How the Credit Mesh

How the Credit Mesh Works

Credie replaces traditional credit bureaus with a dynamic, decentralized mesh of reputation signals. This mesh is distributed across mobile devices, wallets, and oracle nodes.

Each interaction updates a user's Credie Trust Index (CTI), a continuously evolving score based on on-chain + off-chain activity.

Key components of the mesh:

• Wallet Behavior: Frequency, gas usage, diversity of contracts

• Device Fingerprints: Consistency, region entropy, activity burst patterns

• Network Proximity: Interactions with other verified users

mermaidCopyEditgraph TD
    A[User Wallet] -->|tx history| B[Scoring Engine]
    C[Device Metadata] --> B
    D[Oracle Signals] --> B
    B --> E[Credie Trust Index]

The CTI is updated in real-time and never stored centrally, it's queried dynamically per credit request.

---

2. Risk Scoring Methodology

Credie uses a layered model for evaluating repayment likelihood:

1. Signal Layers:

• On-chain: Transaction reliability, token types held, protocol usage

• Off-chain: Mobile consistency, email entropy, social handles

• Behavioral: Time-on-site, scroll rate, interaction confidence

2. Machine Learning Stack:

• A graph-based GNN model processes wallet-to-wallet interactions

• A random forest classifier handles mixed signal inputs

• A risk oracle ensemble delivers a bounded risk percentile

mermaidCopyEditflowchart LR
    A[Signal Collection] --> B[GNN Relationship Mapping]
    A --> C[Random Forest Risk Classifier]
    B --> D[Risk Oracle Ensemble]
    C --> D
    D --> E[Loan Decision + Terms]

Every score is ephemeral and recalculated on request, making it tamper-resistant and adaptive to emerging behavior patterns.