sibelius / ccsseraphini

cc @sseraphini
https://ccsseraphini.vercel.app/
174 stars 50 forks source link

avoid spam on cc @sseraphini #1181

Open sibelius opened 2 years ago

sibelius commented 2 years ago
arturluizbr commented 2 years ago

Use IA to label tweets into topics, then have a white, gray and black lists about topics.

(Is it viable?)

arturluizbr commented 2 years ago

Block recently created accounts into using it. Maybe 24h or less.

akinncar commented 2 years ago

multiple tags more than 5-6 @'s sounds like spam NFT bots never tag only 3-4 accounts

akinncar commented 2 years ago

we should not block NFT or other topics, what about if a developer has a question about NFTs?

sibelius commented 2 years ago

image

sibelius commented 2 years ago

image

sibelius commented 2 years ago

image

sibelius commented 2 years ago
lschaan commented 2 years ago

you could use the tweet's origin in order to help identify if it's from a bot or not and create a blacklist of unallowed origins

Screenshot 2022-11-10 at 11-52-13 🤖 Vi uma vaga (Tech) no Twitter

vniche commented 2 years ago

Tweets with lots of mentions or tags are usually scams, I'd definitely ignore those.

EronAlves1996 commented 2 years ago

Tweets with lots of mentions or tags are usually scams, I'd definitely ignore those.

This is a good Idea. It can add a negative score for every mention on tweet

sibelius commented 2 years ago
ViLourenco commented 2 years ago

you could use the tweet's origin in order to help identify if it's from a bot or not and create a blacklist of unallowed origins

Screenshot 2022-11-10 at 11-52-13 🤖 Vi uma vaga (Tech) no Twitter

Elon is going to remove it soon: https://twitter.com/elonmusk/status/1592178009410531330

sibelius commented 2 years ago

https://developer.twitter.com/en/docs/twitter-api/v1/accounts-and-users/mute-block-report-users/api-reference/post-blocks-create

vinibgoulart commented 2 years ago

only when they use the tag? or when to follow you? or always?

sibelius commented 2 years ago

for this new kind of attack, we can also check new follows