Thanks for this information David --

However, it isn't entirely clear yet how the CryptProtectData /
CryptUnprotectData algorithm works when it's bound to the machine only
-- for the most part the document assumes user credentials to base keys
on, but in my case assuming fixed logons is not an option.

More specifically, I'm afraid that it's too easy to set up a ghosted
machine and then successfully decrypt data that was encrypted on the
original machine.  Just this morning this was proved by a little
experiment (and yeah, getting the ghosted WinXP it to work all the way
was a pain, but then, we're not sysadmins really).

Is this expected behaviour, or is my understanding not correct?

Thanks again,

Koen.