PDA

View Full Version : is SETI's stats system flawed?



FoBoT
08-10-2002, 09:09 AM
I have been reading a thread over at Arstechnica about some easily exploited loop holes in the SETI stats system.
It seems it is quite easy to cheat and rack up big stats without actually doing much work in SETI

anybody here care to comment?

Thread (http://arstechnica.infopop.net/OpenTopic/page?a=tpc&s=50009562&f=122097561&m=4790967925)


This hole is huge, really, really huge, and allows an individual with only modest power to accumulate thousands of WU's of credit in a short period.

The way that this is done is startlingly simple, every day you process 10 WU,

On Days 1 - 10, you submit the 10 you processed,
on day 11 you sumbit day 1 = day 11 == 20 WU
Day 12 == 2 + 12 (20)
Day 21 == 1 + 11 + 21 (30)
Day 31 == 1+11+21+31 (40)

and so on, your stats grow and grow this way and Berkely are not checking more than a few days back for duplicates.

Cobra-ICE got greedy, and tried to submit all of the WU's every time, and eventually got caught out.

Cheat #2 is far more subtle. This involves a ring of users, one of which I have the names of (passed to Berkeley already).

When the client reaches 99% the output files are moved to a safe place, and processing resumes. The safe 99% completed WU's are then given to ~200 users, and they replace the user_info.sah file with their own, crunch the final 1%, and return the WU - getting credit for it. So one WU is credited to ~200 people.

The final irony is that the two cheats can obviosly be combined.

fRaGgLe
08-10-2002, 11:30 AM
The simple answer is that the stats system in not flawed, it is the idiots that exploit it that are flawed.

FoBoT
08-10-2002, 01:12 PM
i hadn't realized you were over here fRaGgLe

i don't check SETI stats much , i only run P3 450 on SETI normally