26

So we have a situation on Travel.

enter image description here

I assure you, none of these are legit. This has been going on for weeks, at something like this volume at the same time every night, consistently. The normal IP-based blocking isn't working for these people.

At one point, Stack Exchange simply blocked all CJK (Chinese/Japanese/Korean) characters on Travel... until someone complained about not being able to write CJK characters in legit posts.

But this spam is almost completely CJK characters - easily 80-90%. I can't think of, and haven't seen, a legit Travel post with that high a percentage of these characters.

So, a solution: Block posts with >x% (probably 70-80, 90 if conservative) CJK characters. I'm not sure if the platform has this functionality yet... but it should.

So, if we do have this functionality, let's use it. Otherwise, can it be developed?

  • It might be easier to find some (likely low) number n such that very few legitimate Travel posts have more than n consecutive CJK characters and some reasonable ϵ>0, and then start blocking posts with >(1+ϵ)n consecutive CJK characters. (At least I assume that most legitimate uses of CJK characters on Travel are for giving local names of landmarks/streets/etc, and such strings would generally be fairly short.) – user642796 Sep 28 '15 at 4:11
  • Since the existing block is regex-based, measuring the length in relative terms isn't straightforward. A simpler but still reasonable criterion would be: 10 consecutive Han characters. – user259867 Sep 28 '15 at 4:56
  • BTW that pie chart would've been a good demonstration itself. – Marshmallow Sep 28 '15 at 17:55
  • To target this specific spam wave one could also add [University|College] to the regexp. – JoErNanO Sep 28 '15 at 17:56
  • 1
    Don't fix this too quickly, please (eyes "Marshal" badge). – Mark Sep 29 '15 at 0:06

You must log in to answer this question.

Browse other questions tagged .