Some time ago, we briefly discussed the fact that a number of people (earlier
described by the SETI people as the "UNIX crowd") are faking and forging a
substantial amount of the results that are being returned to the setiathome
project at Berkeley.

The posting I've included below appeared on sci.astro this evening. One of
the participants/scientists in the program explains a bit about the extent of
the problem. The conversation started in regard to the fact that every WU is
currently being processed 3 or 4 times -- and the wastage of computer
resources that represents.

Wirt Atmar

=======================================

In article <[log in to unmask]>,
David M. Palmer <[log in to unmask]> wrote:
>[[ This message was both posted and mailed: see
>   the "To," "Cc," and "Newsgroups" headers for details. ]]
>
>In article <[log in to unmask]>,
>[log in to unmask] (Eric J. Korpela) wrote:
>
>> I'm afraid you don't understand.  I want to see that at least two machines
>> get the same answer.  Do you even comprehend what a miniscule fraction of
>> the machines out there faking results would do to the completeness of the
>> survey.
>
>Yes, it would reduce it by a miniscule amount.

We have one "user" who every day starts a new account, downloads 25000 work
units and then doesn't process any.  He does so from a large proxy server
from a national ISP, so we can't block it by IP address.  All we can do
is tolerate it.  25000 work units a day is not a miniscule fraction.  If we
only sent out unit once, we'd never see a result from any those work units.

>If you send out 10% of the data twice as verification even randomly,
>then anybody who fakes 40 WU has an exp(-4) chance of getting away with
>it.  If you don't do it randomly, then you can catch everyone.

We also have users who start up a new account daily, send fake results,
and then transfer the stats to another account.  You just can't catch everyone
with this method.

>Just out of curiosity, how many people have you caught cheating? and
>how many of those submitted just a few forged units?

Currently 9 of the top 20 gaussians on the stats page are forged.  We
have, as of today deleted 958 accounts due to obvious cheating.  When we
get to brute force comparisons of results, I'm sure we'll see ten times
that number.

One thing you are ignoring is that the cheaters don't cheat at low rate.
The download work units at high rate, and return forged results at high
rate. The percentage of our capacity being used by forgers is much higher
than the percentage of our users who are forgers.

>> Not at all.  I don't see our track record as poor.  Your opinion obviously
>> differs.  And to our 420,000 current users your opinion means so much.
>
>How many of them know that most of their processing power is being
>totally wasted?

It is not wasted.  And if you think it is wasted, don't participate.

Eric
--
Eric Korpela                        |  An object at rest can never be
[log in to unmask]            |  stopped.
<A HREF="http://sag-www.ssl.berkeley.edu/~korpela">Click for home page.</A>

========================================