Deep learning-empowered intrusion detection framework for the Internet of Medical Things environment

The fusion of Internet of Things (IoT) technology into healthcare, known as the Internet of Medical Things (IoMT), has significantly enhanced medical treatment and operational efficiency. Real-time patient monitoring (RPM) and remote diagnostics enabled by IoMT allow doctors to treat more patients effectively and save lives. However, healthcare devices' interconnected nature makes them vulnerable to cyber-attacks, threatening patient privacy and security. Ensuring the security and accuracy of patient health data is paramount, as any tampering could have life-threatening consequences, especially in emergency situations. To address these challenges, this research focuses on developing robust security models to secure patient data in IoMT networks while meeting the growing demand for efficient healthcare services. Artificial intelligence (AI)-based technologies such as machine learning (ML) and deep learning (DL) have the potential to be employed as the methodology for intrusion detection. The goal of this research is threefold: firstly, the linear support vector machine (LinSVM) model; secondly, the convolutional support vector machine (ConvSVM) model; and finally, the categorical embedding (CatEmb) model, which have been proposed to overcome the issue of security in a network. This article offers the CatEmb model as the first effort to use a DL-based embedding approach to recognize intrusion in the IoMT environment, utilizing patient biometric and network traffic flow data. Our experimental results show the efficacy of the proposed DL models, with the LinSVM achieving a training accuracy of 99.78%, ConvSVM reaching 99.98%, and CatEmb achieving 99.84%. These models outperform existing methodologies by 2.61% in detecting network intrusions, as demonstrated through metrics such as detection rate and F1-score. Furthermore, the proposed approaches are thoroughly compared with the existing state-of-the-art studies.