I'm pretty sure that joemike has posted that full disk encryption in APFS High Sierra encrypts even free space.
FYI in case anyone is wondering "why would you encrypt free space?", it is the ultimate goal of data encryption to prevent someone without the key from gaining
any knowledge of the contents, or even lack thereof. IE you shouldn't be able to identify the one hard drive with all the data on it compared with the other nine in the bin that are freshly formatted. Someone trying to sneak a hard drive out of a secured area that encounters a stack of removed drives would positively
love to be able to identify which of the ten they need to nip off with for analysis, rather than be forced to try to smuggle all ten of them out the door.
I ran into this
exact issue several years ago while working at an AASP retailer. A vendor sent us a demo unit of a new "hardware encrypted drive". It was a USB external drive, but it also had a USB-A port on the back, and came with what looked like a flash drive. But it wasn't a flash drive, it contained the key to the external drive. If the key was not plugged in, the drive was encrypted. Pulling the key was like locking the drive.
Seemed like a nice idea, and my boss gave it to me to check out. I quickly discovered that the drive used a 32 bit block encoding scheme
that had no salt for the block number. (it was
not a stream cipher) This meant that every four bytes were encrypted together, the same way. So regions of the hard drive with no data on them were
easily identified by a pattern of the same four bytes, repeated over and over.
The first thing that stuck me is I could immediately identify a drive that at one time contained information vs a drive that had been wiped, because empty blocks were so obvious. But it then occurred to me I could also start identifying structures on the drive such as partition scheme, number of partitions, etc, because of areas of blocks that were zeros or not zeros. ("this is a fat-32... that one is HFS+" and so on) And since some areas in the partition tables were known fixed values, I could see the encrypted patterns at those locations and knew what that pattern stood for, allowing me to identify it anywhere else it occurred on the drive. ("this is what a '1' looks like. this is the string 'AFPS'", etc)
Allowing me to get ANY of this information would be cause for concern. And in the business of crypto-analysis, every nugget of information your protection "leaks" is another crack in the armor that may eventually lead to a catastrophic failure. (go read up on Enigma, a great example of that IRL)
I took the time to try to contact the manufacturer, and engaged in a short discussion with them. I could quickly summarize our discussion with "They didn't consider it to be a problem."
Needless to say, we didn't stock any of these new drives.
