Clever bugspam
Larger image here, for web interface users
Jessica and I were triaging the MediaGoblin issue tracker and found some bugs that were really confusing. They both mentioned LDAP plugins, which we have, so at first they looked like real bugs... but then upon closer examination, they made no sense in the context of MediaGoblin. And hey, at the bottom there was a link... to a spam site!
I was really surprised at how accurate this bug looked... it was too good for a markov chain. @Mark Holmquist found the original post was a stackoverflow submission with terms that matched some other terms in our tracker. No wonder it was hard to detect as spam!
At the moment, I'm admiring the technique. I'm sure that level of admiration will drop off soon for pure annoyance if this keeps up though...
j1mc, Aqeel Zafar, Tyng-Ruey Chuang, Claes Wallin (韋嘉誠) and 3 others likes this.
Claes Wallin (韋嘉誠) shared this.
yes, basically it can be a markov chain with input being search engine returns for specific key terms and a lower frequency, to keep long sentences and a kind of "meaningful" effect....
just add the deadly link and you're done ... :/
but this is always interesting how this can create poetry out of trash sometime though ... ;)
@olm-e Yeah... it wasn't a markov chain though this time, it was reusing a real technical question from elsewhere. That's what I thought was surprising.
yes ... it's clever indeed to respect full posts, as it keeps coherence and looks like its just not "posted in the right box" ... what I thought was the use of a markov chain of keywords, not directly in the text, to treat websites forms and spam text from public forums content, like question and answer...
@olm-e Yeah, that's an insidiously "successful" strategy, unfortunately...