Protect. Moderate. Purge. Your. Sever.
Please. Captcha by default. Email domain filters. Auto-block federation from servers that don't respect. By default. Urgent.
And yes, to refute some comments, this publication is being upvoted by bots. A single computer was needed, not "thousands of dollars" spent.
You are viewing a single comment
This is all 100% correct. People have already written captcha-bypassing bots for lemmy, we know from experience.
The only way to stop bots, is the way that has worked for forums for years: registration applications. At lemmy.ml we historically have blocked any server that doesn't have them turned on, because of the likelihood of bot infiltration from them.
Registration applications have 100% stopped bots here.
You're right that captchas can be bypassed, but I disagree that they're useless.
Do you lock your house? Are you aware that most locks can be picked and windows can be smashed?
captchas can be defeated, but that doesn't mean they're useless - they increase the level of friction required to automate malicious activity. Maybe not a lot, but along with other measures, it may make it tricky enough to circumvent that it discourages a good percentage of bot spammers. It's the "Swiss cheese" model of security.
Registration applications stop bots, but it also stops legitimate users. I almost didn't get onto the fediverse because of registration applications. I filled out applications at lemmy.ml and beehaw.org, and then forgot about it. Two days later, I got reminded of the fediverse, and luckily I found this instance that didn't require some sort of application to join.
Don't read the first sentence, and then glaze over the rest.
But even then, however, what's to stop an army of bots from just ChatGPTing their way through the application process?
I went to a website to generate a random username, picked the first option of polarbear_gender, and then just stuck that and the application questions for lemmy.ml into ChatGPT to get the following:
I don't know the full criteria that people are approved or declined for, but would these answers pass the sniff test?
I'm just worried that placing too much trust in the application process contributes to a false sense of security. A community that is supposedly "protected" from bots can be silently infiltrated by them and cause more damage than in communities where you can either reasonably assume bots are everywhere, or there are more reliable filtering measures in place than a simple statement of purpose.
As I said in my post-
If I decide I want to write spam bots for lemmy- there isn't much that is going to stop me. Even approvals, aren't hard to work around. Captchas are comically easy to get past. Registered emails? Not a problem either. I can make a single valid email, and then re-use it once on every single instance. Writing a script that waits for approvals, is quite easy.
Wait what's the difference between the suggested auto block and you historically blocking instances without applications? Is there other criteria you use to determine the block?
Not saying I know the answer, just curious.
chatgpt.
Despite all the hype about these things being able to solve all the worlds problems, they can't answer a series of contextual questions.
Boom. Roasted.