datamining and probability mathematics makes even anonymous dataset personal
"Scientists found that four pieces of personal data—seemingly innocent things, like where you ate lunch last week or how much you paid for your new hat—is all it takes to steal your identity. Although we often think of identity theft as a hacker’s game, a new study shows that even public datasets used by scientists and pollsters are exceedingly vulnerable to being compromised.
After examining 1.1 million credit card records, the researchers found that they were able to reconstruct 90 percent of shoppers’ (supposedly anonymous) identities based solely on factors like where they had shopped and how much they had paid.http://www.vocativ.com/culture/science/credit-card-data-t...
that is why you have tools to switch data in datasets so that it is still correct in the form that is needed but it is impossible to have any real identity come out of it