ADAPTIVE NEUTRALIZATION OF CYBERPHYSICAL SYSTEMS STRUCTURAL BREACH BASE ON GRAPH ARTOFOCAL NEURAL NETWORKS

The paper proposed a threat model of cyber-physical systems (CPS), with examples of
attacks and consequences for systems for various purposes. It is concluded that the most critical
consequences of attacks are related to the disruption of information exchange within the system

Thus, the task of ensuring the security of the CPS is reduced to restoring the efficiency of information
exchange. To neutralize the negative consequences for information exchange, it is proposed
to use graph artificial neural networks (ANNs). A review of modern architectures of graph ANNs has
been carried out. To generate a synthetic training dataset, an algorithm was developed and implemented
that simulates the intensity of the network flow and the workload of devices in the system
based on graph centrality metrics. A graph ANN was trained for the task of reconfiguring the graph
of the CFS network