Federated Learning Framework for Privacy Preserving Agricultural Disease Intelligence Across Distributed Farms
Journal of Smart and Sustainable Farming · 25.05.2026 · 092
DOI: 10.64026/jssf/2026009 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Smart and Sustainable Farming, online erschienen am 25.05.2026. Verfasst von Lijuan Xieye. Die Arbeit nennt 18 Literaturangaben.
Abstract (Originalsprache)
Despite the growing importance of agricultural disease detection in the context of deep learning, there are data privacy issues, communication cost and difficulty in adapting to the different farming settings when the training is done in a centralized manner. To address this, this paper presents Privacy-Preserving Adaptive Federated Agricultural Disease Intelligence (PAF-ADI), a federated learning framework that enables each distributed farm to train the classifiers of agricultural diseases collaboratively without sharing raw images of crops and sensitive information of the farms. It is built to incorporate adaptive client weighting for data distribution and update quality heterogeneity, to include norm-based gradient clipping, to support secure aggregation, differential privacy and to implement communication-efficient parameter exchange. The goal of these mechanisms is to enhance collaborative model learning without exposing too much information and without requiring too much communication. The framework is tested on representative disease categories, with tomato, potato, rice and maize categories, including the metrics used for classification, the confusion matrix analysis, the convergence behavior, the disease localization comparison and performance across distributed farm clients. The reported evaluation achieves 95.10% accuracy, 94.80% precision, 94.60% recall, and a 94.70% F1-score. The overhead is reduced from 18.40 MB / round of FedAvg to 12.70 MB / round in this case, 30.98 saved. Results indicate that in a decentralized agriculture the federated optimization approach works as an adaptive, privacy-preserving method for deriving accurate and resource-friendly agricultural disease intelligence while keeping data ownership at the local level.
Zitieren
APA
Xieye, L. (2026). Federated Learning Framework for Privacy Preserving Agricultural Disease Intelligence Across Distributed Farms. Journal of Smart and Sustainable Farming, 092. https://doi.org/10.64026/jssf/2026009
Vancouver
Xieye L. Federated Learning Framework for Privacy Preserving Agricultural Disease Intelligence Across Distributed Farms. Journal of Smart and Sustainable Farming. 2026:092. doi:10.64026/jssf/2026009
BibTeX
@article{Xieye2026,
author = {Xieye, Lijuan},
title = {Federated Learning Framework for Privacy Preserving Agricultural Disease Intelligence Across Distributed Farms},
journal = {Journal of Smart and Sustainable Farming},
year = {2026},
pages = {092},
doi = {10.64026/jssf/2026009}
}
EUVIMED