mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Data

Reply
 
Thread Tools
Old 2019-02-22, 17:20   #12
ramgeis
 
ramgeis's Avatar
 
Apr 2013

7316 Posts
Default

Quote:
Originally Posted by petrw1 View Post
Are there gaps in her range. I've seen cases of users sending in thousands of NF in a batch then all the corresponding F a few hours later.
No, they are just bogus results as it took just a couple of TF attempts to find factors for the claimed NF results. See the other thread: https://www.mersenneforum.org/showthread.php?t=24103
ramgeis is offline   Reply With Quote
Old 2019-02-26, 19:01   #13
Maciej Kmieciak
 
Nov 2018
Poland

2·7 Posts
Default

I think some of the following rules could be implemented to prevent such events:

-New account gets the status "unverifed" or something like that for 30/90 days or until it uploads correct Double-check. In this period it can't post results without assignment ID (and other restrictions like GHz-days limit)
-One result per one assignment ID for ECM and P-1 factoring (this would block what I did)
-If someone posts more than 1000 (or so) trial factoring results without a single factor, some human have to look at them before publishing
-Some checks in timestamps (not too many results in the same second)

Does it make sense? (Sorry for my grammar again)

Last fiddled with by Maciej Kmieciak on 2019-02-26 at 19:03
Maciej Kmieciak is offline   Reply With Quote
Old 2019-02-27, 04:10   #14
Madpoo
Serpentine Vermin Jar
 
Madpoo's Avatar
 
Jul 2014

3×5×7×31 Posts
Default

Quote:
Originally Posted by Maciej Kmieciak View Post
I think some of the following rules could be implemented to prevent such events:

-New account gets the status "unverifed" or something like that for 30/90 days or until it uploads correct Double-check. In this period it can't post results without assignment ID (and other restrictions like GHz-days limit)
-One result per one assignment ID for ECM and P-1 factoring (this would block what I did)
-If someone posts more than 1000 (or so) trial factoring results without a single factor, some human have to look at them before publishing
-Some checks in timestamps (not too many results in the same second)

Does it make sense? (Sorry for my grammar again)
Those are interesting ideas. There are pros and cons to each of them.

I don't know if you saw the other thread on this forum, but we've had someone creating bogus accounts and submitting a LOT of fake results. I hope that's not you trying to "test" our system again.

We've been discussing various things to try and prevent fraudulent submissions, but the problem is that the very nature of the system is that anyone can download the software and do tests, and they can also use other apps besides Prime95 and submit results using those manual pages.

For a new user, they may be able to return results pretty quick depending on the work type and exponent size, and we want to encourage new users to participate. Placing restrictions on them could be detrimental, and to be honest we don't have much problem with people submitting bogus results.

Every now and then someone (like you) will do it, and we find out and have to clean things up, but it's very rare.

Regarding only submitting results once per assignment ID, you'd think that would be okay to do, but sometimes people do actually submit multiple ECM curves under a single assignment ID, and it's also normal to submit a P-1 result with an assignment, and then do the LL test and turn that in with the same assignment (TF can be included in that as well).

Your ideas about the # of results coming in are okay for most people. We do have several heavy duty users who turn in thousands of TF results daily. One of those users will sometimes break the "factor" and "no factor" results into separate batches (not sure why, but that's what I heard) so we might see thousands of "no factor" come in all at once, but then they'll submit another set where it's nearly all factors.

Users like TheJudger and TJAOI come to mind.

When you look at the activity graph, you'll notice that the # of factors being found has a lot of spikes of activity, and that's because the results get queued up for a day or two and then submitted all at once.

I can't blame them for doing large batches... going to the manual result page and uploading the result file isn't something you want to do too often.

