother with an infant was selected, etc. The problem here is that random searches are a way of defeating the Carnival Booth strategy. With random screening there is no way to learn what will allow you to slip past the screening. And for something to be random, this means that periodically grandma might have to undergo further scrutiny, Jr. might have to have his bad of toys inspected, and so forth. After all, if you start exempting certain passangers then the random searches are no longer random.

Some people, e.g. Ann Coulter, have argued that the random searches and lack of profiling is due solely to political correctness. Maybe they are right, but there is a non-politically correct reason to reject profiling, at least by itself and with the simple form many advocates seem to favor. Making profiling more subtle so that those who are profiled are unaware of their being flagged is one solution. Since people flagged are unaware they have been flagged there is no way to learn what the profile is. This could be done by devising better methods of scannning both checked luggage as well as carry on luggage. If this can’t be done, then profiling isn’t all that helpful, and in fact could provide a false sense of security when the reality is quite the opposite.

[Note: Initially I was in favor of profiling, but have since changed my mind as this strategy looks like it is too easy to beat. Note that this isn’t any kind of nonsense PC consideration, but purely a strategic/security issue. Do try to keep this in mind when posting comments.

Also, note that CAPS and CAPS II are no longer in effect, the new system is called Secure Flight, but from what I’ve read so far has been built off of CAPS and some of CAPS II, so it again looks like a profiling system. Funny, those in favor of profiling seem unaware that profiling has been going on.]
_____
¹ From the link above,

This transparency [knowing you’ve been flagged by CAPS] is the Achilles’ Heel of CAPS; the fact that individuals know their CAPS status enables the system to be reverse engineered. You, like Simonyi, know if you’re carryons have been manually inspected. You know if you’ve been questioned. You know if you’re asked to stand in a special line. You know if you’ve been frisked. All of this open scrutiny makes it possible to learn an anti-profile to defeat CAPS, even if the profile itself is always kept secret. We call this the “Carnival Booth Effect” since, like a carnie, it entices terrorists to “Step Right Up! See if you’re a winner!” In this case, the terrorist can step right up and see if he’s been flagged.

6 responses to “The Problem with Profiling”

yetanotherjohn

August 18, 2006 at 3:50 pm

Wouldn’t a policy that combined profiling of most likely and random search be the best? Especially compared to a policy that limits the number of people of a certain profile that can be searched?

You are saying that because roughly 10% of one terror plot would not have potentially fit the profile we shouldn’t profile. This ignores the other 90% that would fit. It also ignores that the need to reach beyond their core group would also increase the chance of terrorist plots being found out. There is a reason why the majority comes from the core group, that is where most of the people who agree with them come from. The chances of trying to recruit someone who isn’t as committed is lower and the chances that they change their mind somewhere along the line is greater. And lets face it, being a suicide bomber (at least a knowing one) is about as committed as you can get.

Further, the need to reach out beyond the core group provides greater opportunities for the authorities to infiltrate the terrorists. While the intelligence services might be a little thin on the ground when it comes to loyal agents who fit the profile (and would pass whatever background check the terrorists might apply), they are probably better set up for white males. So terrorist group practices the carnival booth approach, determines that they need a blue eyed, blond hair male. Guess who’s coming to dinner then when they find a sympathetic guy who wouldn’t be hit by the profile?

I’m not persuaded by your reasoning that profiling should be thrown out as an option.
John Burgess

August 18, 2006 at 4:02 pm

In addition to the weakness of profiling that Steve points out, there’s yet another: institutionalized blindness.

Because attention and resources will be spent on checking those who match the profile, an awareness will creep in on the profilers that they’re doing what they’re supposed to be doing. That which isn’t clearly spelled out in regulation will be neglected in practice. And if there’s a hit on the resources, the non-regulatory actions will be the first to suffer.

Other than those coming from allied countries (which generally take part in the Visa Waiver program anyway), those likely to be profiled coming from suspect countries are already being screened and screened again when they apply for visas.

Now, the “profile” population from a country like Saudi Arabia, Syria, or Pakistan has the application reviewed by–at minimum–State, FBI, CIA, and DHS. In addition to a State Dept. Consular Officer doing the now obligatory visa interview, that officer is now accompanied by a DHS officer. This is by law promulgated in 2003.
Wayne

August 18, 2006 at 4:46 pm

Yetanotherjohn

Excellent post

The idea that any method can be counter is not a good reason not to do it. The idea is to make it harder on the terrorist and to slow them down. Catching a few why they are finding ways to get past our defenses is an added benefit. Profiling is not an end to get all and cares needs to be taken not to rely on it too much but it has and can be quite beneficial.
We need to concentrate most of our resource in areas where we get the greatest return. Profiling gets better return than random searches. Not that we need to get rid of all random searches. Offensive operations and Intelligence are areas where we should concentrate our resources. We will bankrupt ourselves if we try to cover all defensive measures.
Steve Verdon

August 18, 2006 at 5:13 pm

Wouldn’t a policy that combined profiling of most likely and random search be the best? Especially compared to a policy that limits the number of people of a certain profile that can be searched?

The problem the authors of the linked paper note is that there seems to be an upper limit on the level of inconvienence that people will put up with. Hence the number of people who can be searched is limited and since searching either profiled or randomly selected people impose the same costs in terms of slowing things up, the answer to your question is, “No.”

