5.2 Reputation Scoring Vector

Omni-Score () It is a dynamic numerical value between 0 and 1000. It is not a static stored value, but rather based on the user's historical interaction vector The function results are calculated in real time.

5.2.1 The Decay Algorithm: calculation

To reflect the "time-sensitive nature of reputation" (i.e., honesty a year ago is not as important as honesty yesterday), all positive behavior integrals are subject to the Half-Life Algorithm.

Basic Decay Formula:

The user's current reputation $S(t)$ is defined as:

$$(t) = \sum_{i=1}^{n} \left( P_i \cdot e^{-\lambda(t_{now} - t_i)} \right)$$

• : The base points earned from each successful transaction.

• : Current block timestamp.

• : The timestamp of the completed transaction.

• : Decay Constant. Setting,This means that the half-life of the credit score is approximately 6 months.

Mechanism significance:

This means users cannot simply rely on their high scores accumulated early on. If a high-scoring user becomes inactive, their score will gradually decline over time, eventually returning to the baseline. This forces business nodes to maintain consistent and stable performance in fulfilling their obligations.

5.2.2 Multi-dimensional Weighting

In addition to the time dimension, the basic points for a single transaction It is determined by a weighted vector of three core dimensions: transaction volume, interaction frequency, and counterparty quality.

Integral generating function ：

$$P_{Tx} = \underbrace{w_1 \cdot \ln(1 + V_{usd})}_{\text{Volume (Logarithmic)}} + \underbrace{w_2 \cdot \mathbb{I}_{diverse}}_{\text{Diversity Bonus}} - \underbrace{w_3 \cdot \text{Risk}}_{\text{Risk Factor}}$$

1. Volume: Logarithmic Growth

We are concerned about the transaction amount Using natural logarithm Processing, rather than nonlinear processing.

• Reason: To prevent "whale hegemony." The reputation boost from a single $1 million transaction should be far less than that from 1 million $1 transactions. Reputation comes from proven honesty, not simply from displays of wealth.

• Example: A transaction of 100 U earns 10 points, while a transaction of 10,000 U earns 20 points (instead of 1,000 points).

2. Frequency & Diversity

To defend against "Sybil Washing" (i.e., creating two separate accounts to trade and inflate scores), a counterparty diversity coefficient was introduced().

• The algorithm constructs a local social graph 。

• If a user always transacts with the same address, It will decay exponentially to 0.

• You can only earn the full credits by trading with new, high-reputation counterparties.

3. The "Glass Cannon" Penalty Logic

In the Omni-ID model, building reputation is like hiking up a mountain, while destroying reputation is like falling from a great height.

Once a user is determined to be "fraudulent" in DAN arbitration (i.e., loses the arbitration and is marked as malicious), their reputation score will trigger a nullifier:

$$S_{new} = S_{old} \times (1 - \text{SeverityFactor})$$

• SeverityFactor: For serious fraud (such as providing false logistics), this factor is 1.0 (directly reduced to zero); for minor breaches of contract (such as delays), this factor may be 0.2 (subtracted by 20%).

• This asymmetrical design ensures that the cost of wrongdoing is extremely high, making rational users cherish their Omni-Score as much as they would their feathers.