📰 Detecting Clusters of Fake Accounts in Online Social Networks


#1

#2

Boy, if we could only apply machine learning to our own botting activities!


#3

Well, it’s theoretically possible, but what would you want to train your bot to do?


#4

For starters, find the “ideal” target based on a set of criteria.


#5

hmmm, now this is an interesting share indeed :smiley:


#6

Good share, straight from Stanford too…


#7

Thanks for the share @wortime it’s already late here :sleepy: gotta read that file tomorrow.


#8

yukkkk mathematics


#9

But what should these criterias be?

I have a bit of knowledge in machine learning, but I have no clue how you could determine a pattern of accounts that are ideal targets based on their stats.

It seems to me that the ideal targets have pretty similar content to you.
It’s extremely hard to train an AI that analyze the pictures on a base where it can really define the content type on a deeper level.

What I mean is that it’s definetly possible to determine what category an account belongs to, but as you said, not all accounts in the same category are ideal targets.

For me the target search is always trial and error and I don’t think there is really a pattern what makes a target ideal. I might also be wrong, machine learning has often found patterns where it was thought that no pattern existed (like Google’s Deep Mind that plays Go).


unlisted #12

closed #13