Content
@
https://warpcast.com/~/channel/farcasterunion
0 reply
0 recast
0 reaction
Njal
@cryptonjal
I knew that spam accounts are created en masse, but it surprised me that some were very easy to filter out on Dune surprised me. It gives an insight into the success rate of 'spam' accounts. - 3,063 accounts that showed the same behaviour in groups of more than 15; - There even seems to be a group of 500 to 800 accounts that have almost all received a spam label 0 after around 13 replies and 13 likes (disposable accounts); - From these 3,063 accounts, there were 93 with a label 2, and 2 without a label. That's a success rate of 96.9%, the other 3.1% are false positives, and no labels. - A large proportion of the accounts not identified as spam (89) were in smaller groups, which may suggest a different strategy of these groups of accounts. Here, the success rate is 90.4%. 9.4% of accounts slip through.
5 replies
4 recasts
11 reactions
Njal
@cryptonjal
Dataset 1: https://dune.com/queries/5254390 Dataset 2, with only the small groups: https://dune.com/queries/5265275 You can filter out the label 2 spam accounts by searching the query results for ☣. This is not a total overview of spam accounts, but it is clear that there is efficiency to be gained by looking at group behaviour.
2 replies
1 recast
7 reactions
Mikko
@moo
This bot is trying really hard. Maybe it run out of ChatGPT credits.
2 replies
0 recast
1 reaction
Colin Charles
@bytebot
What’s happened I suspect is they were just suspended from X. Basically X did a massive bot clean out and there’s a good bet they setup here verifying their X accounts. If we traverse all the bot activity here with suspended X accounts, you’ll probably realise they’re a match. These are Internet Water Armies
3 replies
0 recast
1 reaction
Njal
@cryptonjal
It's interesting anyway that there are ferivied x accounts on FC that disappeared on X... X-spam-bots recycling turns out to be too easy
0 reply
0 recast
1 reaction
