There are two basic approaches to mitigating massive data collection efforts: reduce your digital footprint, or add noise to your digital footprint. The latter is far more effective, as you fuck them in three ways: you hide your data, you have plausible deniability, and you spam their databases with crap.
However, the "data flak" approach has limits. It is only doable with discretionary data, and even that might be problematic. But it would work well to make access logs useless. While I don't want the system to spam hacker news with nonsense posts about building nuclear weapons, I might be able to harmlessly incorporate intentionally explosive phrases in a post just to annoy them (hah, see what I did there?) For things like flights and credit card usage, there is no way to introduce flak, or really minimize exposure.
Of course, the best mitigation technique is to raise Cain over this and get the entire system dismantled. There is simply no excuse to be doing this sort of thing. The fact is that it would be pretty hard to organize an attack and not trigger (local) alarms. Local police have a feel for the community, for what's what. They can call in the feds if they need to.
Systems like this are going to always have three serious problems: a) bringing the hammer down on the innocent (false positives), b) failing to detect actual plots (false negatives), c) be abused (used as a weapon for politically or personally motivated attacks). And in the long run a) and c) are going to grow monotonically, and that is not acceptable. It's one thing to have the asshole local cop bust down your door for no good reason, it's quite another to have a bunch of feds fly in to nab you based on some data in their computers. They've never met you, don't know you, don't know your relationships to the people around you - but based on data will ruin your life. That's a nightmare scenario if I've ever heard one!