On Fri, Jul 13, 2012 at 1:33 PM, Morten Olsen Lysgaard
<[email protected]>wrote:
A year ago I researched DHTs quite extensively. I found them vastly
interesting for a long time dream of mine. A time-shared cooperative
storage network. A system, where public files can be stored.
At the time i was thinking a gigantic music storage service which would
let people stream songs from other peers as long as they contributed back.
I implemented Chord in Haskell and with it a replicated and fault tolerant
storage layer, but then I realized that I had no knowledge of securing this
DHT. I didn't even know what security would mean on such a network. How do
you prevent malicious nodes from destroying all the data, or creating a
million peers. How do you enforce good behavior so that people share their
disk space and bandwidth? All these questions popped up in my head.
TLDR; I'm wondering if anyone here have experience with, or know of,
techniques for securing the function of structured overlay networks with no
central authority. I find this problem really really hard.
This is of course a massive unsolved problem, and any useful advance you
make is likely to be widely imitated.
It is related to the problem of preventing "search engine optimization",
at which task google keep a very large team of very smart engineers
working full time. The measures they have taken to reduce search engine
optimization have had a significant adverse effect on google's search
results, as has search engine optimization itself.
One of the many problems to be solved is trust. We have an existing
solution that works very well: The git model: This model allows us to
assemble a very large body of true data and protect it from bad people.
Most people choose to trust one guy, call him the king, or unelected
president for life. He trusts a handful of people, and accepts their
version of the truth. Call them the lords. Each of those people trust
a bunch of other people, and so on and so forth, so a big and detailed
collection of true data from a big and diverse bunch of people is
collected together and ultimately approved by one guy.
This does not sound very decentralized, but it is decentralized indeed,
because one can have, and often do have, several kings, and always do
have numerous potential kings, and anyone can choose which king he has
confidence in.
It is not a star network, but rather a network where the number of
really important nodes is manageably small.
And in practice, we have several different skews of the truth, several
different skews of the big pile of truth.
_______________________________________________
p2p-hackers mailing list
[email protected]
http://lists.zooko.com/mailman/listinfo/p2p-hackers
