Skipjack uses Elliptic curve? (was Re: Moderation [Tim,Sandy])

[Adam Back <aba@dcs.ex.ac.uk>]

Lucky Green <shamrock@netcom.com> wrote:
> [...]. How many of you remember the anonymous message posted to this
> list revealing that Skipjack is an elliptic curve cipher? [One of the most
> respected names in cryptography confirmed this to me in private
> conversation. No, the person was not privy to the secret specs. The person
> didn't need to be. :-]

What aspect of Skipjack family is Elliptic curve?

Skipjack itself I thought was a symmetric key block cipher, with 80
bit keys and 64 bit block size.

The key escrow designs (clipper chip and family) included several
additions:

1. check sum to prevent LEAF (Law Enforcement Access Field) forgery
   (16 bits, which is not enough as Matt Blaze demonstrated)

2. government access copy of chip's serial number encrypted with LE
   family key in LEAF 

3. copy of session key encrypted with unit's escrow key in LEAF (the
   escrow key is the key that is stored in the government database
   indexed by chip serial number - the database which is split between
   the two escrow agents).

4. hardware random number key generation

5. undisclosed key exchange mechanism

6. are DSS signatures used?

Presumably the Elliptic curve is for key exchange?  Is there something
about the design which implies Ellitpic curve must be the key exchange
mechanism used?

Another possible area for public key, if they had it on chip, would be
to use public key encryption for the encryption of the serial number.
Otherwise, when the chip is reverse engineered the LE family key would
allow traffic analysis of all clipper traffic.  Public key would
prevent this.

(According to Ross Anderson's paper on tamper proof hardware, at least
one chip manufacturer has reverse engineered the clipper chip)

> If nobody cares about the leaks, why do we need to provide a forum for
> them? Besides, there are other fora that could be used. sci.crypt or
> Coderpunks are both good places to post "found" code.

It is true that sci.crypt and coderpunks do make alternative fora.
Somebody else pointed out that rc4.c was posted to sci.crypt first.  I
think they are correct, and in fact if I remember, it was forged as
from David Sterndark <david@sterndark.com> or some other play on David
Sternlight's email address.

Adam
--
print pack"C*",split/\D+/,`echo "16iII*o\U@{$/=$z;[(pop,pop,unpack"H*",<>
)]}\EsMsKsN0[lN*1lK[d2%Sa2/d0<X+d*lMLa^*lN%0]dsXx++lMlN/dsM0<J]dsJxp"|dc`