Exploring the Effectiveness of Selection of Feature Techniques for Incursion Detection in Cyber-Physical Systems

This review aims to understand and analyze various recent feature selection techniques that are being used by researchers for detecting intrusions within cyber-physical systems (CPS) from 2023 to 2019 and evaluate their effectiveness. This review has been carried out by consulting different research papers from reputed databases such as IEEE, Google Scholar, Web of Science, Science Direct, etc. We have selected only those articles which are relevant to the theme of the study. Then we categorized different techniques for selection of feature used for incursion identification in CPS in tabular form. We have found that the embedded technique for feature selection results in better accuracy of the IDS models in comparison with other available techniques for feature selection. We present a comprehensive review and comparative analysis of popular feature selection techniques available in literature commonly employed in intrusion detection in CPS from 2023 to 2019. These techniques encompass both traditional and machine learning-based approaches, including filter, wrapper, and embedded methods. The strengths and limitations of each techniques are identified, shedding light on their applicability and achievement in the context of incursion detection in CPS. We have also found out the various research challenges for the detection of intrusions in CPS. The novelty of this work is that we have studied the latest feature selection techniques that are being used by the researchers in the realm of intrusion detection in CPS and we have found from the literature that ensemble grounded techniques result in better accuracy.