Review of Cyber-Physical System-Based Security Datasets for Learning-Based Intrusion Detection Systems

Cyber-Physical Systems (CPSs) have been integrated into various sectors to streamline operations through automation. They are also employed in critical infrastructures, where security breaches can lead to severe consequences such as human casualties, financial losses, and damage to reputation. Traditional signature-based Intrusion Detection Systems (IDSs) have limitations in detecting and monitoring cyber breaches, especially in intricate and dynamic attack scenarios where attackers employ evasion tactics. Presently, numerous studies advocate the viability and effectiveness of learning models, including machine learning and deep learning, in fortifying the security of CPS. However, it is imperative to address the heightened computational demands, particularly for Deep Learning Models, by employing optimization techniques. To comprehensively evaluate the potential and efficacy of learning models in safeguarding CPS, it is crucial to assess their performance on the latest and benchmark CPS-based cybersecurity datasets. Thus, this paper undertakes a comprehensive review of significant cybersecurity datasets pertaining to Cyber-Physical Systems, aiming to propel further advancements in this research direction.