For obvious reasons, security, especially airplane security, is back in everyone’s thoughts and we had a lively discussion here recently about what methods would work.
While editing the wikipedia article about El Al, I found this fascinating article in Atlantic Monthly (pdf) on the web, which, while from 2002, is extremely current. The subject of the article, Bruce Schneier, a computer crypto geek and security consultant, explains why most of the approaches to increased secuity since 9/11 fail badly - that is, once they fail, they fail catastrophically. One example: all the additional biometric identification proposed now for our ID cards would not have prevented Mohammed Atta from flying his plane into the twin towers; he and his compatriots used their real identities. Similar things hold true for the tech-intensive attempts to keep bombs and weapons out of the plane. Once the attacker is beyond the barrier with them, the system fails badly.
A system that fails well, he holds, is instead one that does not depend on the attacker being unaware of it. He cites the awareness that, nowadays, passengers would risk their own life to tackle a hijacker. Knowing about this fact does not help the attacker, but knowing that old-style scanner cannot detect liquid explosives almost did. Armored cockpit doors meanwhile are a classic example of compartmentalization—they contain the failure after it has occured. And as the article nicely points out, methods like containment do not cause the widespread damage to civil liberties the way the data gathering approach does.
The third element he notes is the human one, where we come back to how I found the article. El Al does not (only) protect its planes by bomb sniffers and background checks. They talk to you before you board their planes.
Posted on 08/16/2006 at 09:04 AM 
Well, automation will always have a far larger margin of error and chance of failure than actual human testing. I do software testing for a living, and there it definitely rings true. We use both automation and human testing, and honestly, the automation is fairly limited. Since it can’t adapt and change, you have to tediously code every possibility into it to make sure it is efficient enough to be effective. However, there are always issues that you can’t forsee, that a live, thinking human will see instantly. The obvious advantage of automated processes is speed, but in the case of the airport I think that is pretty much being tossed out, so I see little real reason for it.
Datamining is about the same, there will be tons of false positives, and if you don’t have a thinking human being to properly interpret the data, false positives will lead to un-needed arrests, tons of time and money wasted, etc. Given, it does cut down on time, and using a search query to automatically find piece of data in a database matching it is far faster than a person sitting and sorting it, but again, without proper interpretation, it just causes more problems.
Seems to me that increased security measures and tools are just a way to decrease human legwork (and probably eventually cut down on payrolls and such). Sure they do help increase efficiency and accuracy, but if you reduce or eliminate the human equation, they actually become far less effective (since false arrests are a waste of time and resources).
One potential but unlikely scenario I can foresee with all the biometric data is that the human side of the equation is either eliminated, reduced to the barest minimum (just the guy that gets the email or whatever from the system to go make the arrest), or the data interpretation is outsourced to save money. I think the first two possibilities are more likely, I’d hope the decision makers in Washington would realize what a bad idea the latter idea would be, even if it could save lots of money.