When it comes to resolving the vast range of problems that crop up in regular life, the cyber physical system (CPS) is by far the most popular infrastructure option. However, making the right decision quickly remains challenging in the big data era. Transforming the manufacturing sector and other applications will rely on Internet of Things (IoT) or CPS. These advantages of CPS come at a price, however, as companies struggle to deal with the massive amounts of data being produced by Internet-connected gadgets, resulting in a slew of difficulties for individuals. These infrastructures are too complex for even the most knowledgeable individuals to manage, monitor, or evaluate. As a result, there is a pressing need for the convergence of Machine Learning (ML) and cyber security in CPS, which equips experts with the tools they need to monitor the internet for potential threats in record time. Since the proposed study examines the various frameworks used for Cyber-attack detection using a learning method, it demonstrates the significance of ML and Deep Learning (DL) in a CPS for more accurately identifying potential dangers. Many academics rely on security analytics, and the tool may also be used to prioritize alarms and signals. It has been brought to the attention of the researchers that the suggested study on various assaults has also emphasized the need to be mindful of rare attacks that may become highly harmful. Additionally, the pros and disadvantages of various methodologies and datasets used in the study of various works in assessing different assaults are presented to aid in selecting the appropriate strategy according to demand.