Loop-AES, security concerns, stability of file backed loop-aes

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Dear mailing list readers,

I have spent the last days to evaluate whole disk encryption possibilities
under Linux. As far as I know there are 3 ways: cryptoloop, dm-crypt and
loop-aes.

My task is to backup gigabytes of medical and financial data stored on a
couple of smb file servers to a remote machine, whose security is unknown.
Lets assume the worst case that the crypto container on this backup machines
is world readable. I will use rsync, so only changes of the crypto container
are transferred on a daily basis.

It appears to me that loop-aes is the only one choice under linux with real
secure implementation of strong encryption. However, I am no crypto analyst
and would love to read some professional comments about loop-aes. So, my
first question is, if somebody knows a link to a document which deals with
this?

Also, I have questions related to file backed loop-aes encryption.

If I use ext3 on top of a file backed loop device, I understand that the
consistency is in danger because writes are reordered of the underlying fs.
What I dont understand is the claim, that ext3 (top) <-> loop-aes <-> ext3
(underlying with data=ordered or data=journal) should work. Why is the
assumption correct that the underlying ext3 preserves the same write order
of the ext3 on top?

I found some other notes about deadlocks while using file backed loop-aes.
On Linux Kernel mailing list it is claimed that GFP_NOFS is cause of
deadlocks. Ext3 uses this call, ext2 not. The author of loop-aes just used
the phrase "it should maybe work" if ext2 is fs on top. So, does somebody of
you know the stability of this scenario (file backed loop-aes with ext2 on
top) on production servers?

The reason I dont want to use device backed loop-aes is the dependency from
the block device. If I use file backed loop-aes and one server crashes, I
can just copy the crypto container as file to to an arbitrary fs created on
a i.e. a ide, scsi-blockdevice or even software raid of a new server. I
think I wouldnt have this functionality if I backup the (ide-,scsi- or
software raid-) block device with "dd" (maybe I am wrong?).

Kind regards,
A.Engels


-
Linux-crypto:  cryptography in and on the Linux system
Archive:       http://mail.nl.linux.org/linux-crypto/


[Index of Archives]     [Kernel]     [Linux Crypto]     [Gnu Crypto]     [Gnu Classpath]     [Netfilter]     [Bugtraq]
  Powered by Linux