Smart Contract-Based Supply Chain Traceability System with Real-Time Anomaly Detection using Graph Neural Networks and Federated Learning

Project Abstract

The rapid globalization of supply chains has intensified the need for transparent, tamper-evident, and traceable provenance of goods across multiple stakeholders. This research proposes an integrated framework that combines smart contracts, graph neural networks (GNNs), and federated learning to achieve end-to-end traceability, real-time anomaly detection, and robust data privacy in decentralized supply chain ecosystems. The proposed system models the supply chain as a dynamic heterogeneous graph where nodes represent entities such as suppliers, manufacturers, logistics providers, warehouses, and retailers, and edges capture transactional events, material transfers, and custody changes. Smart contracts on a permissioned blockchain automate and enforce governance rules, provenance immutability, and automated compliance checks, while enabling auditable history and tamper-resistant records. To enable real-time anomaly detection, we design a scalable GNN architecture that leverages temporal and relational information to identify suspicious patterns such as counterfeit activity, route deviations, inventory mismatches, and anomalous lead times. The model incorporates attention mechanisms to weigh the importance of different relationships and temporal windows, providing interpretable insights for investigators. Recognizing the sensitivity of operational data, the framework employs federated learning to train global anomaly detection models without centralized data aggregation. Local models are trained at participating organizations and only model updates are exchanged, preserving data privacy while benefiting from collective intelligence. A privacy-preserving aggregation protocol with differential privacy guarantees mitigates the risk of leakage through gradients. The system architecture integrates secure data ingestion pipelines, standardized event schemas, and off-chain storage for large telemetry while maintaining on-chain provenance proofs. Access control, role-based permissions, and notarization features are embedded in smart contracts to ensure accountability and compliance with regulatory requirements. We address data integrity, provenance provenance, and latency challenges by using cryptographic hashes for event verification, sidechains for scalability, and streaming techniques for timely anomaly alerts. The evaluation encompasses a multi-case scenario spanning perishable goods, electronics, and cosmetics to demonstrate adaptability across domains. We measure traceability latency, provenance completeness, anomaly detection precision, recall, and F1-score, as well as privacy metrics such as data exposure risk and communication overhead. Empirical results show that the integrated system achieves near real-time traceability with sub-minute update propagation, a high degree of provenance completeness across nodes, and robust anomaly detection performance with minimal false positives under federated learning constraints. Ablation studies reveal the contribution of temporal graph features, attention-based relational reasoning, and privacy-preserving aggregation to overall effectiveness. The research also discusses deployment considerations, including governance models for consortium networks, governance of smart contract upgrades, and interoperability with existing ERP and MES systems. Potential limitations such as network latency, model drift, and adversarial manipulation are analyzed, and mitigation strategies are proposed. The work contributes a holistic blueprint for secure, scalable, and privacy-preserving supply chain traceability that leverages cutting-edge blockchain, graph-based analytics, and distributed learning to enhance trust, efficiency, and resilience in modern logistics ecosystems.

Project Overview

What This Project Is About

---

This project explores how smart contracts can track products through a supply chain, while using machine learning to spot unusual patterns in real time. It combines blockchain safety with easy-to-understand data analysis to improve transparency and trust.

---

---

The Problem It Addresses

---

Visible, tamper-prone records make it hard to verify the journey of goods, leading to fraud and quality issues. There is a need for automatic checks that can flag anomalies as goods move from supplier to consumer.

---

---

Objectives of the Project

---

1. Design a simple blockchain-based ledger to record key supply chain events.
2. Implement smart contracts that enforce rules and enable traceability.
3. Develop an anomaly detection system that runs in real time and alerts stakeholders.
4. Explain how graph-based patterns help identify unusual routes or suppliers.
5. Demonstrate how privacy-friendly data sharing can be achieved with federated learning.

---

---

What You Will Do Step by Step

---

1) Review basic supply chain processes and identify data points to capture. 2) Set up a simple blockchain environment and write smart contracts. 3) Build a graph model to represent product movement. 4) Integrate a real-time anomaly detector and test with synthetic data. 5) Explore federated learning to share models without exposing data. 6) Evaluate performance, accuracy, and usability. 7) Document design choices and potential ethical considerations.

---

---

Expected Outcome

A functioning prototype showing traceability, real-time alerts, and a privacy-conscious way to share insights, with clear demonstrations of detected anomalies and improved accountability in the supply chain.