[CALUG] Opinions on whole Disk encryption (for Linux)

Dave Dodge dododge at dododge.net
Sat Feb 2 00:11:56 EST 2008


On Fri, Feb 01, 2008 at 11:47:07AM -0500, David A. Cafaro wrote:
> Ok, I wanted to solicit any experience/opinions on whole disk  
> encryption.

I played around with loop-AES many years ago (I think I'm even
credited for a bugfix or two).  It will probably work fine for what
you want.  There's also dm-crypt, which I don't know much about but I
assume can be layered onto a block device in a similar manner.

One thing you have to consider is how you're going to supply the key
at boot time.  If you're worried about the server being stolen out of
its rack, then you can't really keep the key on the drive; so every
time it boots you'll have to walk up and physically type it into the
console or perhaps insert and then later remove a USB stick containing
it.

If you're running swap space you should encrypt the underlying block
device.  Since swap doesn't have to persist between boots, you can use
a new randomly-generated key each time the machine comes up and just
mkswap to reformat the resulting garbage plaintext.

The kernel and startup code itself can't be encrypted, unless the
bootloader supports encryption (and again you've got to figure out how
to get the key into it).  I don't know if any of the normal
bootloaders can do this.

You can probably find plenty of howtos and blog posts about all this.
For example here's some for Fedora Core 5 that covers a lot of it (I
think using dm-crypt) including the randomized swap setup:

  http://linux.ioerror.us/2006/09/encrypting-your-swap-partition-on-fedora-core/

                                                  -Dave Dodge




More information about the CALUG mailing list