Federated Blockchain and Hybrid Deep Learning Framework FB-HDL for Secure and Transparent Agricultural Trading
Year: 2026
Type: conference-paper
Abstract: Digital agriculture is growing fast, but it still runs into problems, like data ownership questions, trust that wavers a bit, and also the whole pricing manipulation thing. To address these issues, we introduce the FB-HDL (Federated Blockchain-Hybrid Deep Learning) framework, which is the answer utilizing the three high-technology tools of blockchain, federated learning, and hybrid deep learning to facilitate trustworthy validation of transactions and secure sharing of data. Blockchain technology is decentralized, immutable, and offers transparent traceability of agricultural transactions. Thus, it has eliminated the need for intermediaries and constrained forms of manipulation, while smart contracts automatically provide pricing and payments. Federated learning allows farmers to collaborate to train models at their local edge node, without needing to share raw data and retain privacy and less dependency on a central server. Our hybrid deep learning model enhances the overall integrity of transactions, as system intelligence is enhanced through the forecasting of commodity prices and tagging transaction links as valid or broken. The system will first preprocess, extract features and validate links using actual agricultural price datasets, prior to the record being written to the blockchain. The results of the experiments were positive, reporting an accuracy of transactions of 98% with a transaction throughput of 20 per second and a confirmation latency of 5 seconds alongside accurate price trends. Running a confusion matrix showed link classification with lower false transaction rates. FB-HDL is a promising decentralized alternative for digital agriculture, as it has greater transparency, scalability, security, and privacy compared to traditional blockchains.
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7203704147