Most of the spam I get on Twitter is from accounts that have not even any followers. Of course you can grow accounts to enter people's social graph but those losely connected can be identified. An impersonator marked as 5th degree contact would have a hard time pretending to be somebody I assume to be following already.
Discussion
I’m all for trying lots of solutions. Sadly like SEO it’s often a game and a moving target. That’s why I liked the POW concept, at least there is a cost. I think we’ll end up with a fair few strategies working in unison.
One difficultly is shifting from a centralised scoring/identity system, to a per client compute approach. Less data, less power, less compute, less storage, etc. I think we’ll get there however.