Matt Blaze: How unpredictable screening helps terrorists

Re:Matt Blaze: How unpredictable screening helps terrorists by pedrocr on Thursday December 31, 2009 @04:10PM (#1024984)

The problem is that with non-random screening the terrorists can be much more efficient. They can get a pool of candidates together and send them on flights without any explosives on them. They can then find out which of the guys get screened at a rate less than chance and send those on the actual attack. That's why random screening is the best system possible because any other system would have to be perfect, otherwise any flaw can be detected on dry runs and exploited for the attack.

This was thoroughly explored last time the TSA tried to be smart about screening and implemented its Computer-Assisted Passenger Screening System. MIT article exploring it:

http://groups.csail.mit.edu/mac/classes/6.805/student-papers...

[ Parent | on HN ]

Re:Matt Blaze: How unpredictable screening helps terrorists by zmimon on Thursday December 31, 2009 @07:21PM (#1025253)

It seems to me that the right approach should always have a significant random element, not as a deterrent, but as a check on how well the non-random component is working. The random part will examine in depth to find anything that should have been found in the non-random part but wasn't (say, body searches to find large metal objects making it through the x-ray screen at the airport). Without that you would be blind to defects in the system.

[ Parent | on HN ]

Re:Matt Blaze: How unpredictable screening helps terrorists by alterego on Thursday December 31, 2009 @04:20PM (#1025007)

I think this article isn't talking about random vs. profiled (which is what the mit paper is about), but random vs. 100% (where there 100% has been analyzed carefully to be sufficient to detect large bombs, etc.).

[ Parent | on HN ]

Re:Matt Blaze: How unpredictable screening helps terrorists by pedrocr on Thursday December 31, 2009 @04:24PM (#1025014)

He was hand-waving at the end about the 100% but that's not a solution. They are doing it randomly because 100% is simply not possible. My point was that in a situation where 100% is not possible, as in real life, any system is potentially worse than randomness.

Maybe this just means we need to come up with screening mechanisms that can be applied to 100%, but given the current capabilities, choosing random screening is better than profiling.

[ Parent | on HN ]