Motivation

The intention of this document is to briefly introduce, discuss and compare proposed mechanisms for registration of nodes in the Blend Network.

The goal is to guide the selection process of the most suitable design for our use case.

The focus is for edge nodes registration as the path for core nodes is settled.

The expected outcome is to have an agreement on the mechanism and some idea on the parameters we intend to use it with.

Introduction

We have investigated many possibilities for incentivization of the Blend Network. Some of them initially looked very appealing and promising like rewarding blend nodes directly by the sender of the message. However, we have discovered that they have led to many difficult problems, that could not be solved easily like:

Rewarding blend nodes for successful delivery of a message enables nodes to link the message based on the fact that they share a proof that is necessary for them to retrieve the message. The proof is the linking factor. Therefore, the message can be deanonymized only by the first and the last node on the blending path.
It is not trivial to construct the blending protocol in a way that it guarantees that the node will receive a reward for blending messages. In the direct rewarding scenario a node may be convinced it has received a unique secret while the secret was already shared (double spent). This could be bypassed but at the cost of additional messaging, which increases bandwidth that we want to keep small.
With direct rewarding it is problematic to incentivize cover messages, especially if we want to make them indistinguishable from data messages, as the cover messages must “simulate” the direct rewarding mechanism in an convincing manner to blend nodes. This opens an avenue for an abuse where cover messages are used without any restrictions to spam the network.

Therefore, we have been thinking to introduce a form of message emission quota which is distributed to nodes based on their role in the network. For the purpose of the generation of the emission quota, nodes must perform a form of “registration”, which can be seen as an interaction with the system that creates such messaging allowance for the node.

The quota makes it possible to:

Limit the total number of messages that can be emitted in the network. Which enables us to control the bandwidth consumption.
Limit the number of messages that can be emitted by registered nodes. Which protects the network from abusive nodes sending excessive number of messages (spam).
Account the number of messages that have been blended by registered node. Which enables us to reward nodes based on their contribution to the blending protocol.

Moreover, the same “registration” might be used as a necessary step to create the Core Validator Set from a set of nodes. This set forms the core of the Blend Network (and Data Availability Network). The “registration” process protects the network from abuse as it guarantees that the set of core validators is backed by a threshold of stake. Additionally, enables node to make protocol-related information publicly available.

For the purpose of this document we are considering two quota models.

The first one is the coupled quota model, where we use the same quota for cover and data messages (for core nodes).