We need better spam detection and blocking tools

Question

With popularity comes spam, and Stack Exchange is popular. It sometimes takes a few repeats before you start to recognize the same URL or product name, and blacklisting frequently spammed URLs is a non-trivial process which moderators can't even do ourselves.

For one-time drive-by spammers, the community tends to be pretty good about dispatching bad posts. Repeat offenders though are not always easy to catch, and result in a lot of work for both moderators and the community at large. We need better tools to combat this.

I would like to request the following tools, kind of in order of complexity to implement:

The relatively easy to add things...

• A search function for posts marked and deleted as spam, similar to deleted:1. Something like spam:1 would do. This would help in finding trends.

• A listing in the moderator tools of posts recently marked as spam. (This could probably even be a 10k or 20k tool). Recently deleted posts are already listed, but a lot of deleted posts aren't spam. Having this tool would help us to find repeat offenders, enabling a more appropriate response.

• Automatic listing of URLs found in posts flagged and deleted as spam. Any URL appearing in more than 3-5 spam-flagged posts (perhaps have a menu, or just filter by # occurances) would show up on the list, with a link to view all posts with that URL.

The more complicated things...

• Automated blocking of repeatedly spammed URLs. If a given URL is present in more than, say, 15 posts marked as spam, it will automatically be added to the blacklist. To avoid false positives, perhaps there should be an exception for URLs that appear at least twice as many times in non-spam posts with positive vote counts.

• Auto-flag from Community when the same URL gets posted x or more times in n or fewer days, if the URL has never previously been seen on the site. This would help to raise attention for URLs which are being spammed but might not be caught right away by the community.

There are probably other things that would be helpful too, these are just some of my initial ideas of things that would help us to catch and stop spam more easily. Feel free to add other suggestions in the comments.

Answer 1

Note: most of the following tools are currently available to our developers; not as moderator tools.

Yes, indeed... if only we could look over all sites, checking all posts flagged as spam/offensive that have been deleted and marked as successfully resolved, extract the links, and aggregate those by domain... if only those tools already existed (cough)

I'm a bit reluctant to do too much by way of automated blacklisting, as without sanity checks it could perhaps be exploited in nefarious ways - maybe that's overly paranoid. However! We also have other more manual tools available to developers, including:

• manual blacklisting of content
• automated analysis of posts by a user and blacklisting of common hosts

I'm open to thoughts on how much of a common problem it is when we get a spate of spam for a single host that we should take prompt action on. From what I've seen, it hasn't usually been big enough (except for very rare and isolated events) to get too excited about. Open to thoughts, though.

Finally, yes, we could perhaps do better with better network-wide rather than per-site blacklisting. That said, I don't want the blacklist to be enormous...

Answer 2

^{To avoid confusion: please only upvote if you think automatic deletion of associated accounts is helpful, and downvote if you think it's making things worse... I know how voting works here.}

---

Yesterday I was about to post the following as a question, but while typing I realized that automatic removal of associated accounts might also delete valuable content or might be problematic for hacked accounts, and hence be a bad idea. Or is it?

A spam account was deleted from Stack Overflow (at least 5 spam posts from May 8th to May 16th), while its associated account on Super User was involved in the same spamming too (at least one spam post, but also a nice non-spam answer).

Some time later, the Super User moderators also deleted the spam, and warned:

If you spam this product again, you will be suspended.

That's fine with me, but still just curious: would it make life easier for moderators to automatically delete associated accounts (and their content), if an account is deleted for spamming on one site?

(That would also remove any subscriptions such spammer might have configured to find future spammable posts, but I guess such subscriptions could also be maintained using a puppet, so not a lot we can do there. Also note that this specific user has associated a new SE account to their remaining other accounts. But I don't think that having a few "old" associated accounts would fool other users into giving a spammer the benefit of the doubt if spamming again.)

Answer 3

It would be nice if a user get's deleted or suspended for something like spam elsewhere on the network, to have his latest posts on other sites automatically be flagged.

Simply put them in the queue of the moderator flags like you do with low quality posts or put them there on /review. So in a way, it's just another trigger for an already existing mechanism. If the post(s) look suspicious, users can flag the user for moderator attention and they can be dealt with like any other problem user.

Like you said, it doesn't happen that frequently, so having some peer review would be good to prevent nasty side effects. But being deleted/suspended elsewhere on the network seems a pretty good marker for being a problem user