[RC5] DES-II-3 [LONG] [Repost]

David Taylor dtaylor at nildram.co.uk
Sat Aug 29 22:21:18 EDT 1998

Sorry if anyone gets this twice, Its been about 24 hours, and it hasn't
shown up, and AFAIK this list isnt moderated, so I am just reposting it.

 [dbakernote on behalf on moderator]

  Moo! The list is moderated.  The current moderator seems to be afk
  for the weekend.  I'm approving posts on his behalf.

	Moo moo ka choo!
	dbaker at cuckoo.com

 [dbakernote on behalf on moderator]


Since we only have 14 hours to crack DES-II-3 in after the contest
data is released, if we want to win the top prize of $10k,
I doubt the current system would work..

For a start.  It takes up to an hour to generate blocks, and propagate
them across the proxy network for clients to download AFTER the data
is released, which has just wasted 1/14 of our time.

Then, many clients connect once every 24 hours, or less, so those
clients may never see a block during the time we can win the $10k.

An idea that I have seen before is putting a timer in the client to
make it connect automatically when DES-II-3 starts, but that would
very probably overload the proxy network.

So, if we started distributing the keyspace up to a week before the
contest starts, (this would allow us to generate the blocks BEFORE
the contest, without wasting time), then simply providing the client
with the 8 bytes or so of Cipher Text it needs + the IV?

Also, so everyone's clock being wrong doesn't wreak havoc, how about
the proxy telling the client that DES-II-3 will begin in
'73.456 hours' or whatever, and the client will store the time
according to THAT particular computer in a checkpoint type file
(to stop problems with multiple computers with different times
sharing ini's)

Offline clients could just ignore it, and Lurk-Only clients
would just ignore it until they were online.  If there would
still be too much load on the proxy network, a further offset
could be embedded in each block, say +25 minutes, so that each
client would try to connect at a different time..

So, basically this is what happens

A week before the contest, DES blocks are created, containing
information on ONLY which area of the keyspace to check, and
an offset (0-100 minutes, say).

These blocks are then distributed (keeping a lot spare so clients
that lose the blocks don't muck it up). The proxy tells each client
how long until the contest begins, the client then adds the offset to
this time, and calculates what time it will be on that computer when
the contest begins, and stores that time.

At that time the client attempts to connect (if it is 'allowed' to),
the client then downloads the IV and cipher text, and starts cracking.
After a relatively short period of time, the client should return the
work it has done, and fetch more blocks.

Now I realise this is rather complicated, and weird, and would require
a complete reworking of everything, but the current system will just not
switch over fast enough to do DES-II-3.  With the current set-up we may
as well not bother, unless the key happens to be in the first 3-4%.

There is probably some massive flaw in this, so just yell at me if it wont
work (though not if the reason that it wont work is it would require
to be redone - I know that, but as far as I can see, its our only choice).

These clients were designed with RC5-64 in mind, which has no time limit,
they weren't designed to have a fast contest switch over (they weren't
designed to have a contest switch over, period), or a low block turnover
time...  We need to redesign it to make DES-II-3 practical...

Can we maybe expect V3 before Jan 13 1999? And if so, is it any better than
V2 at this kind of thing?

Just my $0.02 ( x 10^4 ) :-)

 David Taylor
 dtaylor at nildram.co.uk

To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest

More information about the rc5 mailing list