Dealing with SPAM

[venomgfx]

We can learn a lot from the current spam that blender.community is getting, here we can list the issues.

• [ ] Block certain titles (e.g. including phone numbers)
• [ ] Prevent publishing duplicate titles (also helps to avoid the 'Awesome Post Title' issue)

Spam examples:

[muhuk]

• We can prevent or rate-limit posts by new users.
• We can add a Report button (eventually someone deals with the spammers right? This could help alert them faster)
  • When a critical number of reports are reached a moderator gets an e-mail.

[L0Lock]

• [ ] Block certain titles (e.g. including phone numbers)
• [ ] Prevent publishing duplicate titles (also helps to avoid the 'Awesome Post Title' issue)

I should point out that, while effective against some origin of spam, it is far from enough because very easy to go around and still post anything.

The following would be the bulk of the filtering:

• We can prevent or rate-limit posts by new users.
• We can add a Report button (eventually someone deals with the spammers right? This could help alert them faster)
  • When a critical number of reports are reached a moderator gets an e-mail.

In some websites, there are "critical numbers" that lead to automatic actions. The actions are usually alike hiding the post (+ moderator/admin notification of said action) and can of course be undone at any time by the moderators/admin. And are usually also detecting if the critical number of reports is reached within a few minutes/hours after posting, to better troll-proof misusage of this feature.

Some site also compare the thread contents & comments, and try to recognize patterns like fishy external links, long texts over XX% identical, ...

Spammers will always find something so run through the net, that's why there's a need for a robust reporting system.

In a forum I'm moderator on, spammers got the idea to copy paste very old topics to create a new one, then come back a few days/weeks later and edit their original topic to replace it by their spam message.

[fsiddi]

A few more points about this:

• Lookup for keywords in post content and title, as well as comments
• Tweak the spam deletion system to make accounts inactive, preventing creation of the same account over and over

[L0Lock]

This spammer account seems to have been free for at least 2 months now:

Sophiewilson - @sophiewilson — Blender Community

Is there a pattern that can help to deal with it automatically? Were its post reported as spam significantly more than the average fake-reported post?

We also got this new one yesterday:

Liberty Insurance - @liberty — Blender Community

They are seemingly low-risk spams, but the day someone wants to post something really hurtful and is smart enough to bypass a-priori filters, it will hurt bad.

Could it be considered to have a system like Discourse (I.E. Blenderartists) where a post with too many reports is hidden, but users can still click to see the post, knowingly of the potential issues.

• Flag a post for moderator attention - documentation / users - Discourse Meta
• What happens when you flag a post - documentation / users - Discourse Meta

It's pretty powerful, responsible users can still view content, posters can rectify once (prevents spammers to just edit infinitely to bypass the system) and moderators always have full control.

Edit: new spam accounts today:

• snbaba - @snbaba — Blender Community
• Pandit Aman Sharma - @pandit_aman — Blender Community

Both posting with phone numbers.