very short plaintexts symmetrically encrypted

[lists.gnupg-users at mephisto.fastmail.net]

On Sun, 10 Jan 2010 14:02 +0100, "Werner Koch" <wk at gnupg.org> wrote:
> On Sun, 10 Jan 2010 04:44:35 -0500, vedaal at hush.com wrote:
> 
> > symmetrical encryption is a simple way to avoid signing, while 
> > still maintaining relative reliability of knowledge as to who sent 
> > the message
> 
> That is not true.  For example you can't detect a replay or MitM
> attack.

Forgive me, but how is a MitM attack possible against a symmetric cypher
using a shared, secret key?

A MitM attack is really an attack on key exchange, as it requires the
MitM to intercept at least one public key, and substitute another (one
of his own) for it. Using symmetric crpyto, however, the key must be
prearranged, or exchanged by some other trusted means. Assuming only the
sender and receiver of the message know the secret key, I fail to see
what a MitM can accomplish. Of course, if we just broadcast the secret
key on the Internet, or something, then it's not much good--but anyone
using symmetric crypto should know better.