Abstract

An Intrusion Detection System (IDS) is an essential part of the network as it contributes towards securing the network against various vulnerabilities and threats. Over the past decades, there has been a comprehensive study in the field of IDS, and various approaches have been developed to design intrusion detection and classification system. With the proliferation in the usage of Deep Learning (DL) techniques and their ability to learn data extensively, we aim to design Deep Neural Network (DNN) based IDS. In this study, we aim to focus on enhancing the performance of DNN-based IDS in Cyber-Physical Systems (CPS). CPS combine physical processes, networking, and computation. The integration of CPS components could seriously jeopardise the security of CPS settings because of the physical limitations. The vulnerability of CPS to cyberattacks has grown with the development of IoT and other physical systems. As cyber-physical systems refer to the intersection of your organization’s technology and IT infrastructure and its physical assets, ensuring access and data security through advanced methods prevents any cyber-attack from damaging your physical assets and thereby disrupting your business flow. Conventional cyber and network security procedures fail to guarantee data privacy and security in CPS contexts. This research aims to provide a cutting-edge attack detection method based on learning for CPS environments. The paper suggests using MLP-based smart attack control systems to increase the CPSs' security. Performance analysis is presented in terms of different evaluation metrics such as accuracy, precision, recall, f-score, and False Positive Rate (FPR), and the results are compared with existing feature selection techniques. The effectiveness of the suggested model was confirmed by comparing the outcomes with those of other successful deep learning-based algorithms, including the Gaussian Naive Bayes algorithm, SVM, and logistic regression. Comparative results demonstrate that the suggested method outperforms existing learning models with an exceptional accuracy of 99.52%.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call