@Adam since it's now locked, can you please add the quote format around "See also" section here? @public removed it in several FAQ questions and I'm now rolling back those edits.
In part because it makes people like me want to find out how it works to break it, rather than to suggest improvements to make it work without being secret.
@Magisch I'm sure it is, I'll probably just set up some bots and see if it's something as trivial as a weighted threshold of votes over time or something.
@ShadowWizard It's pretty much never better. If it's so badly designed that mere knowledge of how it works results in it being completely broken, then it's worthless. It should be designed so that knowledge of the working does not make it easier to go through vote fraud.
@Magisch I think people could only abuse it if someone were to actually write a voting script to evade it. Most people don't know how to do that (and that's really the only way to get a lot of votes fraudulently).
Even still, a good design would detect signs of voting fraud, not detect signs of people who don't know how to evade voting fraud. Even something as basic as machine learning could be used to notice simple patterns.
Freenode has tried to do the same thing with its spam detection bot (forget its name right now), and we all know how well that works. :P
נינג'ה ישראל או NINJA ישראל היא תוכנית מציאות ישראלית המבוססת על התוכנית האמריקאית "אתגר הנינג'ה" (American Ninja Warrior) של רשת NBC, שהיא בעצמה ספין-אוף לתוכנית היפנית המצליחה בשם "סאסוקה" (サスケ; Sasuke). התוכנית הציגה מספר רב של מתמודדים המנסים להשלים מסלול מכשולים עם דרגת קושי עולה. בשנים בהן משודרת התוכנית בעולם רק מספר מצומצם של מתמודדים צלחו את המסלול הסופי. המתמודד המנצח, יובל שמלא, זכה ברכב שטח, ואם היה עובר מכשול נוסף היה זוכה גם ב-250 אלף שקלים.
מנחי התוכנית היו אסי עזר ורותם סלע, יחד עם צוות הפרשנים ניב רסקין ורותם ישראל. העונה הראשונה עלתה לשידור ב-25 ביולי 2018 בקשת 12, ונגמרה ב...
@Derpy My first guess is that it generates an interaction graph (forget what those are called) to see if there's an abnormal amount of voting within a specific cell (again still forget what the term is, but those little "clusters" of interaction).
E.g. for a given threshold for vote reversal, the amount the weight for any vote goes up is inversely proportional to the amount of distinct individuals the user has voted for.
@forest I doubt it is very complex. If I had to guess, I would say that probably it is just a scheduled script that checks if any account has received a lot of votes in a short period.
Exactly. I mean when I find out the details, it's not like I won't suggest improvements. Though I tend not to like suggesting improvements for secret algorithms.
> Every night around 03:00 UTC, a voting fraud detection script is run that looks for patterns such as these. It basically looks for users who have voted on another user's posts many times. The threshold number is fairly low within a given amount of time (the exact mechanics are kept secret).
Right now the hardest thing for me is to think of a way to test it out without actually voting fraudulently. Maybe if I create a dozen accounts to vote for each other, but still post beneficial content? But even then that'd be breaking ToS.
@forest If you're serious, just use SEDE and run whatever you want to test on there
use SEDE to find profiles and then look by hand at their voting history. You should be able to find what you want that way without actually sockpuppeting anything
@Mithrandir I don't care. If mods get aggressive because people are "thinking too much", then I don't want to be on this site anyway. If I'll get in trouble while not breaking any of the ToS, then we have a problem.
The attitude mods have towards curious users is what determines whether or not I'll find out the algorithm and suggest improvements, or find out the algorithm and take time to design an easy to use userscript for voting fraud (for education, of course).
@Tinkeringbell I think there's a specific legal definition, but I don't know it.
Cause I have this "funny" co-worker that keeps calling each if-else statement an algorithm... and I really hate that as to me an algorithm is a set of multiple instructions.
