It's interesting, but I'd really like to see the paper expand more on the node/identity.

What's stopping me from writing software that floods the network with hundreds of thousands of identities, all under my control?