risk of stolen cycles, domain ranks, malevolent efforts vs DESCHALL

Eating Before Swimming (mathboy@sizone.org)
Thu, 15 May 1997 04:16:38 -0400 (EDT)

I consider myself having joined the project fully TODAY, not yesterday
when I merely got my home machines running on it. Actually, its not just
me, but members of a local "charity" group of "internet professionals"
under the name of uPowered (microPowered - we drink uBrew, work on
uComputers and work at the neighboorhood political level ;)

(We do free work for technically needy organizations in exchange for
social connections into local govt to thwart socially oppressive new
laws, such as Ontario's proposed multi-mega-smart card driver's
liscence, health-id, welfare-id, all-in-one privacy nightmare, which
will have your THUMB PRINT on it.)

DESCHALL suits our organization's efforts to produce technical issue
awareness in community and government, and may help us get known as well. :)

We are just a half-dozen consultants and employees working at high tech
jobs of one kind or another (a couple of us run our own small startup).
Through our members, however, we have, in a first try, amassed about 20
Million Keys per second worth of cycles:

5 486s
2 Sparcs
24 Pentiums (10 166s, 5 200s, 5 133s, 2 120s, a 90 and 75)
3 Alpha 300s

This totals over 20 Million keys/sec. Not as much as that poor guy
with 40 Pentium Win3.11 boxes has at his disposal but it will GROW!

By tommorow I will add those 8 16hr/day pentiums, and POSSIBLY another
30 Sparc 10s and 5 Indy 100MHz R4000s. This would be some 10-12Mk/s
more CPU. By next week some 5-20+ P100 thru P200s or Alphas or Sparcs or
Indy's may also be at our disposal (arent layoffs grate :( Our goal is
to hit 100 M k/s contributed to the project and make it into the TOP 10
of the rankings without becoming a University ;) [*]

However, its too bad the Supersparcs and Alphas and Indys suffer from
such poor performance! My 486s are equalling the Sparcs, and the Alphas
are getting about 380-420kk/s. The Indys will get 200kk/s. I MAY have
an SGI O2 for use as well, which will possibly mark the FIRST SGI box
I've tried this on which *CAN* run the BitSplice versions, as well as
being our first 1Mk/s+ single machine. All other SGI CPUs cannot run

The sparc performance is really too bad, cuz the 30 machines really add
up, even at their current contribution. However, they have a high
social risk cost factor. Debating their use now. [**]

[*] Rankings:

Our organization has our machines spread out through a number of our
clients' domains. Thus we will never officially be in the top of the
stats unless we total our domains, but only WE'D do that. ;) I'd like
to see our totals appear up there. I suspect that its not "one" group
(I may be wrong) that runs ALL of the top 5 or 6 University efforts.
Its probably more like 4 or 5 big groups, plus a number of
independents, and they just fall under the same domain.

Not that this all REALLY matters, it just sets an internal goal for me
to hit, shows how we CAN make a VERY significant contribution even if
we are NOT a university which sanctions this usage of cycles, even just
as a few independant consultants. I saw the UIUC totals and my jaw
just gaped last night. But I checked our totals and my goals again and
saw that they COULD be within ONLY (heh, "only") a magnitude (base 10 ;)
above us in performance.

My goal of 100Mk/s would bring us to 6th place according to may13 stats
I think. I should check again. We're currently running enough CPU to
put us in 23rd, and by this weekend I expect to be at 20th or so. All
the people involved in our effort intend to split the prize (some money
going towards equipment to help us do our donated work effort).

So, is there a way to create "artificial domains" or the like in the
ranks? That would be nice - tho perhaps then UIUC and PSU can say "well,
GREAT! now we can tie *OUR* other 15 domains we had going on this into
OUR stats as WELL!" and then up the highest ranks by a magnitude again. ;)

[**] Social risk cost factor and WINNING:

How is this winner notified, or is Rocke going to wait for the winner to
announce himself? In some cases, we're, uh, ahem, "sneaking" cycles
(hey they're heavily niced!!), but we'd rather not WIN (heh, AS IF! :) and
then have our admins notified who would thereby not let us know or nuke
our logs of evidence or otherwise.[+] Could it be possible to LET US contact
YOU Rocke if we log a win before you scare our admins into removing
access or coming down with draconian cycle usage policies (or instate
Ban keysearch efforts? CPU usage quotas again! ;)

[+] Every two hours I mail myself my logs on risky machines just to be sure. ;)
This should probably be done with PGP if the SuperSecretMessage will appear
on the screen when we win.

How are malevolent searchers guarded against?

How do you make sure that someone's not reporting falsely on their assigned
keyspace? Say someone took keyspaces, reported that they had an amazingly
fast machine (at the top of believable ranges) and just kept cycling
thru them all saying "nope, not in here" "not in here", etc etc, say
at the equivalent rate of 5 or 10Mk/s or more? How do you manufacture
the evidence that the entire keyspace WAS searched, and how do you
protect this info as it gets back to the server? (56 bit DES? :)



PS: Looks like I had to fix my SIG! Heh. Really, before I meant
40bit keys, I swear! ;)

Ken Chase mathboy@sizone.org Sonic Interzone $free$ email/news Toronto Canada
Join the DES Challenge! Wake up the US Govt!   www.frii.com/~rcv/deschall.htm