Integrating Machine and Deep Learning for Enhanced Security in Cyber-Physical Systems: Challenges and Future Research Agenda

Cyber-physical systems are an essential and frequently used infrastructure for resolving today’s most challenging problems. As we all know, making rapid and accurate decisions in a big data setting (also called a critical/significant environment) is a tough challenge. This chapter’s potential investigates the specifics of one of the most significant technological revolutions developing “Cyber-Physical Systems” and the increasing role of artificial intelligence techniques in these systems. Artificial intelligence techniques are essential in information security because they can rapidly evaluate millions of incidents and recognize various threats – from malware leveraging zero-day vulnerabilities to risky actions that might result in a phishing attack or download of malicious code. We conducted an efficient search in Web of Science, PubMed, Scopus, and EBSCO for articles published up to April 2021 that addressed federated learning, deep learning, machine learning, graph-based approaches, intrusion detection tree approaches, and Signature Based Malicious Behavior Detection in cyber-security. Additionally, the work compared various attack detection techniques in cyber-physical systems against associated challenges and quality metrics such as accuracy, bandwidth, Variance of Noise, Sparsity rate, Pushing recall, F1-score, precision, and recall. It extensively discusses the context for artificial intelligence in cyber-security, including the different cyber-physical security attacks. The unique aspect of this work is that the survey summarizes current concepts and their limitations, focusing on future research potential in artificial intelligence techniques for Cyber-Physical Systems. This research work would facilitate multiple researchers and scholars investigating cyber-physical domains and serve as a basis for further studies.