So whenever he says 'we need to change the entire algorithm', it means change the if-else or for-each loop. Not rewrite the entire code-base :P
In mathematics and computer science, an algorithm ( ( listen)) is an unambiguous specification of how to solve a class of problems. Algorithms can perform calculation, data processing and automated reasoning tasks.
As an effective method, an algorithm can be expressed within a finite amount of space and time and in a well-defined formal language for calculating a function. Starting from an initial state and initial input (perhaps empty), the instructions describe a computation that, when executed, proceeds through a finite number of well-defined successive states, eventually producing "output"...
> In mathematics and computer science, an algorithm (/ˈælɡərɪðəm/ (About this sound listen)) is an unambiguous specification of how to solve a class of problems. Algorithms can perform calculation, data processing and automated reasoning tasks.
@Magisch To clarify, should I expect moderator retaliation if I discover and publish non-copyrightable and non-trade secret information using non-disruptive means?
I know that I've been asked in a friendly way not to before when I looked at it, and I comply mostly because I respect the askers and understand their reasoning
What you consider "non-disruptive means" may not be actually non-disruptive; the problem is that these things are kept secret because if it was public, then fraudsters would have a much easier time of it. So if you happen to find out something, then it is preferred that you don't reveal it.
@Mithrandir By non-disruptive, I mean passive. I.e. using nothing but data which is already public (as opposed to creating vote fraud bots to test it actively).
I can't do anything about it if you publish the details of what you find in a blog post somewhere. Just understand that by doing that you are potentially handing the keys to undetected voting fraud to those who find it.
And... for understandable reasons, that's not something I'd like to happen.
@forest well, please don't write a blog post "How to bypass Stack Overflow vote fraud mechanism". You'll only harm Stack Overflow, and it's just something not nice to do. Leaving legal or not legal aside.
Be your views on security by obscurity as they may, fact is that's how it currently works and there isn't really dev time right now to completly rework it from what I can gather. So if you think that publishing an easy-made way to bypass it is going to make SE prioritize it, that's the wrong way to go about it
Other sites have similar fraud detection mechanisms, and they actually work, even with the information public. The reason is that it detects fraud behavior, not just fraud signatures. In order to evade it, users are forced to slow down their voting or to register a larger number of accounts to accomplish the same purpose.
Once upon a time there was a user named Curiosity.
That user got fair share of questions with what appears to be fair upvotes:
Examples: one, two, three
Then came this question where a user posted a comment that might have offended our Curious user:
Good chances that he/she chose ...
@forest there were always voting rings and there will always be, can't do much about it except keep an eye if possible and alert moderators when we see something.
I think it'd be simple enough to deterministically detect just by comparing the percentage of downvotes to a specific tag for a given user against the average.
@forest nahhhh... much better than that. It is pretty clear that recently someone on Sci-fi&Fantasy downvoted all MLP related posts. Or at least attempted to.
I noticed that most of the recent questions that are not closed since Oct 12 until now (Oct 16) have been downvoted at least once:
Can 2 elements be combined into a new release without a "clan based Kekkai Genkai"?
Is Gaara able to control sand because of his mother or shukaku (the 1 tails)
Try...
@ShadowWizard It is also possible that I were wrong. I think there was indeed a pattern there (all votes came at the same time) but probably they were too few to really justify any work
So, while it may have been valid on another post (but in that case, flagging would have been a better option) it wasn't really appropriate on THAT post.
It's so tempting to say 'I'm sorry, I looked but couldn't find a backlock to go through', if this co-worker didn't already act like I'm stupid. ON the other hand, perhaps I should just go with the flow!
@ShadowWizard Freenode is the largest software-development-focused IRC network. -R is a status for a user (you can set statuses on yourself) which means anyone can send you a private message. If you're +R, they have to be registered to send you a message.
-R on Freenode = thousands of new spam messages from a variety of the server's regular well-known fairly racist spammers. I.e. wowaname and that one kid whose name I forget but I really should remember who keeps making those offensive channels.