Protecting the Cyberphysical System Against Intrusion Dection

As CPS applications have a growing influence across a wide range of industries, CPS security has emerged as a crucial area of study. CPS security calls for a different approach than conventional IT security due to the interdependence of cyber and physical elements. As part of our comprehensive CPS security methods, we offer a federated learning-based threat detection technique in this study. First, we design a convolutional neural network with a gated recurrent unit to provide a novel deep learning-enabled intrusion detection method for industrial CPSs. To build an advanced intrusion detection system without sacrificing individual data privacy, Our second contribution is the development of a federated learning structure for use by a wide variety of industrial CPSs. To further ensure the safety and confidentiality of model parameters during training, a secure communication protocols is developed using the Paillier cryptosystem. Extensive tests on a real-world industrial CPS data set demonstrate the CNN-GRU scheme's superiority over cutting-edge methods and its high effectiveness in recognising various cyber-attacks against industrial CPSs.