Cheating at Seti@home

Megor writes "Well it was bound to happen, people are cheating on Seti@home to inflate their work unit statistics, and the people who administer Seti are ignoring the complaints. ZDNET has an article explaining how they are cheating."

Interesting paper on this subject

[taviso]

Theres an interesting paper on this subject available here [distributed.net]. well worth a read.

Re:SETI Checking?

[phragle]

Seti does check,

As I understand it, for each unitl they send a number of redundant units out and then compare the evetual results taking the most popular result to be the correct result for that unit.

Re:Motivation?

[Enocasiones]

YOU should read the article:

Nealon has also identified ways in which the cheating could pervert the accuracy of data received by the project.

"Pervert the accuracy" means "corrupt" in my dictionary.

Re:Motivation?

[simong_oz]

My guess would be that it's the same sort of motivation that drives f@%$wits to beat up a pensioner for the $1.50 they might find.

OK, it's a dodgy analogy, but I'm just saying that I'm not sure it's possible to understand the motivation without having it yourself. To you and I (and many others), corrupting the results of the research is a terrible thing, but these arseholes just don't give a shit.

Re:Ahem.

[taviso]

one of the major culprits for this was actually Microsoft, they had a scam going where they optimized the SETI software for Windows and then published the results to show how well their platform performed.

Theres more information in the setifaq [setifaq.org], section 1.3.6

and on this usenet thread [google.com].

its a very interesting topic.

Re:Motivation?

[Anonymous Coward]

I'm not an anonymous coward. My handle in seti is EG, I crunch for EXDC and am a member of ARS technica.

One of the cheats actually indexes the name of the wu. the name of the wu contains data that locates the data to a chunk of the data tape.

by indexing the same file name over and over to bypass the duplicate checking routines, you are introducing wu's that do not correlate to their proper location in space. Resubmit such a wu thousands of times with the name indexed each time it's sent destroys the baseline of the science database. Copy the same wu to 1,000 other people for submission multiplies the error a million fold. it is very conceivable that there might be whole chunks of wu's results that cannot be scientifically normed to ANY location on the starmap! It does effect the science data.

Re:Old news

[Anonymous Coward]

Why does it seem someone /always/ has to play the This Is Old News card? Yippie for you dipshit, you heard about it a while ago, here's your prize. You've contributed zero. Hopefully you've at least made yourself feel better.

I will contribute something to this. Another scenario of cheating goes like so:

Set up a batch script to download multiple (like 100) work units at a time, each is 350k or so.
Another script steps through each unit, completing the work required.
When all units are done, result files are only 10k per work unit.
Prior to submitting, make a copy of the result files.
Proceed with submitting the results.

Repeat the process, except next time you're ready to submit, you can re-submit the prior 100 units as well.
As long as you maintain a copy of the completed units, you can keep submitting them.
Don't know if there's a limit on the frequency or quantity of resubmits, but your results go up exponentially: 100,200,300,500,800,1300,2100...

Re:EVIDENCE ?

[JDax]

The problem is not "bringing more machines to bear". A big issue has to do with those purposely re-submitting the same, already-processed results, over and over again, without ever having or needing to download a new WU.

Read this thread [infopop.net] and this one [infopop.net] and do try to follow the links to the graphs showing the suspicious results.

And do sign the petition against cheating here [teamprimerib.com].

More info.

[Faile]

While the article does a fair job explaning about the possible patches/hacks here's [hccnet.nl] a little more detailed version, and a few old usenet postings from S@H staff on the subject.

Re:whyd they

[Jugalator]

Not that I have anything against germans and their lives (heh), but the only seti@home work units on eBay [ebay.co.uk] seem to be from three germans??

Seti@Netherlands has some comments

[jacoplane]

here [seti-nl.org], though they're in dutch. I would translate, but I have no time. Basically, they say that they don't want to win by cheating, and that if they have evidence to support the claim that someone is cheating, that person will be removed from the team.

Official Word: not a problem for the science.

[SETIGuy]

Yet another overblow cheating report. Frankly, it doesn't really impact the science. The cheaters only process a small fraction of the total data and candidate identification doesn't rely on either a single result or a results from a single work unit.

Lets keep the the scope of the problem in perspective. What these guys are worried about is being in first place in the stats. I understand their concerns, but right now we have neither the funds nor the manpower to share them. Perhaps when SETI@home is shut down, and SETI@home II is running, we will go back and adjust the totals. Perhaps not.

SETI@home II will run under BOINC and will have more immunity to such exploits. The cost of such immunity will likely be a GUID for each machine running BOINC, in addition to a per user key pair. This, of course, will get slammed by privacy advocates. Hell, if Microsoft were doing it I'd slam them.

Right now our priorities are

• Keeping everything running.
• Identifying candidates for reobservation at Arecibo (sometime in the next 4 months or so).
• Building the SETI@home II data recorder.
• Getting Astropulse running as part of the BOINC beta.
• Finding enough funding to keep us running.
• Designing the SETI@home II analysis code.

Sorry, but fixing the stats can't be a priority right now. The extension of that to "SETI@home doesn't care about cheating" is extrapolating too far.

Re:SETI's a scam. Always was. So?

[Surt]

Ok, I know this is a troll but:

Care to venture how many tax dollars go to seti right now? I'll venture an impressive number: 0. Seti is not currently publically funded. That's right, its all voluntary donations.

I don't think anyone speaking for SETI has really made a claim such as: If we do X we will find life. Many times certainly they have said we _might_ find life.

The drake equation and its variables are not 'made up'. They are estimated as accurately as possible from star surveys of the galaxy. Only the last couple of variables are really open to any sort of scientific question, the others are pretty well settled. The question that should be asked is whether or not there are enough hospitable star systems out there that it might be possible for intelligent life to exist outside our solar system, and the answer is clearly yes.

Seti also has a clearly falsifiable premise: Seti claims that there might be other intelligent life inside of our galaxy (I believe that seti is willing to settle for our galaxy, since talking to anyone in a different galaxy is currently really beyond the realm of possibility for our technology). This claim is trivially falsifiable. Send a small probe to every start system, and survey any planets found. If no life is found, I believe that most if not all seti scientists will be glad to consider the question answered. Now sure, falsifying this way is a bit too difficult for us right now, so we're trying some other methods, but certainly within another century or so we should be well prepared to consider attempting a direct refutation of the seti premise.

Or you can view this from the other direction. Seti is a project engaged with the attempted refutation of the scientific premise that there is no intelligent life in our galaxy outside our solar system. In this case we might say that they are simply using the best tools available to them right now, and we can expect that within the next 10,000 years or so that the question will be pretty well settled by advances in our technology, but it might be settled much sooner than that, and the potential value of settling the question early is tremendous.

To clear up a last point, there are 2 questions I think seti would like to answer:

1) Is there life outside our solar system, inside our galaxy.

2) Is any of (1) that we discover, also intelligent.

I think both are clear and falsifiable.