[Cryptography] Imitation Game: Can Enigma/Tunney be Fixed?

[Ray Dillinger]

On 01/05/2015 05:29 PM, Peter Gutmann wrote:
> Jon Callas <jon at callas.org> writes:
> 
>> Variants of the Enigma were used up until the early-to-mid-1990s, with the
>> eleven rotor NEMA.
> 
> There have been even more recent attempts to design (at least as gedanken
> experiments) secure rotor machines, see Ross Anderson's "A Modern Rotor
> Machine" from FES'93, available via
> http://link.springer.com/chapter/10.1007%2F3-540-58108-1_6.

I had another thought about securing Enigma (or designing
a rotor machine) last night.  This follows from thinking
about the "not enough different rotors" problem.  The rotors
were actually somewhat intricate and difficult to manufacture;
deploying with 30 of them would have been difficult.  So the
problem is to come up with something simpler to manufacture
and fairly easy to use, but cryptographically harder to break.

Suppose that each rotor contains one or more sockets for
interchangeable subassemblies that can be added in any
of several different orientations. One possibility is an
inner 'disk' not very unlike the rotor itself electrically
speaking.  The advantage would be that they can be easily
repositioned relative to the rest of the rotor, according
to the key of the day.

But these inner disks could be mechanically very simple.
All the mechanical difficulty of making contact while in
rotation, and being driven, and for that matter rigidity,
is already taken care of by the construction of the
rotor itself.  These inner disks could be very simply
connected inside them, just by being screwed down in
position each day.

Manufacture would be simple; you could stamp electrical
traces out of thin sheet copper, stick them through
prepunched holes in a "middle" insulating  disk, bend
them so the ends reach contact positions on the edge,
bend them again so contact areas stick through slots in
an insulating 'cover' on each side, and then glue the
extending contact ends down on the outside of the covers.

With a clockwise spiral on one side, pass through at a
chosen point, and counterclockwise on the other side, you
make a crossbar configuration - so any arbitrary combination
of connections is possible, no fancy printed-circuit
routing or shapes required.

In fact you could completely replace all the wiring (other
than the external contacts) in the rotor with one of these
disks, and just allowing a rotation of the disk relative to
the driving teeth on the rotor would make the code breakers'
job harder.

But there's no reason to stop there.  The disks would be
thin and fairly flexible, so you could make a stack of
them, using the extra thickness at the electrical contacts
to make sure each disk in the stack makes contact with
the next.  The job of rewiring a rotor for each day's key
is then taking the cover off the rotor, building a stack
of disks into it like stacking a deck of cards, putting
the cover back on, and screwing it down to clamp the stack
in place.

So, instead of just figuring out which of 8 rotors is in
each of four positions, the codebreakers now have to
figure out how the rotor in each of four positions is
composed from a choice of some (say) four out of the 30
or so disks that have been distributed, each of which
may be in two different orientations and any of 26
different rotations relative to each other and relative
to the driving teeth.  This would extend the key space
to a point where I'd be confident of it even today,
and be easier to manufacture per machine than the eight
rotors (with all the sliding contacts and contact springs
and driving lugs and individual wiring etc that each
rotor had to have). People with screwdrivers would
still have to connect all the springs and contacts,
but on just four rotors per unit rather than eight.

				Bear



-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: OpenPGP digital signature
URL: <http://www.metzdowd.com/pipermail/cryptography/attachments/20150106/7c0b993d/attachment.sig>