MYK-78

[Arthur Abraham <a2@well.sf.ca.us>]

4-May-93
  
Eric Hughes writes:

>It is possiqK another set of LEEF's; that's not
clear to
>me, but Arthur thinks you can.  

Sometimes I have a bit of trouble following Eric's written
syntax.  This
passage would make more sense if "can" was replaced by "must")

>If, however, one can just change the
>CV at will, one could send the LEEF's in the clear and then
>immediately change the CV (session key).  Now the LEEF has
been sent
>but the conversation makes no sense.  My money is that this
is
>interlocked with IV generation, though.

To quote myself: 
>[CV,checkword,L1,L2,IV] is a self-checkin+nit 

which means that these data items must be loaded in this order
(MSB first) and must all have been properly constructed via
the Skipjack algorithm, else the MYK-78 will yank on its ERROR
line.  To build a proper protocol, CV (= session key) and
checkword would be encrypted, and to be a proper citizen,
L1,L2,IV would be transmitted in the clear.

The funny thing is, that after loading the "self-checking
unit", and checking it,  MYK-78 will still accept a different
CV.  I specifically had this verified.  It will encrypt funny,
since the CV won't properly match the rest of the loading, but
it shouldn't be too hard to use this encryption mode. 

There are actually several ways to abuse this chip -- another
obvious one is tonever power it off,  and run with the first
LEEF you ever get -- it just seems to cry out in masochism.  A
real consumer crypto chip wouldn't be this flexible.

I hope to be able to say something about the true protocol for
law-abiding American citizens in the near future.

-a2.

ps: though I will continue to post to Cypherpunks, if you wish
me to read something, please send it to me directly.

-a2.