Multiagent Reinforcement Learning for Efficient Cyber Attack Detection on the Internet of Medical Things using tSNE- ZOA based Dimensionality Reduction

Technological developments in information and communication (ICT) have completely changed a paradigm in computing. The creation of several new communication channels, including the Internet of Things (IoT), is a consequence of these advancements. Sensitive data can be transmitted between medical devices using the Internet of Medical Things (IoMT), it serves as a component of the Internet of Things. However potential dangers such as internet hijackings, impersonation, lack of service violence, password inference, and man-in-the-middle make security a problem, IoT technology for healthcare improves patient contact and treatment. Safeguarding IoMT ecosystem against malware attacks is crucial for patient safety and maintaining high touch and care. This paper presents a method to detect cyber-attacks in the medical field using Multiagent Reinforcement learning. Initially, data for processing is collected from the NSL-KDD and preprocessed using Improved Generative Adversarial Imputation Network (I-GAIN) for missing value replacement. Ordered Quantile normalization (ORQN) is used to converts an arbitrary vector into a normal (Gaussian) vector. Then, t-distributed stochastic neighbor embedding (t-SNE) is then used to reduce the dimensionality of the dataset. Number of components in t-SNE is optimally selected using the Zebra Optimization Algorithm (ZOA). Multiagent Reinforcement learning (MARL) is a technique utilized to detect data attacks based on the simulated environment. According to simulated research, the proposed approach achieves 95.3% accuracy, 4.7% error, and 93.5% precision. The developed technique outperforms existing methods in performance, resulting in a better prediction to detect cyber-attack in medical data.