Application of an Autonomous Object Behavior Model to Classify the Cybersecurity State

Abstract

This paper considers the issues of ensuring the cybersecurity of autonomous objects. Prerequisites that determine the application of additional independent methods for assessing the state of autonomous objects were identified. Side channels were described, which enable the monitoring of the state of individual objects. A transition graph was proposed to show the current state of the object based on data from side channels. The type of sound signals used to analyze and classify the state of information security was also shown. An experiment intended to accumulate statistical information on the various types of unmanned object maneuvers was conducted using two audio recorders. The data obtained was processed using two-layer feed-forward neural networks with sigmoid hidden neurons. The autonomous object behavior model can be used as an additional element to determine the state of cybersecurity. Using a segmented model, it was possible to improve the accuracy of determining the cybersecurity state. The proposed model enabled the identification of differences in the states of autonomous object cybersecurity with probabilities that were, on average, more than 0.8.