Originally Posted by Spiff
The MIT study in question is a binary comparison of random harassment vs. profiling. Neither is acceptable.
It's like asking someone whether they would rather be hit over the head with a lead pipe or poked in the eye with a sharp stick.
I agree, I don't like either but, to me, the message of the paper is that a pure profiling system with no random elements is exactly the same having no security if an adversary takes the time the probe the system. Considering that an incremental increase in random screening brings such poor results (shown in the paper's simulations) I'd be more inclined to move toward better automated screening.