How do we prevent miraculously true DC results from manual submit?

[thyw]

Originally Posted by GP2

Also, is there any restriction on choosing a public name in PrimeNet, or can you choose a name that matches someone else's? If I see a result returned by "MadPoo" I'll trust it more than Joe Random.

If you take a look at results.txt or prime.log, (you probably know this already) you can see it's sending:
UID: name/cpu, completed xx..., We4... AID...
It is sending it by your login(private) name, not the public one. Also, looks like the public name nowhere to be found in local.txt or prime.txt.
(Login) name changing requires "...and email George so he can manually update your login username." Not something you can do too often invisibly.

^nevermind realized you were talking about manual submit, where it's only displayed in recent : public name ; Manual testing. Even less (no cpu name) on exponent page. It still requires an account to log in afaik. It can be blow away if you have access to the database.
If you set the same public name, you will get a warning "Tip: 1 other users also use this public name"
Is there a security rule (like both cannot be Anonymus) so both cannot be manual submits?

[GP2]

Originally Posted by thyw

If you take a look at results.txt or prime.log, (you probably know this already) you can see it's sending:
UID: name/cpu, completed xx..., We4... AID...
It is sending it by your login(private) name, not the public one. Also, looks like the public name nowhere to be found in local.txt or prime.txt.
(Login) name changing requires "...and email George so he can manually update your login username." Not something you can do too often invisibly.

^nevermind realized you were talking about manual submit, where it's only displayed in recent : public name ; Manual testing. Even less (no cpu name) on exponent page. It still requires an account to log in afaik. It can be blow away if you have access to the database.
If you set the same public name, you will get a warning "Tip: 1 other users also use this public name"
Is there a security rule (like both cannot be Anonymus) so both cannot be manual submits?

The project administrators have access to various hidden data that is not made public, for instance the shift count, that helps them better detect anomalies. But from the point of view of an ordinary participant, limited to only the public data, if I notice something that seems a bit odd in the data then I'm reassured if I see, for instance, Madpoo did a triple check. Except what if that wasn't the real Madpoo?

[kladner]

Originally Posted by GP2

The project administrators have access to various hidden data that is not made public, for instance the shift count, that helps them better detect anomalies. But from the point of view of an ordinary participant, limited to only the public data, if I notice something that seems a bit odd in the data then I'm reassured if I see, for instance, Madpoo did a triple check. Except what if that wasn't the real Madpoo?

This way lies madness, poo or no poo.

[GP2]

Originally Posted by GP2

The project administrators have access to various hidden data that is not made public, for instance the shift count, that helps them better detect anomalies. But from the point of view of an ordinary participant, limited to only the public data, if I notice something that seems a bit odd in the data then I'm reassured if I see, for instance, Madpoo did a triple check. Except what if that wasn't the real Madpoo?

To give a concrete example:

Consider M1048507. Note that "Tom Cage" and "Cornelius Caesar" returned identical bad residues here.

We can speculate on the reasons why: maybe they are really the same person, maybe they both used the same buggy software, or some other reason. In the end, it doesn't matter: because of this precedent, we can't consider any exponent that was double-checked by these two users to be "truly" verified unless it's triple checked.

Now consider exponents 1117307, 1420519, 1421011, 1421027, 1421153, 1421159, 1426367. By the above definition it seems that these small exponents in the 1M range were not "truly" verified until 2015 (think about that for a moment), when Madpoo went and triple-checked basically everything below 2M. So it's important to know that the user with that public name is really Madpoo and not an impostor.

Other than these two users "Tom Cage" and "Cornelius Caesar", there are several dozen other cases where two user names returned the same bad residue. In many or most of these cases, the second user returned all the results on the same day. Earlier in this very thread, George posted that there was an old version of mprime that would somehow forget user names occasionally and then randomly assign a numeric "S" username, and that, along with the usual glitches that sometimes cause a result to be resubmitted, probably explains this phenomenon. So far, looking for cases where these pairs of apparently-identical users both submitted the same good verified result under different names, all of them appear to have been triple checked (i.e., actually double checked) by another user.

[kladner]

So it's important to know that the user with that public name is really Madpoo and not an impostor.

Granted. Good point.

[LaurV]

Originally Posted by retina

There is currently no problem. So there is currently no need to change anything.

