Modelling-Based Classification of Sensor Nodes in WBAN Using a Hybrid Backpropagated Mask Convolutional Neural Network

Abstract

Wireless Body Area Networks (WBANs) have enhanced cornerstone technology in modern healthcare, enabling continuous monitoring of patients through interconnected wearable sensors. However, their reliability and security remain vulnerable due to untrusted or malicious sensor nodes that can degrade data quality and delay medical responses. This study introduces an integrated intelligent framework that enhances node trust classification indoors healthcare-oriented WBAN environments. The proposed approach combines Principal Component Analysis (PCA) for dimensionality reduction with the Dragonfly optimization algorithm algorithm for optimal feature selection, followed by a Hybrid Backpropagated Mask Convolutional Neural Network (HB-MCNN) for node trust classification. The model was evaluated using a real WBAN dataset under controlled simulation settings. Results demonstrated a classification rate of 99.2%, an accuracy of 98.5%, a packet delivery rate of 98.6%, and a latency of 7.2 ms, while reducing energy consumption to 55.5%. These outcomes confirm the framework’s ability to enhance the reliability and security of medical data transmission by accurately isolating untrusted nodes, reducing communication delays, and improving overall trust management in healthcare monitoring systems.