"Reputation budget" standard document format proposal

[bshanks]

Here is a first-draft of a proposal for a distributed reputation document standard. Not a specific trust metric, but just a proposal that everyone chooses a standard document format for posting lists of

Comments

why the format supports lists

[bshanks]

The reason the format is a list, rather than just an individual assertion of a single trust certification, is that I want to be able to support "reputation budget" metrics like TrustFlow, by which I mean metrics that normalize the value of each of your outgoing links by the sum of the values of all of your outgoing links (i.e. if you have 3 friends, Jane, Bob, and Amy, and you rate Amy 3, Jane 5 and Bob 4, that's equivalent to rating Amy 9, Jane 15, and Bob 12). In order to do that it's important to be able to say, "here's my list and this is the ENTIRE list" (otherwise, if everything is supposed to be normalized, it's impossible to interpret what you mean when you say "Amy is a 3"). The benefit of normalization is that it creates scarcity in trust values, preventing "trust inflation".

Re: preventing trust inflation

[bshanks]

> preventing "trust inflation"

at least, I think it would help with that :)

Re: why the format supports lists

[ciphergoth]

Yes, such budgets are essential for crapflooding resistance. Interesting stuff - will read further, thanks for posting it!

Re: why the format supports lists

[bshanks]

Thanks. What is crapflooding? From the name I'd assume it means signing up a bunch of new fake users --- but they don't have any effect unless they can convince trusted users to certify them (bad nodes have no power without "confused" nodes) --- so does crapflooding also imply a specific way of getting certifications, or some mathematical form of how many such certifications are gotten, or something like that?

Also, are there any particular review papers or books you would recommend I read to get into this stuff?

thanks

Re: why the format supports lists

[ciphergoth]

http://en.wikipedia.org/wiki/Crapflooding

Supposing you can exploit the confused nodes to get one person trusted. In order to prevent crapflooding, there must be a bound on the number of people that can become trusted as a result of exploiting the confused nodes. Without your lists I don't think that's possible...

Re: why the format supports lists

[bshanks]

So the idea is, at the beginning the attacker is given links from confused nodes to one node that they control. They are then allowed to

Re: why the format supports lists

[ciphergoth]

The point about crapflooding is that we want more than a bound on the trust an attacker can create for any single node - we want a bound on the number of nodes they can cause to be trusted...