Hybrid deep learning-based cyberthreat detection and IoMT data authentication model in smart healthcare

Abstract

The Internet of Medical Things (IoMT)-based medical devices and sensors play a significant role in healthcare applications, enabling on-site and remote monitoring of vital parameters in patients and alerting medical personnel in critical situations. However, these networks are vulnerable to cybersecurity threats, resulting in issues such as patient safety concerns, data breaches, ransom demands, and device tampering. Detecting cyberthreats efficiently is challenging because IoMT generates large temporal data. Furthermore, cyberattacks typically involve imbalanced classification, where classes are not equally represented. The absence of data authentication can lead to severe consequences, including threats to patient privacy and financial ramifications, ultimately eroding trust in the healthcare system.

This paper proposes an improved deep learning-based model for cyberthreat detection and IoMT data authentication in smart healthcare. First, it introduces an embedded Ensemble Learning (EL) technique to select important features of IoMT, which trims unnecessary features and reduces the possibility of overfitting by classifiers. These scaled inputs are fed into the proposed One-Dimensional Convolution Long Short-Term Memory (1D-CLSTM) Neural Network to classify cyberthreats. The random undersampling boosting technique has been applied to address issues like imbalance classification. The PoAh consensus algorithm is applied in the fog layer for data authentication. The proposed model is evaluated based on various performance metrics and compared to state-of-the-art techniques such as 1D-CNN, LSTM, and GRU. Evaluation results show that the proposed 1D-CLSTM achieves 100% accuracy with the WUSTL-EHMS-2020 and 98.55% test accuracy with the ECU-IoHT datasets. The PoAh-based authentication takes 3.47 s at average 9th iteration.