From: Hal Finney (hal@finney.org)
Date: Tue Jul 23 2002 - 22:05:27 MDT
We have occasionally discussed the use of profiling as a security measure
in airport security, attempting to identify those people who would be
more likely to be terrorists. Harvey Newstrom has often cautioned against
this approach, writing for example on May 23,
> As a security professional, I really must insist that standards require
> search of everyone or random searches. You cannot let guards try to
> detect the possible "guilt" of people by looking at them. They do not
> have that skill, and it is not effective enough to base a security
> policy on. Security profiling must be based on individuals, meaning
> behavior or situation. Groupism that includes or excludes whole genders
> or races will instantly fail because the bad guys then have a magic
> profile that will let them through. Just choose a person who looks
> right as your agent, and you get through security. Such a security
> policy would be invalid according to any security standards I know.
Today there is an article going around which illustrates the wisdom
of Harvey's advice, from
http://swissnet.ai.mit.edu/6805/student-papers/spring02-papers/caps.htm.
This paper shows that, under plausible assumptions, the profiling methods
currently being used by the airlines, under government supervision,
are actually *less* likely to detect terrorists than random searches.
The paper provides a method called "Carnival Booth" to allow terrorists
to exploit profiling systems and escape detection along much the same
lines discussed by Harvey above. The authors even offer evidence that
the 9/11 hijackers had used a similar method to assure that they would
be able to evade profiling and to carry out their plan without detection.
The article is very well written, for an academic paper, and the first few
sections are non-mathematical and well worth reading even for the layman.
It certainly calls into question the methods currently being used by
the airlines.
Hal
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 09:15:38 MST