A Federated Learning Framework for the Secure and Collaborative Optimization of Multi-Layered ZnO Thin Films

Authors

  • Ms G. Subbulakshmi Author
  • Dr M. Balachandramohan Author
  • Dr Janarthanam S Author
  • Dr.Mamta Sanjay Koban Author
  • Dr. Kishore Kumar Kamarajugadda Author
  • Dr Movva Pavani Author
  • Elangovan Muniyandy Author
  • Koppuravuri Gurnadha Gupta Author

Keywords:

federated learning, differential privacy, ZnO thin films, distributed optimization, privacy preserving machine learning, materials science

Abstract

Distributed manufacturing networks face fundamental challenges in collaborative optimization: data privacy constraints, communication bandwidth limitations, and heterogeneous local data distributions. This paper introduces a novel federated learning framework specifically adapted for multi-layered zinc oxide (ZnO) thin film fabrication across four geographically distributed nodes. The proposed system implements secure model aggregation via additive secret sharing and enforces formal privacy guarantees through ε-differential privacy mechanisms. We present two comprehensive benchmark datasets: Dataset 1 containing 2,400 simulation-based fabrication parameter sets, and Dataset 2 comprising 1,800 real experimental samples collected over a 12-month period from four research institutions. Experimental evaluation demonstrates 36.2% average prediction error reduction compared to isolated node training, while maintaining strong privacy guarantees (ε = 2.0, δ = 10−6 ) with only 5.8% accuracy degradation. Communication efficiency improves by 99.0%, reducing total data transmission from 345.6 MB to 3.456 MB across 50 communication rounds. Cross-dataset generalization analysis reveals 94.3% performance transfer from simulation to experimental domains. This work bridges the gap between federated learning theory and practical materials science applications, providing both methodology and reproducible benchmark datasets for the research community.

Downloads

Download data is not yet available.

Downloads

Published

2026-09-22

How to Cite

A Federated Learning Framework for the Secure and Collaborative Optimization of Multi-Layered ZnO Thin Films. (2026). Minimax Theory and Its Applications, 11(2s), 650-658. https://journalmta.com/index.php/jmta/article/view/444