I'd like to see those 3rd party clients like mfakt* and gpuowl be able to use the API so they can submit results automatically like Prime95 does. Of course, they may have firewalls that don't let them reach the server, but it's still a good idea in general.
Madpoo is offline   Reply With Quote
Old 2019-02-27, 19:18   #15
Maciej Kmieciak
 
Nov 2018
Poland

168 Posts
Default

Quote:
Originally Posted by Madpoo View Post
I hope that's not you trying to "test" our system again.
No, that's not me, I'm not going to do it any more. To be 100% clear, I also tried this ECM cheating few months ago on a much smaller scale on the following exponents:
4249327
12423997
12424001
12424007
12424039
12599963
I know it was stupid, but it's better to let you know instead of hiding.
My point is, you can detect large bogus results, but not small. For example, fake "no factor" results leading to not necessary LL test. But as I understand, we're assuming that people play fair.

Last fiddled with by Maciej Kmieciak on 2019-02-27 at 19:25
Maciej Kmieciak is offline   Reply With Quote
Old 2019-03-02, 21:12   #16
Madpoo
Serpentine Vermin Jar
 
Madpoo's Avatar
 
Jul 2014

3·5·7·31 Posts
Default

Quote:
Originally Posted by Maciej Kmieciak View Post
No, that's not me, I'm not going to do it any more. To be 100% clear, I also tried this ECM cheating few months ago on a much smaller scale on the following exponents:
4249327
12423997
12424001
12424007
12424039
12599963
I know it was stupid, but it's better to let you know instead of hiding.
My point is, you can detect large bogus results, but not small. For example, fake "no factor" results leading to not necessary LL test. But as I understand, we're assuming that people play fair.
Yeah, I didn't really think that was you doing anything. I was pretty sure of that.

Thanks for pointing out those other cases, I may tidy those up.

Ultimately this project is geared towards LL and PRP tests which do involve matching double-checks. Bogus factoring work is a pain but doesn't ultimately do much damage. Yeah, you may prevent someone from finding a factor if the work was actually done, so that would be a bummer. In the case of ECM work those tend to be smaller exponents that have already been LL tested and verified. I know some people love finding factors on the smaller stuff though.

LL and PRP testing relies on keeping the residue masked/hidden until it has been verified, and I am very much against having people double-check their own work because it breaks that aspect of it.

I also do some other data spelunking to make sure people aren't trying to game the system in other ways...things I won't go into detail about, but I do consider that some people are mean and may do nasty things, so I'm keeping an eye out, as are others.
Madpoo is offline   Reply With Quote
Old 2019-03-02, 21:37   #17
Madpoo
Serpentine Vermin Jar
 
Madpoo's Avatar
 
Jul 2014

3×5×7×31 Posts
Default

By the way, I just queried for other cases where people submit multiple ECM results with the same assignment ID.

Like I thought, there are people who do it, legitimately. It might be 3 curves at a time under the same AID.

User Yxinity is one such who has done this many times, and I'm pretty sure it's all fine.

Ultimately though, there's really no good way I could think of to be able to verify that factoring work is actually done (in a way that the server could easily confirm when each result is submitted).

Maybe the people talking about this in the other thread will come up with something... they lost me when they got deeper into the math.
Madpoo is offline   Reply With Quote
Old 2019-03-03, 03:46   #18
ATH
Einyen
 
ATH's Avatar
 
Dec 2003
Denmark

ADE16 Posts
Default

You are actually often handed a worktodo line automatically with 3 curves with 1 assignment ID.

This is from ECM CF, but it also occurs for "normal" ECM if the exponent is low enough you get 3 curves in the same line:

ECM2=7BC509AE4EC6A99F6B0B49A0B7B01CE0,1,2,4626463,-1,50000,5000000,3,"2711107319,1228319819568841"
ATH is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Results ET_ Operazione Doppi Mersennes 604 2020-03-26 15:17
Results ET_ FermatSearch 5 2019-12-02 06:09
gfn results ET_ FermatSearch 8 2019-06-02 12:09
CPU Results last 24 hrs Unregistered Information & Answers 3 2010-07-26 00:49
0x results... Mike PrimeNet 11 2004-05-23 12:55

All times are UTC. The time now is 21:50.

Sun Mar 29 21:50:09 UTC 2020 up 4 days, 19:23, 2 users, load averages: 1.34, 1.35, 1.40

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.