Meta community. Discuss about this lemmy instance or lemmy in general.
The botted account proliferation is escalating fast. Almost all top instances one can see in the fediverse observer are now bot-filled ones
I contacted their team to ask them to gather more details so that we can use the Overseer to combat this in advance, but they're disinterested in helping so no assist will come from there. Fairly disheartening I must say.
I was not planning to get into this sort of api polling, so I'll have to see how feasible it is for me to develop it the overseer itself.
DbZ, how feasible is it to have some sort of analysis tools to detect bots regardless of their home instance.
Right now they are proliferating in specific instances but those are easier to block. More concerning is the use of bots via established instances.
Since we can see the user profiles of even remote users, how likely is it that we can detect bots accounts by running some sort of analysis? The benefit here is that it should be able to detect even human-run astroturfing accounts.
I have no idea how I would be able to achieve this tbh. If someone tells me the methodology, I could implement it
I'm not familiar with locally run text analysis models, but there should be a way to analyze comments and flag them based on certain topics.
This is a far more complicated approach but I suspect it could be ultimately necessary.
Say you've got a user who's been shilling a certain product regularly. They've been active 24/7 without seemingly sleeping and/or share the same IP, ISP or IP geolocation as other users with a similar pattern. You should be able to have a confidence metric that this user is involved in an astroturfing campaign. (Although you won't and should not get IP addresses for remote users).
Slightly offtopic but a funny thought: I wonder if there's a way to engage with these bots, ask them something that an AI would respond to in a flawed manner and detect them based on their replies or lack of replies.
Edit: I just remembered that I read that even if not displayed in the UI, upvotes are public because they correspond to favorites on Mastodon. It should be much easier to find a correlation of certain accounts consistently upvoting or downvoting certain posts
As an side not to what's going on with bots. I'm noticing a LOT of communities being open by the same person and the content within feels AI generated (usually images). Anyone else noticing this or am I too deep in tin foil hat territory?
I've noticed this happening on Facebook too, my feed is suddenly full of generated art. It's super annoying and non of it is tagged as AI generated, but it obviously is. A lot of stupid boomers seem to think it's real judging by the comments, though I suppose the comments could be bot accounts too. If social media platforms don't build tools to counter AI generated content we are all going to drown in it before long. I don't mind it if it's tagged as such and in it's own community I suppose. Short of some sort of whitelist system, I have no idea how they are going to stop it - I mean moderators simply won't be able to keep up with the volume of AI generated posts at some point.
I think we're in for a bit of a flood with this stuff. I had to unfollow a and subsequently block a bunch of nsfw instances because it was just awful AI image spam.
which person, which communities?
Not on this instance. Was more of a general comment.
But as an example..
the User AkhuyanATlemmy.world
Ah you mean elsewhere? Yes it's very possible spammers are already "building up" accounts for later.
That sucks, but I guess inevitable.
Unless I'm the one misunderstanding, they aren't opening a bunch of communities, they're a mod of [email protected], where they make posts alerting users that new communities have been created. Their posts look weird because they're using a template that grabs the post title from the community description, so if it says something strange then their post will too, but those aren't their communities.
I thought they were creating those communities as well.