9004 Problems of information security. Computer systems Проблемы информационной безопасности. Компьютерные системы 2071-8217 13 Anomaly detection in cyber-physical systems using graph neural networks Обнаружение аномалий в киберфизических системах с использованием графовых нейронных сетей Vasileva Ksenia 0000-0003-2849-4682 Lavrova Daria lavrova_ds@spbstu.ru Peter the Great St. Petersburg Polytechnic University 30 03 2021 1 117 130

The paper proposes the application of convolutional graph neural networks to detect anomalies in cyber-physical systems, developed a graph model reflecting the dynamics of changes in the state of devices, presented an algorithm for data preprocessing, which provides the formation of the graph based on the studied sample of telemetry values. The optimal parameters of the neural network were established experimentally, the applicability and effectiveness of the proposed model for detecting anomalies in cyber-physical systems were shown, and the ability of the model to detect and distinguish between classes of attacks was confirmed

 graph neural networks cyber-physical system anomaly detection convolutional neural networks information security telemetric data analysis