Protecting Files at Home Using Encrypted Containers (Linux Journal)

Posted Oct 5, 2005 19:26 UTC (Wed) by eludias (subscriber, #4058) [Link]

Posted Oct 5, 2005 20:29 UTC (Wed) by Wummel (subscriber, #7591) [Link]

1. The containers are not resizable, so you have to know the size beforehand, perhaps leaving some safety space for data that gets added later.
2. The container file size is restricted when one wants to burn a container to DVD/CD (filesystem iso9660). Limits vary from 800MB to 4GB. On my system I was finally able to write 4GB container files with growisofs from the dvd+rw-tools to DVD.
3. The key file (crypto.key in the article) is small, but extremly important. My scripts make multiple copies in case one gets corrupted.
   A solution would be to use LUKS as partition format.

Posted Oct 5, 2005 22:24 UTC (Wed) by Ross (subscriber, #4065) [Link]

a) The script examples are really bad. Overuse of cat, using control-c where it is not needed, etc.

b) Key handling. This is important. They key needs to be both protected from disclosure to unauthorized parties and protected from loss. If you lose the key you lose the data. It shouldn't be stored in plaintext on the hard drive. That's equivalent to storing the original data there. An alternative which isn't really discussed is using passphrases to either a) generate the key or b) encrypt the key for storage. Using a USB memory stick also works as long as you physically protect it and don't leave it plugged in when the data isn't mounted.

c) Memory and swap. When you mount a filesystem portions are loaded into RAM and various portions are cached. Keep in mind that if you do not shutdown the system (and even then to some extent) the data is still present. Similarly userspace accesses to the mounted filesystem will result in portions being loaded into swappable RAM. It is likely that your swap-space will contain portions of the data. Try grepping for one of your passwords in your swap partion some day. Using crypto swap, not using swap at all, and erasing swap partions on shutdown are all ways to address this problem (the latter is the least effective... what if the plug is pulled?).

Posted Oct 6, 2005 9:55 UTC (Thu) by lamikr (subscriber, #2289) [Link]

So far I have only seen some emails from Fruhwirth Clemens and Jari Ruusu.
But it seemed that both of them had very hostile attitude for each others and
that seemed to overrule all technical discussion from the advantages/disadvantages of both implementations.

Could somebody explain this for me?
(I have used and I am still using cryptoloop, as Mandriva has been kindly patching their kernels with cryptoloop patches and I have not understood why I should change to something another method...)

Mika