You are saying that because roughly 10% of one terror plot would not have potentially fit the profile we shouldn’t profile.

Not quite, I am saying that these people present a problem for those who favor profiling that is never ever addressed. Further, there is a very logical reason to suspect the efficacy of profiling outside of the short run. You are basically peeling off one part of the argument and ignoring all the rest.

There is a reason why the majority comes from the core group, that is where most of the people who agree with them come from.

People who are muslim come from all over the world. You can find muslims who don’t look the slightest bit “Arab”, in fact there are hundreds of millions of them, IIRC, in South East Asia where you have at least two terror groups there. So this “core group” argument is not accurate.

I’m not persuaded by your reasoning that profiling should be thrown out as an option.

Funny, you haven’t mounted a single argument against my position. Your claims about the core group are false. For example, are you ready to profile Michelle Malkin…after all, Mindenao has quite a few muslims and its very own extremist group. In fact, IIRC, the Bojinka plot was formulated by AQ members in Manila and quite possibly included member of Abu Sayyaf.

Also, profiling is fine so long as the person profiled doesn’t know it. If they don’t know they have been flagged via a profile then they can’t learn what the profile is.
yetanotherjohn

August 18, 2006 at 7:13 pm

Steve,

I think you are being a bit disingenuous in your response.

1) Limited number of “searches”. So half selected by profiling and half by random would impose additional delays how? I didn’t call for doubling the number of searches. If we are only doing one search, however we decide to do the search, then we have a whole other set of problems.

2) Since what I am suggesting is to do both, then I think I am addressing the issue of the 10%. If we can come up with a way to weed out the 90%, then we have the random searches to go after the 10%.

If we have 100 people on a plane, two of whom are terrorists, then I have a very slight chance of catching either terrorist. If on the other hand I can use profiling to increase my chances of catching one of the terrorists, I have made progress. If they have managed to recruit two people who don’t fit the profile, then I have lowered my chances of catching them. But in the part of my argument that you seemed to have ignored, I have increased the terrorist cells chance of detection by getting them to recruit more of the people they are recruiting today. The profile has also made it easier for us to potentially infiltrate the cell, since the terrorist are more likely to recruit different people than they are now.

If one of the two terrorist in the hypothetical does match the profile, then I have increased the chance of catching the terrorist tremendously. If both match, then I have increased it even further.

2) While profiling can be determined, every attempt and determining the profile raises the chance of detecting the attempt. Finding the first thread to start pulling is a key part of stopping terrorist before they hit.

3) I really thought that pointing out the greater risk to the terrorist cells being detect or infiltrated was an answer to your carnival effect. Yes the terrorists can potentially determine the profile and then look to recruit people who don’t match that profile. But that recruitment beyond who they have been recruiting would leave them more vulnerable.

4) Lets look at pictures of the 9/11 hijackers. Gee, I think I could identify a core group profile from looking at these. I suspect I could do a core group profile out of England that included young Pakistanis that would have highlighted most (admittedly not all) of those arrested so far. Yes there are Muslims who can match any racial profile in the world (though they might be a bit light among Kung tribesmen). But not all Muslims are the enemy. We are most concerned about those who are planning to die in the terrorist attack.

Did you notice that the latest foiled plot was looking only for the death of those on the plane. No indication of trying to hijack the plane to crash it into a building or hold the passengers hostage until the US is out of Iraq or Guantanamo detainees are released. Why? Because enough people who fly since 9/11 have had to ask themselves what would they do in a similar situation and sitting quietly in their seats is not the general answer.

So you have to find someone willing to end their life violently. Look at the suicide bombers and there certainly seems to be a core group who is willing to do that. I suspect for every instance that you could find a suicide bomber who didn’t match the profile, I could find ten who do. Its not that profiling would eliminate all the threats, rather that it would seem to address one of the most dangerous ones.

Now there certainly can be the case of an unwitting suicide bomber. They would likely be less effective since they would not be consciously trying to further the plan. They would likely be easier to detect as if the little old lady found a bomb in her purse, she might mention it to someone before take off. I’m not sure random searched makes the unwitting dupe that much more likely to be caught either.

5) The combination of random and profiling would make the detection of the profile harder. At a minimum, just because someone was stopped doesn’t mean they fit the profile and just because the person wasn’t stopped doesn’t mean that the profile wasn’t triggered (e.g. having an air marshal sit near the suspect).

6) I really am not sure why you think I didn’t address your arguments. It seems to be that a profile can be detected and thus potentially evaded. Granted. But what about my argument that trying to evade the profile could make it harder to recruit, makes it more like to have someone have second thoughts and provides a better chance of infiltrating. So even if the profile becomes known, it still has a barrier effect that complicates the terrorists job and eases the counter intelligence job.

Now obviously profiling can’t be static. If you were looking to profile for a hijacker in the ’70s, you wouldn’t use that same profile for a hijacker today (or more dangerously a suicide bomber today). If profiling did nothing more than delay current plans, force terrorists to discover what the profile is, then force the terrorists to recruit non-profile terrorists, it will have done more than the current random searches. If you want, randomly switch the percentage of profiling and random searches. If you made profiling only 1% of the searches, you would likely make the terrorists have to reconsider their plans.

Just because there exists a detection technique, that doesn’t mean the whole idea should be thrown out.
RJN

August 18, 2006 at 8:58 pm

yetanotherjohn:

You have it right.