This one is making the rounds, but I figured I'd point it out too.

Security is always a trade-off: How much security am I getting, and what am I giving up to get it? These data-mining programs are not very effective. Identifiable future terrorists are rare, and innocents are common. No matter what patterns you're looking for, far more innocents will match the patterns than terrorists because innocents vastly outnumber terrorists. So many that you might as well not bother. And that assumes that you even can predict terrorist patterns. Sure, it's easy to create a pattern after the fact; if something identical to the 9/11 plot ever happens again, you can be sure we're ready. But tomorrow's attacks? That's much harder.

http://www.newsday.com/news/opinion/ny-vpsch213503428oct21,0,3927478.story

Schneier has it exactly right, as usual. Data mining sounds snazzy and exciting and high tech, but it's not the right tool for this job. In fact it's just about the worst tool possible—no real benefit, and a very high cost, both pecuniarily and in terms of further encroachment on people's rights and liberties and ability not to be treated a priori like a criminal etc.