Optimal Asynchronous Federated Learning for the Internet of Battlefield Things (IoBT)

Desta Haileselassie Hagos, Earl Tankard, Manish Bhurtel, Danda B. Rawat

Abstract: Federated Learning (FL) is a distributed machine learning technique that enables multiple devices or systems to collaboratively train a distributed model without sharing raw data. This is particularly useful for the Internet of Battlefield Things (IoBT) as it enhances the performance and capabilities of military devices through distributed training of models without compromising data privacy or security for involved parties. In this paper, we delve into the potential of FL to enhance military devices within the IoBT. Distinguishing our work from other works in this line of research, we introduce a tailored battlefield dataset tailored specifically for battlefield scenarios. Our approach combines both synchronous and asynchronous FL techniques, demonstrating their practicality in the context of the IoBT using this custom dataset. Our study not only explores the benefits and challenges of employing FL in battlefield scenarios but also highlights the capability of asynchronous FL to achieve comparable accuracy to synchronous FL while significantly reducing resource consumption. By addressing these critical aspects, our research advances the use of FL in military applications and contributes to the ongoing development of secure and efficient systems for the IoBT.

External IDs:dblp:conf/icmla/HagosTBR23