3 ms·
I suspect there is an opportunity in working on the other side of this equation, providing a service to identify suspected spam accounts. You could either charg
by chatmasta 12y ago
I suspect there is an opportunity in working on the other side of this equation, providing a service to identify suspected spam accounts. You could either charge for access to the list, or publish it as a public service, in the same way that abusive IP addresses are published.
You could identify spam accounts in a few ways. A naiive, but probably effective approach, would be buying the "fake follower" services yourself on places like Fiverr [1]. But that would cost a lot of money quickly. A more sustainable solution would be applying any number of machine learning techniques, which have been shown to achieve 80-90% accuracy. apply any number of machine learning techniques with a fair accuracy. [2]
Personally, I would like to do some research into this area. I would assume you could get pretty far just by graphing the networks of accounts, and comparing the shape of them to authentic accounts.
[1] fiverr.com
[2] http://scholar.google.com/scholar?as_ylo=2013&q=fake+twitter+account&hl=en&as_sdt=0,5 http://scholar.google.com/scholar?as_ylo=2013&q=fake+twitter...