DL-Based DDoS Attack Detection in SDN-Assisted Heathcare 4.0 Telesurgery Networks

Abstract

The transition of healthcare to smart healthcare has brought about a paradigm shift in the provision of medicinal services. The most notable contribution by the latest iteration of smart healthcare, i.e, Healthcare 4.0, is the advancement in telesurgery. It has exceeded expectations and has found application in a vast array of medical disciplines. The integration of robotic machinery in performing procedures has greatly increased the accessibility of healthcare services. However, one of the primary causes for concern is the vulnerability of such technology to malware attacks. A wide range of such attacks are in existence, ranging from simple spam to ransomware, which can be extremely detrimental to the condition of the patient under care, leading to permanent damage or even death. In this context, we propose an approach focused on detecting distributed denial-of-service (DDoS) attacks on software distributed networks (SDN) of robotic machinery for telesurgery frameworks. We take into account different deep learning (DL) algorithms aimed at detecting whether a signal being passed through the SDN to the robot is desired or malicious. We consider a fully connected neural network (FCNN), a 1D CNN, and a TabNet. We train all three on the same dataset, and compare and contrast their performance based on parameters like accuracy, receiver operating characteristic (ROC) values, learning curves, and shapley additive explanations (SHAP) values. This represents a significant advancement in integrating latest technology with healthcare, providing a dependable solution for the cyber security of telesurgery machinery networks.