As some of you might know, I am part of a team of people moderating a Mastodon server. And as so many of those servers we’ve been having some problems with fake signups lately. This is about how we mitigated the issue and how it was beneficial beyond that. So for those who don’t know […]
I think Lemmy and Piefed are handling the spammers and ai reposts well enough. I’m also on Discuit and it’s ramped with reposters, ai bots, meatspace proxies, and I assume bot upvotes, which is sad.
Lemmy definitely has LLM bot accounts, but the Lemmy is operating at a scale of tens of thousands of users where Mastodon has a couple of million. There’s a lot of manual action by Lemmy admins that probably won’t scale well when there are millions of users.
Only if that scale happens across the current number of servers. If the server base grows along with users the manual management may scale much better
Unfortunately Lemmy has one major problem when it comes to banning spam bots. Only the instance that a user belongs to can ban the user in a way that federates to other instances. This means if that particular instance doesn’t quickly ban the user, then hundreds of other instances need to do it manually. Currently, a pretty significant percentage of spammers get banned individually by many instances.
If we had 100x as many users and 100x as many instances, we would have 100x as many bot accounts for every instance to ban.
The solution is probably something like shared ban lists or ban mirroring (setting an instance up to ban every user that some trusted instance bans). But right now it’s just people clicking, and it’s not scalable.
This could be mitigated quite easily by having a clear separation between instances run for users and instances for communities.
Hmm, yeah that seems like it does need a technical solution
Thank you for taking the time to explain! ☺️