How Secondary Markets Are Manipulated
Circularity fraud in South Asia takes three primary forms:
- Image Re-upload (Duplicate Billing): A single truckload of copper scrap is photographed at multiple angles, generating 5 separate PDF invoices sold to 5 different brand owners claiming ESG credits.
- Wash-Trading Cycles: Shell recycling aggregators pass fictitious invoices in circular loops ($A \to B \to C \to A$) without moving a single kilogram of physical metal, inflating statutory EPR fulfillment numbers.
- Weighbridge Tare Tampering: Truck drivers carry hidden water ballast tanks dumped immediately after gross weighing, artificially inflating scrap mass by up to 35%.
Agent 05 is the gatekeeper. If any fraud signature is detected, the transaction is rejected at the API layer with cryptographic evidence logged on-chain.
The Sentinel Mathematical Engine
1. Perceptual Image Hashing (pHash) & Hamming Distance
Gate 1Computes discrete cosine transform (DCT) frequency fingerprints of every uploaded scrap photo. Measures bitwise Hamming distance against the global registry of historical lots:
2. Directed Acyclic Graph (DAG) Wash-Trading Cycle Detection
Gate 2Maintains a real-time adjacency graph $G = (V, E)$ of wallet addresses and depot IDs. Executes Tarjan's strongly connected components algorithm before every trade to detect circular transaction loops.
3. Digital Weighbridge Telemetry Z-Score Outlier Analysis
Gate 3Compares visual volume contours from Agent 01 against reported weighbridge mass using density constants ($Z = |M - \rho V| / \sigma$). If $Z > 2.5$, physical tamper inspection is flagged.