30

There have been some rather unsavory things posted almost mechanically as of late, and we're seeing an increase in spam / trolling on a couple of other sites. As we can't rule out that these things are related, we're making some tweaks to help you dispose of it with quite a bit less effort.

Effective 3 minutes ago, it only takes half of the usual spam / abuse flags for a post to be deleted and the spam system to be trained. We hope this will make it easier for folks to clean this stuff up, and to teach the system faster as long as this continues. We're going to leave this in place for the next 7 - 10 days, run some numbers, and see if it's time to go back to normal.

Special thanks to all of you that have been helping to keep this site welcoming and clean, we're here with you and we believe that this will help put an end to it.

If you have questions or concerns, please just leave a comment or an answer :)

15
  • 1
    Tim, I saw the same message on The Workplace Meta. Could you please provide us a list of all sites for which this policy is in effect?
    – Glorfindel
    Dec 7, 2016 at 16:33
  • 3
    @Glorfindel - AFAIK, only workplace and here at the moment.
    – Oded
    Dec 7, 2016 at 16:34
  • @Oded thank you.
    – Glorfindel
    Dec 7, 2016 at 16:35
  • @Glorfindel What I'll likely do if this continues to be a need is make a post on MSE letting folks know that we occasionally do this when the need arises, and simply not notify individual sites. What I essentially just did was tell a troll to try again in 7 - 10 days :) I announced it because it's something we've never done before, but keeping stuff that we use to fight this sort of thing closely guarded is usually a better idea.
    – Tim Post
    Dec 7, 2016 at 16:37
  • And in the interim, we're putting quite a few resources into improving the way we deflect this kind of stuff. We still do a fairly good job of it, but we want to get back to being way ahead of it.
    – Tim Post
    Dec 7, 2016 at 16:39
  • 5
    Tim, while The Workplace has seen a slow increase in spam/abuse over the last month, EL&U has actually been decreasing apart from a one-time spike yesterday. Is the threshold cut here necessary?
    – ArtOfCode
    Dec 7, 2016 at 16:43
  • @ArtOfCode A lot of posts are being aggressively deleted on ELU now. Especially from a few days or a week ago.
    – user140086
    Dec 7, 2016 at 16:49
  • @Rathony Those graphs I linked are statistics from SmokeDetector, which catches spam and abuse before it gets deleted with a high rate of accuracy. It's typically good on EL&U in that it catches the vast, vast majority of posts that end up deleted for being spam or abuse. So, those stats are most likely an accurate representation of the number of posts that have been deleted by spam/abuse flags in the past month.
    – ArtOfCode
    Dec 7, 2016 at 16:51
  • @ArtOfCode meta.english.stackexchange.com/questions/9805/….
    – user140086
    Dec 7, 2016 at 16:53
  • @Rathony What about it? A raw link doesn't give me any context about your meaning here.
    – ArtOfCode
    Dec 7, 2016 at 16:54
  • 1
    @ArtOfCode After looking into it, it's looking a lot like something that's being timed. I'm going to be checking on this daily with the intent of rolling it back, but I'm not going to do that until I'm certain it was as isolated as a cursory glance would lead someone to believe - if that makes sense?
    – Tim Post
    Dec 7, 2016 at 17:20
  • @TimPost That does make sense. It also gives me an excuse to continue playing with statistics, so I won't object :)
    – ArtOfCode
    Dec 7, 2016 at 17:22
  • 1
    @TimPost This is Ask Different. Might be worth considering doing this there too?
    – ArtOfCode
    Dec 8, 2016 at 18:03
  • 1
    What's the latest on this? The waves seem to have abated. Has the system learned how to block them automatically, or the attacker decided the returns aren't worth the effort? Have the flag weights been returned to the status quo ante?
    – Dan Bron
    Dec 13, 2016 at 14:07
  • Has this tweak ended?
    – iBug
    Oct 3, 2017 at 10:01

3 Answers 3

14

I am very happy these questions are fast-deleted and am glad to see we are being given more powerful tools to deal with them.

That said, I have noticed the majority of the questions being quick-deleted are about comparing present vs past perfect, or other benign topics (though still not up to our standards).

I have not noticed any advertisement or abusive language. A quick spot check of some of the responsible users' profiles also didn't turn up anything particularly suspicious.

So, on their face, from a normal-user perspective, these posts and users do not appear to have any ulterior motives. Without revealing information which would let future miscreants game the protections on the site, can you share some details on what makes us think this is an intentional attack, either automated or coordinated, rather than a run-of-the-mill spike in "Eternal September"ness, like last time?

Also, if we are fairly confident these are intentional attacks, do we have any insights or suspicions about what the motivations are? Are the responsible parties simply testing our defenses ("casing the joint") in preparation for future attacks where they could get some meaningful benefit?

