Potential spam auto-detector

Apparently, a spam auto-detector has been added to Answers, which flags certain messages as "potential spam" and gives the following dialog buttons,

But when I confirm the message as SPAM and hit the delete button, I find that it takes a long time for it to respond. Is this normal?

[Star Strider]

Yes. I believe Image Analyst created it, so I’ll leave the definitive Answer to him.

I hope that it creates a log of spam messages, the characteristics of which will be used to improve detection in the future. Note that for spam that isn’t caught and marked as potential spam, there is an option in ‘Delete’ that has a checkbox to ‘Report this question as spam’. That may serve the same function.

[Matt J]

I have fresh doubts about whether this feature is working correctly. I've used it to delete the spammer "ehtasd" many times now. Not only hasn't this individual been blocked, but now the filter doesn't even flag him/her anymore. See the post (if it's still there),

http://www.mathworks.com/matlabcentral/answers/159100-great-ways-to-solve-geometry-and-algebra

[Star Strider]

Under the ‘Delete’ option you still have the ability to ‘Report this question as spam’.

I agree, the spam filter needs work. It also reports some comments as false positives, and doesn’t display spam answers in the ‘Flagged Content’ -> ‘Review questions automatically flagged as spam’. It needs to.

I just looked at that link and it flagged this Question, that to me looks perfectly legitimate: http://www.mathworks.com/matlabcentral/answers/159096-how-to-read-a-text-file-and-store-in-matrix-form

[Image Analyst]

Star's link is gone. Matt's is still there. I can only guess that since there's nothing inherently spammy about the content (no links or anything), that it didn't catch it. I don't think being flagged as spam notifies them about the author at all. I was catching kolli and some other bad actor repeatedly until I actually sent a message to John and Randy. Then I didn't see them anymore so I presume they disabled the accounts. Only once have I seen non-spam in there and it was by Chad Greene - I couldn't figure out why it caught it but I marked it as non-spam. It seems like if an author has more than 1 or 2 spams deleted, then at the very least it should notify the Mathworks so they can review/disable those accounts. Right now it does a good job of catching them but disabling the accounts is still a pretty manual process.

[Chad Greene]

Thanks, Image Analyst! :)

Answer 1

[Image Analyst]

I didn't create it. I merely logged it here. I think it must have been either Randy or John.

You're right - it does take a very long time. I don't know how a message gets flagged like that. I don't know if that happens if someone flags it as spam or if it's more automatic. I do notice in the "Flagged Content" list that there are some orphaned items that have flags that you can't get to the actual posts anymore because they were deleted. So in the past I've been looking for flagged spam posts (or closed posts via the Recent Activity link) and calling them up and first removing the flag, then removing all tags, and finally deleting the post just to make sure there will be no orphaned flags. I don't know if all that is necessary anymore with their new way of handling spam.

[Star Strider]

I know you were working on some spam filters, so I though I’d give you at least some credit for it.

I don’t remember if you mentioned this, but I serendipitously discovered that there is now under ‘Flagged Content’ in the upper-right Review questions automatically flagged as spam. I didn’t see that before, and when I explored it found that it either copied or trapped and quarantined nearly 100 spam posts over the past two weeks. They were all spam, and I deleted them all appropriately. I encourage others to look in on that from time to time and delete to your heart’s content.

[Image Analyst]

I just made up a simple m-file to do a search for terms I found often in spams. It's attached. I don't have any power to change any of the web site since I'm not a Mathworks employee.

Thanks for the note about the spam trap - I didn't know about that. There are 24 in there now and I'll get to deleting them. Not sure if you need over 3000 reputation points to see that link though.

[Star Strider]

My pleasure!

I didn’t know about it either until yesterday evening.

I thought mayhap your routine contributed to the development of the spam trap.

[Matt J]

I didn't create it. I merely logged it here. I think it must have been either Randy or John.

OK. Well, John, Randy, if you're reading this, I hope you can have a look at the issue. It seems like a good tool for combatting spam, but with the sluggishness of the buttons, spam sometimes get's generated faster than we can delete it.

[John Kelly]

MathWorks implemented a new spam filter in September

[Image Analyst]

It's interesting that comments will get noticed and flagged as spam, such as this one but don't show up in the spam trap. Does it get only Questions and Answers but not Comments? This question is actually not spam, but I left it (didn't say it was not spam) so you could look at it. I only noticed it because the date of last activity was way prior and not in chronological order.

Answer 2

[Sean de Wolski]

I like the way it makes spammers' accounts have negative questions counts :)

For the pause, I just hit refresh and it still logs it.

[Image Analyst]

Can we wire it up to send a mild electric shock to the chair of the spammer?

[Sean de Wolski]

I think we even have a toolbox to model that!

[Matt J]

For the pause, I just hit refresh and it still logs it.

Sean, you mean the delete or nondelete action goes through, and faster, if you hit refresh?

[John D'Errico]

Can I vote for a more than mild shock?

[Sean de Wolski]

@Matt, Yes, hit the button, it'll appear to hang, hit F5 (or whatever your browser's refresh is) and it still logs the button press.

@John, I'm sure you could optimize it :)