Creating a traceability system for the entire food supply ecosystem has been a challenge in recent years especially during the pandemic. Current supply chains are under particular substantial challenges and many companies have to procure products and services in the context of a supply constrained situation. Moreover, additional challenges remain in the current supply chains to monitor the quality and safety of food from source (supplier) to end consumer (fork). In this case, having a modern supply chain system that can proactively monitor the quality and safety of the food from source to end consumer is a marvel.
FT-Chain aims to create an innovative system to enable federated learning and artificial (edge) intelligence (ML/AI) for complex food supply chains supported with a trust management platform. This system will be designed based on private permissioned blockchain networks for managing data and producing analytics from the transactions throughout supply chains and for external participants like regulatory bodies.
In this project, we are using food quality monitoring sensor equipped  IoT systems,  and collecting  data within multiple steps/ supply chains stakeholders while respecting inherent business risks. FT-Chain supports supply chains by combining Federated Learning (FL) architecture and private permissioned blockchain with the smart contract to keep and trace the information of different stakeholders in a privacy preserving manner with a trust management platform.

Outcome

FT-Chain has introduced an innovative system that leverages Federated Learning and artificial intelligence on Private Permissioned Blockchain (Hyperledger Fabric) networks to enhance decision-making within Supply Chains, with a specific emphasis on the food supply chain. This comprehensive system seamlessly integrates IoT sensors for monitoring food quality, ensuring data integrity across various supply chain stakeholders, thereby mitigating business risks and establishing a robust traceability system.

Within this project, Hyperledger Fabric enables event-based asynchronous updates in federated learning, offering a flexible and powerful architecture for constructing real-time systems. This approach leads to improved performance, reduced communication overhead, and facilitates continuous learning and adaptation.

To address the significant challenge posed by the high cost of on-chain storage in blockchain, especially when dealing with numerous and large federated learning models, the project adopts the InterPlanetary File System (IPFS) as an off-chain distributed storage solution. The proposed system architecture is depicted in Fig.1

By combining the Federated Learning architecture with smart contracts, FT-Chain establishes a platform to maintain stakeholder information in a privacy-preserving manner, further bolstered by a trust management platform.

Publications

N. Dadkhah, X. Ma, K. Wolter, G. Wunder, DBNode: A Decentralized Storage System for Big Data Storage in Consortium Blockchains,accepted in IEEE 9th International Conference on Big Data Analytics, (best paper presentation award), 2024

M. Gulati et al, BETA-FL: Blockchain-Event Triggered Asynchronous Federated Learning in Supply Chains, accepted in 5th IEEE Int. Conference on Blockchain Computing and Applications, 2023