But if for some reason you really want to worry yourself about theoretical problems that don't yet exist then don't display any residue until they are matched.

+1. We proposed in the past the version that the residue is NEVER shown unmasked. Wanna guess? Be my guest... This will also forbid credit whores to do fictive tripple checks for DC credits (i.e. reading the residue from the database, and reporting a fake TC with the same residue, which actually gives you free DC credit).

[Madpoo]

Originally Posted by LaurV

+1. We proposed in the past the version that the residue is NEVER shown unmasked. Wanna guess? Be my guest... This will also forbid credit whores to do fictive tripple checks for DC credits (i.e. reading the residue from the database, and reporting a fake TC with the same residue, which actually gives you free DC credit).

It's entirely possible to simply reject any triple-check result of an already verified exponent. Exceptions could be made for when there's a valid assignment ID and it simply checked in far too late to be of any use (i.e. it expired, someone else picked it up and turned it in, all before this original DC computer finally finished... it happens with some regularity).

To answer some other questions... yes, users can use the same "public name" as other users. Only the private login id needs to be unique. There are a lot of users who list their public name as "Anonymous", for instance, which is why in the reports I changed how those are displayed to differentiate them from the actual "anonymous" user (people who don't create their own account).

Regarding other potential monkey business, I've tried to think like a trickster and ran queries to weed out potentially strange/suspect results. Things like "temporal proximity" where the matching results are turned in strangely at the same (or close) time. Sometimes that's a certain "LaurV" running first/second checks in parallel and turning them both in at once. Which, you have to admit, does seem fishy and it's why I still do an independent check. Not that I don't trust you, but if we wouldn't allow that sort of thing from some random dude on the 'net, then we should hold the same standards across the board.

As we found out from the whole independent triple-checking thread, there was only one example of a verified result turning out to have a different *actually verified* residue and that wasn't from someone scamming the system but from some system error (drat, I forget the details, but it wasn't that big a deal I seem to recall).

There have been other things I looked into as potential vectors for monkey business but I'd rather not go into detail because anyone reading about that might get some ideas. LOL

[bloodIce]

How come LaurV not checking neighbouring exponents instead of a DCing the same? It seems that it prompts a TC and one cycle of DC is lost. Double check is always independent. I respect him as well, but in the real DC we trust (or even in a factor :-)).

[ATH]

He wants a spotless account without any bad results for some reason, so if his 2 runs does not match residues, he does not turn in the results at all.

[GP2]

Originally Posted by Madpoo

It's entirely possible to simply reject any triple-check result of an already verified exponent. Exceptions could be made for when there's a valid assignment ID and it simply checked in far too late to be of any use (i.e. it expired, someone else picked it up and turned it in, all before this original DC computer finally finished... it happens with some regularity).

There are valid reasons for doing triple checks of already verified exponents, you yourself have done a bunch. Simple paranoia is sufficient justification. New LL results under 1M are already blocked, for the time being I don't think there's any need to raise that threshold.

I think it would be perfectly OK however if the server gave no credit for redundant triple checks unless there is a valid assignment ID. Liberating yourself from the tyranny of chasing credits frees you to try interesting things (paranoid triple checks, running P−1 on exponents that already have a known factor, or looking for PRP cofactors).

[kladner]

It's entirely possible to simply reject any triple-check result of an already verified exponent. Exceptions could be made for when there's a valid assignment ID and it simply checked in far too late to be of any use (i.e. it expired, someone else picked it up and turned it in, all before this original DC computer finally finished... it happens with some regularity).

I just completed a DC on which one, of seven expired Anons, had gotten this far:
2016-07-29 -Anonymous-LL double-checkLL78.3 % 2016-08-312016-10-10

Perhaps, if triple checks are sometimes useful, the limit could be set for gratuitous quad checks. That is, higher numbered checks would be accepted if there were mismatches below.

On the other hand, the system may have been set up that way so as to always provide gratification to anyone who completes an assignment, in the hope that they will stay with the project and do more work.

And, as pointed out, credits are immaterial to the actual work. They are there as incentive for fools like me, who still take pleasure in getting points. I might not have poured so much electricity into GIMPS early on in my stay here without that little tingle.

I know that Xyzzy was in a hot 3 or 4-way competition at TF, maybe with Craig, Jerry, and The Judger. for quite a while. A huge amount of work got done very fast.

Positive Reinforcement does do some good.