Vulnerability Detection in Cyber-Physical System Using Machine Learning

Abstract

The cyber-physical system is a specific type of IoT communication environment that deals with communication through innovative healthcare (medical) devices. The traditional medical system has been partially replaced by this application, improving healthcare through efficiency, accessibility, and personalization. The intelligent healthcare industry utilizes wireless medical sensors to gather patient health information and send it to a distant server for diagnosis or treatment. The healthcare industry must increase electronic device accuracy, reliability, and productivity. Artificial intelligence (AI) has been applied in various industries, but cybersecurity for cyber-physical systems (CPS) is still a recent topic. This work presents a method for intelligent threat recognition based on machine learning (ML) that enables run-time risk assessment for better situational awareness in CPS security monitoring. Several machine learning techniques, including Nave Bayes (65.4\%), Support Vector Machine (64.1%), Decision Tree (89.6%), Random Forest (92.5%), and Ensemble crossover (EC) XG boost classifier (99.64), were used to classify the malicious activities on real-world testbeds. The outcomes demonstrate that the Ensemble crossover XG boost enabled the best classification accuracy. When used in industrial reference applications, the model creates a safe environment where the patient is only made aware of risks when categorization optimism exceeds a specific limit, minimizing security managers' pressure and efficiently assisting their choices.