#Banning spam links

1 messages · Page 1 of 1 (latest)

tiny lark
#

I am currently banning automatically for people who use @ everyone and either use one of the links I have on a blacklist or say one of a few certain phrases like “free nitro giveaway”

#

For a while this has worked perfectly but recently they’ve stopped using @ everyone.

#

I don’t think blocking links completely is a good idea

#

Most of the spammers are server users who have been in the server for a long time but they accidentally got their account hacked

#

So we can’t do it based on server age or something, and I think just completely banning all links isn’t good because these spammers won’t be stopping any time soon

#

I’m going to continue watching the situation and seeing if there’s something I can do to adapt to the spammers

#

One option would be to remove the requirement that they say @ everyone and watch for exact phrases that they’re repeating, but then we’re still just chasing them as they modify their messages

#

Furthermore, I added the @ everyone requirement because on another server it falsely banned a very active user who was asking something about the spammers that happened to match the banned phrases

#

I also have the requirement that the user has sent less than 50 messages in the last month, but that user was also on a hiatus

subtle pawn
#

Could you create some list of allowed websites whose links users can share? Like Youtube, Instagram, Wikipedia, etc. If possible, you could see which legitimate websites people are sharing the most links for and then use that as the basis for the approved list.

tiny lark
#

I think the list of sites which people can share is too big. We’ll have people being muted for sharing links to new learning sites, grammar pages, etc

#

If they have removed @ everyone from their spam messages I think best thing to do is for me to just keep adding the new links they use

subtle pawn
#

Well it wouldn't be for auto-mute, just preventing those links from being posted? But yeah, I suppose that makes sense.

tiny lark
#

Ah so like it would delete all messages with links that aren't on the whitelist? Hmm, I see but I think it might still be slightly overzealous and annoying. I'm going to add a few of the new links in tonight and we'll see if they pop up again with new links in a problematic way too soon

vital berry
#

Honestly, the most effective solution would be to require 2FA, but I know it's not practical because a lot of people are going to hate that

tiny lark
#

Requiring 2FA only prevents people from performing moderator actions. The 2FA requirement doesn't do anything to normal users, you can't require it for the permission to send messages

#

In general 2FA is a great idea though, all my accounts have it

#

And we do require it for our moderators

vital berry
#

It indirectly stops most fake/hacked accounts from getting in the server in the first place, because it's harder to create burner accounts with 2FA on. But I get that it's a disruptive because it would require 'everyone else' to comply and change (as opposed to just targeting scammers) ||and because people don't like to secure their accounts for some reason||

tiny lark
#

Ahh I see, however, the problem is all these scammers are actually all hacked accounts that have been in the server for a while

#

They're old users that are inactive then suddenly they start spamming out of nowhere

vital berry
#

Is there a centralized list of scamming offenders? Like a "bad url list" kind of thing? Kinda what antivirus programs do to check for major threats.

tiny lark
#

If it were new users joining then I could be really lenient with keywords to watch for and just ban anyone that says them within their first 10 minutes of being in the server or something

#

I'm personally maintaining a list of URLs here:

links = ["freenitros", 'discord nitro for free', 'free nitro', 'airdrop discord nitro',
                     "hi, i'm tired of csgo, i'm ieaving", 'nitro distribution'
                     'discord.ciick', 'discordgiveaway', 'discordnitro', 'discordairdrop',
                     'discord.oniine', 'discordgift', 'bit.do/randomgift',
                     'stmeacomunnitty.ru', 'steamcommrnunity.com', 'rustiic.com']
tiny lark
vital berry
tiny lark
#

Yeah it's about 80-90% of the spammers that are accounts with less than 50 messages in the last month

#

2FA is certainly good, but from the moderation side we actually have no ability/tools to see who has 2FA, nor do we have the ability to require 2FA to send messages

#

2FA as a requirement only applies to moderation actions like Manage Messages, Manage Channels, etc

#

And we do have that enabled

unknown.png
vital berry
#

ok so I just saw one first hand and maybe it could be done like hardcore mode works? with sklearn?

left wigeon
left wigeon
#

i mean, it won't contain all of the links but it could be a good beginning

tiny lark
# vital berry ok so I just saw one first hand and maybe it could be done like hardcore mode w...

Hmm, it would be tempting to try, but the only ones the bot would recognize as valid are ones that are exactly the same as ones it’s seen before, and if we loosened it up enough to where it started branching out and guessing which new links are spam links, that learning process would scare me haha as the bot owner, I don’t want it to falsely ban a single person.

Basically, there’s not a reliable enough training method

If we leave it up to admins to feed the learning program new links, then it’s no different from admins just adding the links to a list and the bot just checking the list.

tiny lark
vital berry
#

Maybe if that blacklist is shared between bot developers and they also help train the dataset? Like YAGPDB, etc.