clipper and public key

[Eric Hughes <hughes@soda.berkeley.edu>]

Tim mentions that the Clipper chip requires public key in order to be
useful.  This is not quite right.

The clipper chip is only a symmetric-keyed block cipher with a
peculiar (and condemnable) key setup feature.  the chip _per se_ does
not involved public key.

The problem is that you have to get the same key on both end of the
link without transmitting it.  There is a "public key" way of doing
this: Diffie-Hellman key exchange.  That would require licensing from
RSADSI.

This is not, however, the only way to do this.  If you have a
symmetric cipher and a secret system key not known to the
participants, i.e.  embedded in hardware, then you can also transmit a
session key simply by encrypting it.  Of course if you know the system
key then you can read the traffic, LEEF's aside.  Such a system master
key could fairly easily be discovered, unless it's burned into the
chip by the manufacturer and the secret ends there.  (Yeah, right)

Hence in order for a reasonably (?) secure implementation of a
telephone which uses the clipper chip, D-H seems to be necessary.  In
fact, the AT&T 3600 phone does use D-H for key exchange.

Some have asked how come AT&T doesn't get sued by RSADSI.  Easy:
they're a licensee.

In summary: Does clipper require public key?  In itself, no.  In
implementation, likely.

Eruc