[OTR-users] pidgin-otr: passphrase private key and sign public keys

[Ian Goldberg]

On Wed, Jul 09, 2008 at 11:36:04AM +0200, Julian Dibbelt wrote:
> Hello,
> 
> On Wed, Jul 9, 2008 at 12:10 AM, Ian Goldberg <ian at cypherpunks.ca> wrote:
> > I recommend using
> > encrypted filesystems, swap, and backups; it's easy to set those up on
> > most OSs by now.
> 
> That's not the same: OTR would only need to decrypt the private key
> during initilization of each conversation. The encrypted file system
> would be mounted (and thus be decrypted) all the time pidgin is
> running. I know that that is not much of a difference in theory but it
> adds up. It's about the effort an attacker has to make. Starting up vi
> and editing otr.fingerprints while I am e.g. running for the door is
> not the same as installing a key logger and hiding it from the process
> list.

And this is exactly the sort of thing that the OS should do for you.
It makes no sense for every application to have to protect itself from
this scenario.

> > Users have no reason to enter a pass phrase when using
> > IM now,
> 
> Hmm, I do enter a password connecting to the IM server. And I dont let
> pidgin "remember it" because as far as I know pidgin is missing a
> master passphrase to protect my IM passwords.
> 
> Also in the CodeCon2005 presentation you compare yourself with pgp -
> and with pgp I have to enter the pass phrase all the time because pgp
> does encrypt the private key. It also signs my local ring of public
> keys with my private key - so that I can be quite sure that they dont
> get manipulated with out me noticing.

Every time you force a user to enter a passphrase, you have lost.  Users
don't wake up in the morning thinking "I'd like to enter a passphrase
today".  They think "I'd like to talk to Bob".  The earliest security
usability studies were exactly about PGP and showed that even
computer-savvy users often couldn't use it securely.

Our SOUPS study of pidgin-otr showed that users had no problem
initiating encryption (since it's automatic and pretty much
transparent), though they did have some problems with authentication,
which we've addressed in the 3.2.0 version.

> > At least on Linux, one's .purple directory is
> > unreadable to all except the user and root.  (I don't know what the
> > Windows situation is.)
> 
> Ubuntu 8.04:
> -rw-r--r--  1 jdibbelt jdibbelt   406 2008-07-09 10:41 otr.fingerprints
> -rw-r--r--  1 jdibbelt jdibbelt  1984 2008-07-08 16:44 otr.private_key

But the directory those files are in is drwx------, so the perms on the
files themselves don't matter.  [That having been said, we've committed
a patch to make those files be -rw-------, for consistency with other
files in that directory.]

> Overall, I get your point of view and I will stop nagging now.

OK.  :-)

   - Ian