DETECTION AND IDENTIFICATION OF CYBERATTACKS IN CPS BY ‎APPLYING MACHINE LEARNING ALGORITHMS

Abstract

In general, cyber-physical systems (also known as CPS) consist of networked components that allow for remote access, monitoring, and examination. Because they were integrated into an unsecured network, they have been the target of multiple cyberattacks. In the event that there was a breach in internet security, an adversary would be able to damage the system, which may have devastating effects. Thus, it is extremely important to maintain the credibility of the CPS. It is becoming increasingly difficult to identify assaults on computerised policing systems (CPSs) as these systems become more of a target for hackers and cyberthreats. It is feasible that Machine Learning (ML) as well as Artificial Intelligence (AI), may also make it the finest of times. Both of these outcomes are plausible. Technology based on artificial intelligence (AI) can play a role in the growth and success of a wide range of different types of enterprises in a variety of different ways. The goal of this type of data analysis is to avoid CPS assaults using machine learning and artificial intelligence techniques.   A new framework was offered for the detection of cyberattacks, which makes use of machine learning and artificial intelligence (ML). the process of cleaning up the data in the CPS database is starting by performing normalisation in order to get rid of errors and duplicates. This is done so that the data is consistent throughout. Linear Discriminant Analysis is the method that is used to get the features, and it is known as that (LDA). As a mechanism for the identification of cyberattacks, The suggested used process was the SFL-HMM process in conjunction with the HMS-ACO procedure. The new strategy is evaluated using a MATLAB simulation, and the metrics obtained from that simulation are compared to the metrics received from the earlier methods. The framework is shown to be substantially more effective than traditional techniques in the upkeep of high degrees of privacy, as demonstrated by the outcomes of a number of separate investigations. In addition, in terms of detection rate, false positive rate, and computation time, respectively, the framework beats traditional detection methods.