Another network of porn spam bots. They get the attention of users by liking and retweeting their Tweets. Pinned Tweet with spam URL. They can be identified through the URLs and phrases they use. All URLs forward to datewith-me1 dot com. @TwitterSupport
/via @element_arb
/via @element_arb
I followed to urls, so you don't have to. After datewith-me you get to sexclub6 dot com and from there to wwa.6 dot dating. This seems to be the final location because after singing up you finally get the option to give the spammers money.
Others noticed them as well: @smeschmesch and likes on this tweet by @leanderwattig
Looked through 3k Tweets and found 14 URL they use:
mommy2 info
me2much info
checkmama info
gotohook pro
higuys3 info
mojojoken info
qo2hook pro
t4keme info
move4 pro
whynot2 info
m4go pro
hook4me info
sobaken pro
g1rlshot info
NSFW search to see them: twitter.com/search?f=tweet…
mommy2 info
me2much info
checkmama info
gotohook pro
higuys3 info
mojojoken info
qo2hook pro
t4keme info
move4 pro
whynot2 info
m4go pro
hook4me info
sobaken pro
g1rlshot info
NSFW search to see them: twitter.com/search?f=tweet…
I don't know yet how they choose Tweets to Like/RT.
Other bots target accounts based on the username.
Other bots target accounts based on the username.
I analyzed 11k Tweets by 11k accounts. (Domains as search query).
Most of the accounts are old. From 2011-2014. As far as I remember it was much easier to create accounts back them. I don't know if they are old spam accounts or hacked and scrubbed accounts.
Most of the accounts are old. From 2011-2014. As far as I remember it was much easier to create accounts back them. I don't know if they are old spam accounts or hacked and scrubbed accounts.
Most of them follow 60-90 accounts.
They have nearly no followers.
Few Tweets (<20).
50-100 Favs per account.
5k en, 1k es, 1k fr, 500 ko, 500 pt, 300 it, 300 pl, 260 de, 230 hu, 200 th.
They have nearly no followers.
Few Tweets (<20).
50-100 Favs per account.
5k en, 1k es, 1k fr, 500 ko, 500 pt, 300 it, 300 pl, 260 de, 230 hu, 200 th.
is on them as well.
I threw a dozen of spam bots at @Botometer. Only one was clearly identified as a bot. The rest got probabilities between 31% and 57%. My own account gets 33%. I am more likely to be a bot than [NSFW] @susan58318147.
@pewinternet found 37% most accurate pewinternet.org/2018/04/09/bot…
@pewinternet found 37% most accurate pewinternet.org/2018/04/09/bot…
Botometer is probably the most notable bot detection tool in the scientific community. Like most ML, I find it useful to get a bigger picture (ratio of likely bots in a big group of accounts) but mostly useless when used on individual accounts.
twitteraudit.com isn't intended to identify bots but fake followers. I am glad that the four bots I tested got nearly flawless scores. Only the realest of real followers.
I wanted to give botcheck.me a chance as well, but it seems broken at the moment.
/cc @Pudingversteher
/cc @Pudingversteher
My own @accountanalysis tool completely fails as well, because it only looks at Tweets.
Another ML-based tool. github.com/mkearney/botrn… by @kearneymw. It's the most accurate one I have tested yet. Better training set or different features? Will have to look through into the code.
/thx @raymserrato for the suggestion.
/thx @raymserrato for the suggestion.
You don't have to install R to use it. There is a web app: mikewk.shinyapps.io/botornot/ (too late for me, hello Rstudio, what will we do next?)
