Anomaly detection in cyber-physical systems using graph neural networks

This paper presents a security study of networks with dynamic topology. As a solution to the problem of attack detection, an approach to attack detection in networks with dynamic topology based on adaptive neuro-fuzzy inference system was developed. A software layout of the system that implements the proposed approach has been developed and its effectiveness has been evaluated using various metrics. Experimental results confirmed the validity and effectiveness of the developed approach for attack detection in networks with dynamic topology