Based on Tim's comment, and tchrist's collection of examples, I am going to restricting my flagging to questions which are:

  1. New registered users (if their profile says (unregistered) then they are not part of the scheme),
  2. with 1 rep,
  3. who are evidently non-native speakers,
  4. asking about:
    1. how to learn English grammar more easily, or
    2. auxiliaries, or
    3. modals, or
    4. present or past perfect, or comparisons of present and past [perfect] tenses, or
    5. has, had, have been and so on,
  5. which are short and exhibit no research.
9
  • 7
    There's a pattern of malicious users trying to build up 'just enough' rep to downvote. I'm going to be monitoring this daily, and if it all truly looks benign, I'll roll it back. Meanwhile, I will plainly state (for the general audience) - Spam flagging stuff you just don't like is grounds for a long term suspension as you're intentionally using your abilities in a subvert way. We're giving you tools to combat cronies, if you use them on one another, we're going to need to have a talk. This was done following a discussion with your moderators; I don't think they'll be happy if abused.
    – Tim Post
    Dec 7, 2016 at 17:26
  • 3
  • @TimPost Thanks, differentiating between stuff I should flag vs run of the mill nonsense was my concern. I think I'll restrict my flagging to short posts on tenses (past vs past perfect, had va have) for now, as that's what all the quick-deleted posts seem to have in common.
    – Dan Bron
    Dec 7, 2016 at 17:31
  • 1
    @DanBron Trust your instincts is the best advice I can give. We'll monitor for folks trying to dispose of old answers they just don't like, but once in a while you just have to put more power to shields to keep everyone safe.
    – Tim Post
    Dec 7, 2016 at 17:37
  • 7
    @Rathony That's what normal close-voting is for. Flagging is for suspected spammers, & we need to be particularly careful with it now that it's got a multiplier. Remember that a successful spam-flag carries heavy penalties for OP, so overuse risks catching innocent users who simply don't know how to ask good questions, & makes it so that even if they want to adhere to our standards, they'll be locked out. Also it mis-trains the system, which is detrimental. These attacks are scripted, and scripts are hard to change, so it's likely future attacks from this source will hav these characteristics.
    – Dan Bron
    Dec 7, 2016 at 18:26
  • 6
    @Rathony - because questions are here to stay, users come and go; because this is a Q&A site, not a Josh and Rathony site; because if high-rep experienced users are reluctant to ask questions we have to make do with questions from less experienced users and their potentially good requests.
    – user66974
    Dec 7, 2016 at 19:03
  • 1
    @tchrist, Could you post screen shots or text snippets of those questions, so that those of us without 10K on this site can see the pattern? Dec 8, 2016 at 5:14
  • 1
    @BrockAdams I'd rather not just right now for fear of risking having it morph to something harder. The 10kers can see the pattern. If you're not sure, everyone should err on the side of caution and let those of who are sure deal with it.
    – tchrist Mod
    Dec 8, 2016 at 5:43
  • 1
    @tchrist, fair enough, I guess. As long as this pattern doesn't spread/apply to other sites (where we do have 10k). ;) Dec 8, 2016 at 6:08
2

One technical improvement that might be useful is for the number of close flags for each question to be displayed (with the up and down votes). Making available a breakdown of the reasons given would also be useful. Why? It just might discourage regular users from answering LQQs if they see that other users have flagged them. I know that it would be helpful for me, and would make me think twice before dashing off one of my "answers as comments". I have now started checking all low-rep users to see if they are noobs.

4
  • 4
    We need to be careful to not go overboard - unless some have special treatment, we were all 1-rep (or 101-rep) when we first joined. It's probably better for us non-mods to simply take questions at face value, flag or vote up / down / open / close / etc; and comment, edit or answer as we normally would. The mods and staff seem to have things well in hand.
    – Lawrence
    Dec 8, 2016 at 6:42
  • 1
    Mick: (1) FYI, as you’ll soon get to see, users with ≥ 3000 rep can see close (and reopen) votes now.  (Also, users with ≥ 250 rep can see close votes on their own questions, but I guess that’s irrelevant to your point.)  (2) The philosophy seems to be that users with any level of investment in Stack Exchange should learn to recognize low-quality questions on their own. @Lawrence: I’m not clear what your point is. We have been encouraged by mods not to answer LQ questions. What do you mean by “simply take questions at face value”? Dec 12, 2016 at 5:43
  • @Scott My point was that we non-diamonds shouldn't adjust our posts etc to 'work the system' against supposed offenders - this skews the stuff that goes into the repository for reasons other than content, and SE is all about content. Exceptions should be handled by those designated to handle them. For example, if we all stop responding to recent 1-rep members, they won't be able to progress very far; it would be unfair to genuine newcomers, and in the long run, harmful to the SE / ELU community - especially when we don't even know whether a particular newcomer is a miscreant. ...
    – Lawrence
    Dec 12, 2016 at 8:19
  • ... On the flip side, suppose some miscreant posted a great question, and you responded with a great answer. As I understand how SE is supposed to work, that's a great result. If they then use their newfound points and privileges to get up to no good, the diamonds have the tools to drop them right back to 1-rep, or impose whatever other penalties they deem fit. So whichever way I look at it, it's better to just take posts at face value. By that, I mean that we should deal with all questions as we would genuine questions (vote, ignore, answer, etc), all answers as genuine answers, etc.
    – Lawrence
    Dec 12, 2016 at 8:24
1

In the long run, please consider making a flag from high-rep users, e.g. over 15K or 20K, count double than other users. It will be more effective as I don't question their integrity and commitment to the site.

It will really help the site respond to increasing number of low-quality posts and spams.

3
  • 6
    I'm considering that heavily.
    – Tim Post
    Dec 7, 2016 at 17:42
  • 1
    @TimPost Thanks. I think 6 flags are too many, especially when there are very obvious spam or rude/offensive/low-quality posts. It will be better if there are some countermeasure against those who abuse their privileges. I am sure they will work as deterrent for some users.
    – user140086
    Dec 7, 2016 at 17:45
  • 4
    @TimPost Please don't do that. Dec 8, 2016 at 13:11

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .