8

Say I have two partitions on a Linux system, a root partition (/) and a /home partition. Both of them are encrypted with LUKS.

The /home partition needs only to be decrypted if / is unlocked and mounted. At boot time, I enter the passphrase for the root partition. Now I have found several possibilities to avoid having to enter multiple passphrases:

  • Use decrypt_derived from Debian. This retrieves the (master?) key from (secure?) memory and uses that as passphrase for the second partition. It has been mentioned on Unix.SE.
  • Create a key file from random data, store it on the main partition and use it as passphrase for the second partition. This is easy to setup.
  • Somehow read the passphrase and write it to the standard input of cryptsetup. This means that the passphrases need to be the same for both partitions.

What are the security implications of each method and which method is recommended?

Improve this question
0

2 Answers 2

Reset to default
7

decrypt_derived is slightly better than the keyfile approach for these reasons:

  1. You could accidentally store the keyfile with the incorrect permissions and thereby leak the contents to non-root users on the system, whereas the master key for a LUKS permission can only be accessed by root.
  2. You could accidentally delete the keyfile, whereas the master key can't be accidentally changed or removed.
  3. If the root filesystem gets corrupted you could lose the contents of the keyfile, whereas the master key would not be affected by this.

Basically, I would use decrypt_derived if it is available (i.e. I'm using Debian/Ubuntu), but wouldn't go to the effort of installing it manually if it's not. In either case I would still add an additional passphrase key to the home partition in case the root partition gets corrupted and the master key is lost.

I certainly wouldn't use the third method you suggested, since it is more work and you want to avoid storing the passphrase anywhere.

Improve this answer
2
  • 1 is not an issue as the keyfile in question is put into /root and read-only for root only (400, root:root). 2 can be prevented by a good backup, the master key can also be "accidentally" wiped using wrong cryptsetup or dd commands. 3 is also avoided by making backups and/ or making the rootfs read-only. The encrypted partition already has a passphrase, losing the keyfile isn't a big deal either.
    – Lekensteyn
    Jul 26, 2012 at 9:55
  • @Lekensteyn Sure, those issues can be addressed, hence why I said "slightly" :-)
    – mgorven
    Jul 27, 2012 at 6:01
0

Consider using decrypt_keyctl instead because:

  1. LUKS2 compatibility. Don’t use decrypt_derived blindly on new installs. It might result in having a static error message instead a secure derived passphrase in your key slot.
  2. Passwords expire by default after 60 seconds. The derived passphrase would be always available to root.
  3. It have almost all advantages decrypt_derived brings over keyfiles. The downside is, that you could not use it for disks attached later to the system.
Improve this answer

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .