Re: [gnukeyring-discuss] development

[Martin P. <mb...@sa...>]

On 23 Oct 2003, Florin Andrei <fl...@sg...> wrote:
> On Thu, 2003-10-23 at 10:58, Jochen Hoenicke wrote:
> 
> > No, the development hasn't stopped.  I'm preparing a new version with
> > a new crypto algorithms and it takes some time, especially as I can only
> > work on weekends.  I hope to have a test release, soon.
> 
> Awesome!
> 
> Jochen, can i ask a few "hard" questions?
> What's the actual strength of the encryption algorithms on a device
> running PalmOS? Applications running on Unix clones make use of various
> network- and system-events to feed some pseudo-random numbers
> generators. Is there anything like that on PalmOS?

Yes, Jochen added a randomness pool based on the Linux /dev/random.
Graffiti input is a good (low baud) source of input that is hard for
an attacker to precisely predict.

> I guess what i'm asking is, are there any inherent weaknesses in an
> application that runs on PalmOS and uses strong encryption, just because
> it runs on PalmOS?
> 
> I'm quite aware that the encryption algo might not be the weakest link
> in the chain, so perhaps there's no point in worrying too much about
> that (*) but i thought i should ask anyway.
> 
> (*) - Much more important seems to be the master password on the Keyring
> application. If that's too weak, the database is easily crackable.

I think this is the limiting factor.  I think most people would not
have the patience to write a passphrase of more than about 5-8
characters.

Since an astute attacker can try to decode it on a PC, they can cycle
through combinations much faster than somebody could write them.

Of course there might also be bugs.  This was the first real crypto
program I wrote (quite a few years ago now), though other people and
especially Jochen have checked it later.

> Also, a resourceful group might read micro-scratches on the Grafitti
> area and deduce often-used string patterns.

Probably only a problem if a large fraction of your writing is the
passphrase.  This might happen if Keyring is the only app you use and
you rarely add new information.  But I would expect most people would
write in other information at least sometimes.

What I would really like to see is all of the data across all
applications strongly encrypted.  (Like mounting an encrypted volume
on unix.)  Doing so would probably require hairy PalmOS hacking, and
is certainly out of the scope of Keyring.

> Finally, if the application "leaks" information into the palmtop's
> memory, that could be used as well to do some guessing.
> 
> In any case, a palmtop is never to be considered totally secure, but i
> feel that, if the encryption stuff doesn't have any inherent flaws, the
> security of the hardware should be better than that of the PCs (it's a
> lot harder to install "keystroke loggers", it's a lot harder to snoop on
> the RF fields generated by the display and by the input device,
> etc.).

Yes, I think so.  In particular it's pretty hard for somebody to
remotely compromise a palmtop, but probably more than 1% of Windows or
Linux machines in the world are compromised at any time.

-- 
Martin