Research work · 2026

Blockchain empowered federated learning for intelligent analytics in cattle livestock farming

Frontiers in Blockchain

Abstract

Introduction Dairy industry is playing a vital role in rural economies as well as global food industries. Ensuring compliance with food safety standards remains a constant challenge where manual inspections and periodic audits are not accurate because of vulnerable record maintenance systems. There is a pressing need for immutable and decentralized compliance monitoring mechanism that is suitable for small-scale farmers and a tamper proof record maintenance system for ensuring compliance. Methods This research proposes a compliance monitoring system with Federated Learning (FL), CattleML (a lightweight edge inference runtime designed for microcontrollers) and a consortium Blockchain network. CattleML implements the instantaneous examination of cattle health, hygiene, and milk quality using multimodal sensor data (temperature, accelerometer features, ammonia levels and pH), which supports the on-device edge intelligence without exposing raw data. Federated learning with Graph AT-based dynamic clustering and DBSCAN-based fusion enables context-aware personalization across heterogeneous farms. Compliance results are cryptographically validated using ZK-SNARKs and recorded on a Layer-2 Polygon zkEVM Blockchain, where smart contracts manage farm identities, violation logs, audit triggers, and regulatory interactions. Results and Discussion The framework is validated using the Shahhet2812 dataset across 16 parameters. Experimental results demonstrate a federated global model accuracy of 96.94%, a 97.7% reduction in communication payload (4.25 KB per round), and high resilience against sensor noise, maintaining >90% accuracy under 20% noise injection. Results show that the proposed system establishes a trustworthy and automated compliance monitoring for system having resource constraint IoT devices.

DOI: 10.3389/fbloc.2026.1815609