"targets in database" doesn't read as "PRISM case numbers" (to me). In fact, it sounds specifically different in order to indicate unique individuals.
But who really knows, I guess.
But who really knows, I guess.
It's easier when you can associate IP addresses with multiple accounts, of course, but there are a lot of traits (and I'm sure NSA has more than I am aware of) that can be used as similarity metrics to create a probabilistic hierarchy of account agglomeration.
Throwing clique analysis into the mix, which is of course where the most important analysis goes, also helps establish the probability that multiple accounts may be controlled by one person.