integrating GPG with deniable steganography

Bernd Jendrissek berndj at
Thu Mar 22 15:13:01 CET 2001

On Thu, Mar 22, 2001 at 12:22:57PM -0000, Marlow, Andrew (London) wrote:
> > From:	Bernd Jendrissek [SMTP:berndj at]
> > Yes, you're both right.  Skewing pictures won't help anyway, and neither
> > will taking lots of pictures.  Mallory will still be able to see the same
> > message the intended recipient should.  You still need encryption to
> > force Mallory to guess that there *is* a message at all.  
> 	[Marlow, Andrew (London)]  Pardon? I don't understand. This thread

Neither do I, really.

> started when I said that I want to see a way of integrating GPG with
> deniable steg. This means it is encrypted first, then steg is used to hide
> the fact that encryption is used. This is because RIP (an anti-encryption

But steg and encryption are separate processes; one could say encryption
operates at a lower level.  Steganography hides the fact that information
has been altered *at all*.  Whether you've hidden an encrypted message or
a plaintext greeting doesn't matter, vanilla steg hides the fact that it's
there.  To be *deniable* you need to make sure that the alterations to
your original information change *none* of the characteristics of the
original message that Wendy can measure.

> law recently passed in the UK) makes users of public key cryptography
> potential suspects. Wendy the warden (who is Mallory???) can see the message
> that carries the GPG msg, and because she has knowledge of all steg
> algorithms she can extract the message. But she has to prove that it *is* a
> GPG msg. The GPG headers and trailers will be absent. Only the body can be
> recovered and Bob and Alice will argue that bytes likes these can be
> 'recovered' from any ASCII file that does not have constant spacing between
> the words.

Maybe any ASCII file can carry bytes that look like what Wendy found.  If
Wendy measures the message in some clever way that reveals that it is
unlike other "innocent" messages, Alice and Bob have some explaining to
do.  Your example is irregular spacing between words.  Fine.  An innocent
message from Bob to Alice may have a 1/2 space distribution of 70%/25%
If Wendy intercepts a message with a distribution closer to 50%/50%, she
will know something is amiss. This may or may not be proof of the presence
of a concealed message.

The only difference between Alice/Bob and Wendy is that Alice and Bob have
a decryption key, Wendy does not.  The idea is that *this* makes the steg
*deniable*.  The key is the only thing that reveals more information about
a message than Wendy can measure.

Bernd Jendrissek

More information about the Gnupg-devel mailing list