An optimal attention PLSTM-based classification model to enhance the performance of IoMT attack detection in healthcare application

Abstract

The Internet of Medical Things (IoMT) has revolutionized the healthcare industry by allowing remote monitoring of patients suffering from chronic diseases. However, security concerns arise due to the potential life-threatening damage that can be caused by attacks on IoMT devices. To enhance the security of IoMT devices, researchers propose the use of novel artificial intelligence-based intrusion detection techniques. This article presents a hybrid alex net model and an orthogonal opposition-based learning Yin-Yang-pair optimization (OOYO) optimized attention-based Peephole long short term memory (PLSTM) model to distinguish between malicious and normal network traffic in the IoMT environment. To improve the scalability of the model in handling the random and dynamic behavior of malicious attacks, the hyper parameters of the PLSTM framework are optimized using the OOYO algorithm. The proposed model is evaluated on different IoT benchmark datasets such as N-BaIoT and IoT healthcare security. Experimental results demonstrate that the proposed model provides a classification accuracy of 99% and 98% on the healthcare security and N-BaIoT datasets, respectively. Moreover, the proposed model exhibits high generalization ability for multi-class classifications and is effective in reducing the false discovery rate. Overall, the proposed model achieves high accuracy, scalability, and generalization ability in identifying malicious traffic, which can help improve the security solution of IoMT devices.