Keywords for use with PieFed’s auto-ban system

When someone creates a post in a community on PieFed, copies of it are sent out to thousands of servers. Spammers love this because it’s effectively a spam amplification device. When an admin on the spammer’s instance bans them, their spam posts are removed on all servers. Or if a community moderator on the instance hosting the community the spam was posted into deletes the spam, it’s removed from all servers. No problem, so far.

But when the spammer’s admin is absent, or the community moderators are absent, it falls to the admins on every receiving server to manually and individually clean up the spam, on each server, separately. Half the time this doesn’t happen because the admins there are absent / burntout / etc. So quite often, in some places, the spam remains.

To reduce the burden of cleaning up this kind of spam, where the source admins and the community admins don’t react, PieFed v1.9 added an ‘auto ban’ feature, where instance admins can designate other instances to monitor for bans. This feature uses bans rather than content deletions because usually the quickest way to deal with a spammer is to ban them, which removes all their content too.

Setting this up is quite simple.

In the first box, type the domains (without the http:// on the front) of an instance or two that you trust. In the second box, put a few keywords for the types of ban reasons that you want to accept.

Then, whenever someone on an instance you trust bans an account, that account will also be banned (and it’s content removed, if the banner selected that option) on your instance.

Here are the ban reasons we will be using on piefed.social. If you put piefed.social into “Instances to follow”, use these keywords one word or phrase per line, in the Keywords field to accept that type of ban:

  • Spam
    Posts with a commercial purpose, done by a brand new account with no community history
  • Scam
    Similar to spam although with deception involved
  • Vote manipulation
    An account showing clear signs of inauthentic behaviour, alone or in a group, intended to alter post or comment rankings.
  • Harassment
    Repeated attacks over an intended period, intended to silence or intimidate a victim.
  • Troll
    Inflammatory behavior intended to cause disruption. Includes bigotry and hate speech.
  • LLM written posts
    Repeated use of a LLM to generate content in posts or comments.
  • LLM agent
    Every post and interaction is LLM generated, creates new posts autonomously and replies to comments.
  • 4chan
    Repeatedly posts 4chan screenshots on a regular basis.
  • Sockpuppet
    Same as vote manipulation although tends not to post or comment – account exists purely to vote.
  • Serial downvoter
    Follows a victim around and downvotes everything they post, usually in quick succession and sometimes using automated means to cast many votes.
  • CSAM
    Disturbing content regarding children.

Leave a Reply

Your email address will not be published. Required fields are marked *