From: Paul Koning               1695 <pkoning@chipcom.com>
Date: Tue, 24 Oct 95 09:36:24 PDT
To: Cypherpunks mailing list <cypherpunks@toad.com>
Subject: Re: Reformated: How secure....
Message-ID: <308D3F77@mailer2>
MIME-Version: 1.0
Content-Type: text/plain



(Re Steve O's Privasoft followup)

Your reply didn't really answer all the questions that have been raised.

You mention a 2 digit key extension, so does that mean the key is 11 digits?
If not, how is a key extension different from a key?

Given that the program is exportable, why is it any stronger than any other
exportable system?

Re how you would recognize readable text: if the scramble is good, the
output would tend to look like white noise.  An invalid decrypt would have 
the
same general look.  But a valid one would have the properties of a normal
fax, which is long run lengths.  So if you do a run length check on the 
data,
you could readily distinguish a valid decrypt from an invalid one.

Incidentally, I looked at the sample encrypted document on the web.  It has
an amazing amount of structure in it, and certainly doesn't look at all like 
the
white noise I expected.  That in itself suggests a weakness.  (Notice the
dashed lines running down the page, 7 of them...)

Also, you forgot to address the known plaintext attack.  If I have both 
plaintext
and cyphertext, then recognition of the correct decryption is no longer an
issue (it's simply a compare).  For the system to be valid by modern 
standards
it has to be secure against a known plaintext attack, and preferably against
a chosen plaintext attack.  (In other words, if I get you to fax an image of 
my
choosing I should still not be able to recover the key.)  It doesn't appear 
that
it is secure in that sense.

Finally, the property you mention of being able to cope with damaged images
doesn't sound at all unique.  You would fail to have this property if you 
use
a block cypher with long blocks, a cypher with error propagation (e.g. CBC
mode in DES), or any cypher applied to the compressed form of the fax.
But any stream cypher or transposition cypher (as I gather you use) applied
against the raw bitmap has this property, since any flaws in the encrypted
image would simply become noise in the plaintext image.  (A black line
would become a gray line in a stream cypher; in a transposition cypher it
becomes pixel dust on the page.)

     paul