Green federated learning empowered drug manufacturing mechanism for the pharmaceutical industry

The Internet of Medical Things (IoMT) emerged as a result of the close connection between the IoT which is the Internet of Things and the medical field. In the pharmaceutical industries, drug production is carried out by deploying IoMT by assessing the data gathered through smart devices by utilizing AI-powered systems. However, the inherent design weaknesses of conventional AI technology could result in the leakage of drugs' private information. A privacy-preserved global model can be produced through federated learning(FL). Despite this, FL continues to be susceptible to inference attacks, and energy consumption is a further concern. For this constraint, we could use green federated learning a novel and crucial research area where carbon footprint is an evaluation criterion for AI, alongside accuracy, convergence, speed, and other necessary metrics. In this paper, to address the above-mentioned consequences, an energy-conserved and privacy-enhanced technique incorporating Green FL which involves optimizing FL features by Walrus optimization algorithm(WaOA) and making design choices to minimize the carbon emissions consistent with competitive performance for IoMT is proposed. The proposed work shows improved performance with 91% global model accuracy, reduced carbon emissions, and better privacy in drug manufacturing. Furthermore, participants received rewards based on data quality, similarity, and richness, as validated through simulation trials. The findings indicate a convergence accuracy of up to 90% for local models and an increase in participant incentives proportional to data quality. These results confirm the effectiveness of the approach in balancing privacy, accuracy, and energy efficiency in the drug manufacturing Industry.