cross-posted from: https://sh.itjust.works/post/1823812
This is an update to my previous post about suspicious inactive accounts on a handful of instances: (https://sh.itjust.works/post/998307).
I ended up messaging the admins at the 16 instances show in the attached image. I pointed out their wild user numbers, and referenced the lemmy.ninja post detailing how that instance scrubbed suspicious accounts from their user database.
6 admins responded. They had all noticed the odd accounts and either thought the numbers were wrong, or weren't sure how to purge the suspicious accounts without nuking their databases. In the end they managed to delete a combined total of about 338k dormant accounts from their instances. (One of the instances seems to have gone down since then.)
I never received a reply from the other 10 instance admins, though 8 of those 10 instances appear to be down (as of 27 July 2023). 2 instances are still up and unchanged.
Between the actively removed accounts and the downed instances, this represents a loss of 930,004 inactive Lemmy accounts!
You can see the drop in the graphs on The Federation. The total number of Lemmy accounts has been cut in half over the past 3 weeks, from a peak of 2.18M to today's 1.09M. The change is mostly from these 16 instances.
I have to admit, I did not expect such a large change when I started this! Hopefully this bodes well for Lemmy's future as a place where actual humans interact, rather than a cesspool of automated comments and upvote/downvote brigading.
That's all I have for now. Keep your stick on the ice; we're all in this together.
When an account is signed up, is there information such as client ip address that could also be used to spot more inauthentic activity? And more generally, sign up should probably be made resistant to automated bots by randomizing HTML layout & ids and using captchas so it's not so easy to drive sign up through scripts.