Re: [PATCH 0/5] eCryptfs: Filename Encryption

[Michael Halcrow]

On Thu, Nov 06, 2008 at 02:52:26PM -0600, Dave Kleikamp wrote:
> On Thu, 2008-11-06 at 14:27 -0600, mhalcrow@xxxxxxxxxxxxxxxxxx wrote:
> > On Wed, Nov 05, 2008 at 04:57:54PM +0100, Pavel Machek wrote:
> > > On Tue 2008-11-04 15:37:54, Michael Halcrow wrote:
> > > > This patchset implements filename encryption via a
> > > > passphrase-derived mount-wide Filename Encryption Key (FNEK)
> > > > specified as a mount parameter. Each encrypted filename has a
> > > > fixed prefix indicating that eCryptfs should try to decrypt the
> > > > filename. When eCryptfs encounters
> > > 
> > > That is 'interesting'. What happens if normal filename has that
> > > prefix?
> > 
> > If the lower filename has the prefix but does not have a valid tag
> > 70 packet following the prefix, then eCryptfs will complain in the
> > syslog and then pass through the lower filename as-is.
> 
> I'd recommend hiding this kind of syslog verbosity behind a debug
> config option.  I think it would be very easy to create a DOS attack
> against ecryptfs by putting all sorts of clever things in the lower
> file system.

I instead prefer leaving the default behavior as-is, while perhaps
introducing a module option to suppress such warning messages at the
user's explicit request. In general, I would imagine that the majority
of people would really like to know when a malicious attacker is
managing to write bogus mangled eCryptfs-ish files to their lower
filesystem, even if the means of transmitting the information about
the attack includes the possibility of filling up the syslogs. In
other words, if a malicious party is able to write to your filesystem
under eCryptfs, you probably have much bigger problems to worry about
than just your syslog filling up.

I can imagine some circumstances where filling up the syslog really is
worse than ignoring corrupted eCryptfs files in the lower filesystem,
but I submit that such circumstances are in the minority. I can also
contrive some circumstances where the lower files get mangled through
other non-malicious means, but I do not think such circumstances will
occur frequently enough to justify the risks from staying quiet about
file corruption (malicious or otherwise) in your lower filesystem. So,
if I had to pick a default behavior to do the greatest good for the
greatest number, I would leave the default verbosity where it is,
while providing the admin the option of disabling the verbosity when
encountering mangled lower files.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/