20. From Detection to Sybil-Resistance

The answer was never better detection.

That is the door the last chapter walked you to and refused to close. Chapter 19 did the proof in full, so I will not re-derive it here: Douceur showed that without something costly binding an identity to the actor behind it, no amount of inspecting identities from the outside can collapse a thousand puppets back to the one operator, and EigenTrust drew the corollary that the durable object of trust is the actor, not the artifact. Every defense in this book treated the artifact — the message, the link, the account, the behavior — as the thing to judge. That was the category error. The artifact is downstream. The actor is the cause, and the actor was never the thing we were measuring.

So this chapter does something the previous nineteen did not. It stops trying to win the fight we have been losing and changes the question being asked. For the entire history of open-community defense, the question at the door has been: is this message spam? Or some refinement of it — is this account a bot, is this link malicious, is this behavior anomalous? Every one of those is a question about an artifact, asked after the artifact exists, answered by a defender who must be right every time while the attacker must only get through once. We are going to replace it. The new question is the one about the actor — who is this, and what is their standing? — asked before the artifact, and answered not by the defender's cleverness but by the actor's history. There is not one clean answer to that question. There is a family of them, converging, and naming that family is the work of this part of the book.