On Tue, Mar 27, 2012 at 5:45 AM, Michael Rogers <m...@gmx.com> wrote:
> On 27/03/12 03:41, Tony Arcieri wrote: > > By only allowing one degree of separation, instead of: > > > > [You] -> [Trusted Partner] -> [Potential Sybil] -> [Potential Sybil] > > -> [Potential Sybil] -> [Potential Sybil] > > > > you instead just have: > > > > [You] -> [Trusted Partner] -> [Potential Sybil] > > Why is one worse than the other? Either way the loss is the same, and > either way it's my trust in the trusted partner (and nothing else) > that's affected. So let me say what I'd really like to do: I would like to use collaborative filtering to make predictions about "third party" peers based on your direct experiences with peers you know. A specific example of collaborative filtering algorithms is Slope One. Singular value decomposition can also be used for collaborative filtering. The general idea is that you can index the interactions of peers (i.e. their "block chains") into a database, then input your empirical data about your past interactions with peers on the network, and it can generate recommendations for peers you can interact with. This allows you to cull valuable predictions about good peers from a noisy data set of potential sybils. -- Tony Arcieri
_______________________________________________ p2p-hackers mailing list p2p-hackers@lists.zooko.com http://lists.zooko.com/mailman/listinfo/p2p-hackers
