Problem with that comparison is that while a government could make and store nukes without people knowing, it's hard to start exploding them without attracting attention. Accessing a database, on the other hand..
Right so what I'm imagining is basically (for lack of a better term) a QA framework that tests that the instrumentation for tracking access...etc is legit. You could say that blowing up even one nuke without proper oversight is unacceptable and thus we have mechanisms in place to make sure the right bells and whistles are trigged.
EOD launching a nuke and using PRISM data are keystrokes on a computer.
No - launching a nuke results in a thermonuclear explosion detectable from space - not just keystrokes on a computer. Keystrokes on a computer can be hidden, those mechanisms can be by-passed - and no satellite from space is going to detect all that and alert us to the problem in the reliable and impartial way that would be necessary.
I see. I understand that a little more. I still disagree, however. There's a huge incentive for the government to keep those mechanisms working, and we'd all know it if they didn't. That's definitely not the case with the NSA. How will we know there isn't some secret backdoor?
Sure they're keystrokes, but if the software/audits/etc. are run by the same people who might want to abuse it, it doesn't mean much. And put yourself in their shoes: if you abuse a secret database, you likely won't get found out. If you abuse the nukes you most likely will.
Right so it's more difficult to do, but you can separate out the testing team or do other things to be rigorous about detection. At the very least you can make it more difficult than sitting down at a terminal and hitting go. One sort of simple solution I can think of is having passwords just like nuke authcodes that have similar restrictions that are passwords for access and decryption...etc.
And if the government (right at the top) decides they want to start accessing it regularly without the public knowing (like the current situation), they control those separate testing teams, and can authorize secret access at any time?
I think if it involved Presidential and Congressional approval and had things like exact query logging and required declassification prior to prosecution than that would be reasonable.
