Trust Labels

Introduction

Trust labels constitute integral aspects of the Proof of Reputable Observation (PRO) system.

They impact several facets, such as determining the nodes from which to obtain snapshots. Trust labels are off-chain information concerning the security of nodes and their potential to harm the network. These labels are local bias values that you supply to your nodes during the joining process to the cluster (metagraph). They are specific to each node. Consequently, different nodes can exhibit varying degrees of bias.

note

On a blockchain, off-chain information refers to data and transactions that are not recorded directly on the blockchain. It describes any data or processes that occur outside the main blockchain network.

Trust labels are numeric values between -1 and 1.

• -1 is the highest level of distrust
• 1 is the highest level of trust.

A node’s trust label can be described as its rating, similar to a customer’s review of a product or movie.

Example

• A node you know to be very secure can get a high trust label rating (e.g., 0.9).
• Nodes operated by someone you trust (e.g., yourself) might get scores of 0.6 or 0.7.
• If you know a node is not a bot, but are otherwise unsure of that particular node, you might give that node a rating of 0.5.
• If you suspect a node of being in a botnet, you might assign them negative trust label ratings (e.g., -0.6).

Negative Trust Labels

Assigning a negative trust label to another node should indicate that you have some information external to the network suggesting that the peer might be involved in, coordinating, or associated with malicious activities or negative actors. Negative trust labels don't prohibit a node from participating in the network, and they don't act as a centralized blacklist. Instead, they function on a per-node basis, preventing the spread of additional scoring information in the model.

Example

If Node B views Node A with a -1 trust, Node B will download from Node A unless a trusted node is available.

They are also used in situations where multiple versions of the network (forks) exist to reduce the influence of votes associated with a specific version. For instance, you should consider applying a negative trust label if a node clearly unrelated to the project or community starts promoting a particular version as the "correct" fork, even when it involves an obvious attempt at double spending or changing balances.

Neutral Trust Labels

0 indicates neutral trust and is distinct from the absence of a trust label. It means you know about the node and have no reason to trust or distrust it.

Note

In general, avoid giving a node a 0 trust label. Instead, don’t provide a trust label.

No trust label means you don't know about the node.

Do Not

Do not provide labels for nodes you do not know about.

Trustworthiness

As your nodes engage with other nodes on the cluster, they will form individual assessments of each peer's trustworthiness. Consequently, the trust labels you initially supplied become less relevant. Trust labels wield their greatest influence when no prior knowledge or interaction history with other nodes on the network exists. These labels represent your unique perspective on your peers' trustworthiness and contribute to the security of your nodes.

It is not necessary to duplicate another node’s trust labels, as PRO automatically assimilates data from peer nodes, obviating the need to replicate scores from others.

Supplying Trust Labels

When your node initiates a join to the cluster (metagraph), use the --ratings argument and provide the path to a flat file containing the trust labels.

The file is a text file with a single trust label entry per line in the following format:

Node ID, Trust Label

An example flat file is available here.