Detecting and defeating assignment server cheaters

[v00d00]

I dont see what the issue is to be brutally honest. People have been gaming the system since the project started, and i dont see it ever stopping regardless of what Pandegroup says or does (assuming they had the time to be bothered about it). The rules are just a list of things that various break on a regular basis. Sure some of us obey them, but take the borging issue. DId it ever stop? Did people on certain teams read the rules and decide to stop borging? Highly unlikely. Same goes for those automated installers that shouldnt contain the client. They are still available on certain team sites. So someone hacking their system for fun and points, is even more unlikely to stop. And in the end, will Pandegroup shut them down if they knew who they were? Highly unlikely, since regardless of how they did it, if they are turning in valid workunits and valid science, then PG is getting what they want out of the free computer power thats being donated to them. No one getting something for free is going to do something that jeapordises the results, unless the way it is being generated is illegal, which gaming the system isnt last time i looked.

As long as they keep turning in workunits, then i dont see a problem with it. Work gets done, science continues, life goes on.

[Jesse_V]

V00d00, I know of one big example where the PG has done some serious whacking at a team who had written illegal installers and secretly installed F@h on computers without permission. After an in-depth investigation, they threw a couple of lightning bolts. Please help us identify the teams/forums which offer illegal installers and the like. Not only do those violate F@h policies, but they make the project look bad and degrade F@h's security. That's an entirely different issue than changing a configuration so that the AS gives one different WUs.

[PinHead]

I must be missing something also. So, Punchy, are you upset because a BA12 AS is assigning WU to 16, 32 and 64 core folders? Or are you upset because a BA12 is assigning WU to an 8 core i7?

The thread seems to imply the former and not the latter. I thought the AS worked off of minimum cores, 12 core minimum or 16 core minimum.

[Punchy]

The issue, in a nutshell, is as follows: the BA12 AS is giving out p6901/3/4, the BA16 AS is giving out p8101, a 16+ core system should be receiving work from the two servers in a ratio of 1:10 (i.e. one assignment from BA12 AS for every 10 assignments from BA16 AS), but some 16+ core users have figured out a tweak or are using various known hacks to receive work only from the BA12 AS in order to get higher PPD.

[kiore]

The issue, in a nutshell, is as follows: the BA12 AS is giving out p6901/3/4, the BA16 AS is giving out p8101, a 16+ core system should be receiving work from the two servers in a ratio of 1:10 (i.e. one assignment from BA12 AS for every 10 assignments from BA16 AS), but some 16+ core users have figured out a tweak or are using various known hacks to receive work only from the BA12 AS in order to get higher PPD.

I see people asking so what is the problem.. Well This may be messing with the projects priorities, or maybe that is no problem. Where I do see a problem is it results in a heavier load of 'undesirable' workunits going to people who do not use this (let us be generous) work around. Still no problem? Well this IMO leads to frustration and possibly dumping or an arms race on work arounds so that more and more people are manipulating assignments in more and possibly harmful ways.

[DoctorsSon]

I see people asking so what is the problem.. Well This may be messing with the projects priorities, or maybe that is no problem. Where I do see a problem is it results in a heavier load of 'undesirable' workunits going to people who do not use this (let us be generous) work around. Still no problem? Well this IMO leads to frustration and possibly dumping or an arms race on work arounds so that more and more people are manipulating assignments in more and possibly harmful ways.

The problem is that Punchy is upset because the team he is accusing of cheating is going to overtake the team he is on.

[ChasR]

The incentive to cheat could be eliminated quite easily by realigning the value of the p690x series to produce the same ppd as p8101. It leads one to believe that it is no big deal to PG or that the happiness of donors running steady streams of p690x outweighs the unhappiness of those stuck running p8101. I think eveyone should be equally unhappy and the points realigned, instead of continuing the extaordinarily (ridiculously?) high ppd of the p690x series.

[texinga]

I think ChasR is on the right track here about how to approach things, and I would suggest a slight modification to his suggestion. PG could adjust the 8101 WU points to match that of the 6904 instead. There are those of us that never felt that the 6904 (or any WU requiring that much effort) was out-of -line points-wise for the effort required. How about taking positive road on this thing and correct the 8101 so that everyone that is working them will get a better reward. That way the people that are running 12-thread systems can still help PG finish off that series of work and still receive the same points they have been getting. No matter where you land on Punchy's issue, we all should suggest a "positive" solution to the oft-complained about 8101 points/effort structure.

[ChasR]

I strongly disagree that p8101 should be increased. THe p690x series marginalizes all other forms of folding, and the "effort" is trivial compared to what it takes to make the same ppd folding regular SMP and GPU on SP machines.

Edit:
Bruce needs to remind us about staying on topic.

[texinga]

I strongly disagree that p8101 should be increased. THe p690x series marginalizes all other forms of folding, and the "effort" is trivial compared to what it takes to make the same ppd folding regular SMP and GPU on SP machines.

Somehow, I figured you would take that position. Trivial to one person can be meaningful to another, but then again you or I don't get to make the rules do we? I was just adding my opinion on the matter and attempting to send it in a more positive direction than taking things away from people. The strategy of taking things away doesn't exactly spur people onward as much as considering the reward side of thinking.

[bruce]

STAY ON TOPIC.



I was just noticing that everybody has had a few opportunities to express their opinions and we're well into page 3. Drift and redundancy are creeping in. I know that this is a hot topic for many folks and not important to others. So far you've managed to keep the discussion (mostly) friendly so moderating the topic has not been required. (For me, a little discussion drift is a lot less bothersome that angry words if I have to rank various forum policies.)

Whatever happens next is up to the Pande Group. All I could do is remind you of what they've said in the past and suggest they'll probably continue to move in the directions indicated previously. (You can search that history as well as I can.) I know they are aware of the issues and I don't think that anybody has expressed an opinion in this topic that they haven't heard before.

How about bringing the discussion to a close.

[Punchy]

Would sneakernetting for the purpose of getting the higher scoring WUs (as opposed to its original purpose of reaching non-networked or firewalled machines) be considered a Best Practices violation? In other words, downloading work on a 12-core system to be run on a larger system. I would guess that this falls under the category of manipulating AS logic for "points enhancement" and would probably be a violation, but I'm curious as to what others think.

[DoctorsSon]

How wide spread of a problem do you think this is?
Do you have actual proof this is being done or are you guessing.
Personally I don't care because the work is still getting done and I fold on a team that doesn't pay members for folding.

[Afterburners]

It is widespread enough to PG to already be aware of it and have heard all that is to be said about it. The sad part is all this energy could be spent on the project instead of policing those that choose an un-ethical approach/attitude.

Not to mention many from multiple teams have left Folding all together because of similar actions. As soon as you allow the erosion of or lack there of quality and fairness to creep in, it takes a lot of effort to correct it. Just ask USC or UW or now PS...

Seems to me a few are taking one side or another vs. a solution. To the many that are trying to better understand and or look for a solution, Thank You!

[PinHead]

PG provided a solution quite some time ago, 690x are nearing their end and the issue will go away. That being said, and after a steady diet of 8101's, I suddenly received back to back 6903's and I'm just running the client straight. Now, after the blessing, I am back to my regular diet of 8101's.

So I am wondering where this 10 8101 to 1 690x rule came from? And is it factual or an guesstimation?