Hi everybody,

I was at the event at which Mike spoke and the proposal seemed pretty clear to me.  Here's my recollection of how he laid it out.

1) We need to improve protection against certain classes of Sybil attacks.  That is: we need to make it harder for one "actor" (person, entity, whatever) to masquerade as multiple "actors".   e.g. if I am connecting to eight peers, I'd like some reassurance that they are controlled by different people and not actually the same person pretending to be eight different people

2) There are some interesting ways of achieving this.

3) One way is "proof of sacrifice":  you could devise a scheme whereby creation of a unique "node identity" (my loose term - Mike didn't use this phrase) requires visible destruction of some small number of satoshis.  This is easy for you to do if you only want to present one such identity to the world but very expensive if you wanted to create 10,000 different identities.  So.... if you had this system, a client could make sure to connect to nodes with different identities and they could be more sure that they were controlled by different actors.  Not perfect but it would probably be OK.   Big problem though:  nobody wants to throw away their money!

4) So is there another way?

5) Mike's insight:  why don't we ask ourselves this question:  "what do most people have one of and would find exceedingly difficult to have 10,000 of?"   I guess some answers might be a house or a car or something like that... but Mike added the additional condition: "what do most people have one of and would find exceedingly difficult to have 10,000 of and *which they can prove they have over the internet*?"

6) He then pointed out that the spec of most modern passports calls for them to have an embedded chip and for the chip to have the option of including a private key that can be used to sign arbitrary challenge messages.

7) A ha!  So we already have a widely-deployed infrastructure that maps (roughly - not perfectly) one person to one private key.

8 ) So.....   you could come up with a crypto scheme that allowed you to create a node identity that everybody could see could only have been created by the holder of a passport... and which would be different for each person.... but it would not reveal anything about the person or their passport... just that the controller of that node *has* a passport.

9) Unfortunately, most passports don't implement the signing function so it looked like the idea was dead in the water

10) However, a paper presented at the May BTC conference showed that it may be possible to work around this problem and still achieve the same ends (the details are complicated and I didn't understand them).

Bottom line:  this part of the talk was all about a really interesting approach to preventing a particular type of sybil attack.