Thread by @golemproject

So! It seems our latest article on low level defense anti-sybil mechanisms for p2p networks picked some traction (btw thanks @evan_van_ness for sharing it and tracking it!) so let's break down what Lukasz & Marcin presented in a few tweets.

One of the biggest threats to p2p networks are Sybil attacks. In a high level, this is the action of one (or many) bad actors creating multiple identities and subverting the reputation systems.

Ps. did you know? Sybil attacks are named after a multiple personality disorder popular case that spun off into a book and movie, and the patient's pseudonym was Sybil. Then later it turned out... it was fake. So basically she Sybil attacked her doctors. npr.org/2011/10/20/141…

Back to attacks: as our network grows, we need to work on better defence mechanisms. We are also building these models generalized enough to be used on other networks, because Sybil resistant tools are really needed. Shut up and take our research ;)

Model 0, the codename for the algorithm in question, identifies 2 threats. One is Sybil (it's always you, Sybil!), and the other one, well, just smart people that find it profitable to cheat.

Our approach is to minimise the risk threshold so that our providers and requestors can operate on a reasonably secure network. and this can scale."The loss rate needs to be small enough to make operating in the network profitable for honest nodes."

Take a retail store - they can't avoid theft, but surely they can count a minimum loss threshold in their accounting so that the loss is mitigated (NO, WE ARE NOT TELLING YOU THAT'S OKAY TO SHOPLIFT)

So, here's how it looks like to use Golem in terms of negotiation between requestors and providers.

(once again: NOT OKAY TO SHOPLIFT, NOT OKAY TO CHEAT // Golem teaches you to be excellent with each other)

Unknown requestors pose the biggest threat to the network and the solution - it's not that we don't like you, but we just don't know you :)
Anyways this is the logic behind the approach. If there are none or a handful of past transactions, the requestor is considered unknown.

The provider assumes all unknown requestors to be one requestor, who is then treated according to certain rules. To read them, and find out how Model 0 behaves in a high level, click the link. (not everything is spoon-fed, DYOR!) blog.golemproject.net/model-0-low-le…

If you would like to dig deeper, here is the paper we wrote on this. If you are researching Anti-Sybil, we want your feedback. If you want to implement Model 0, go for it. But send us feedback! github.com/golemfactory/g…

Ps. I had hand-drawn a Model 0 explanation for Babies but the Golem Babies did not approve it. Anyways here it is, for the lols.