Using the confidence level parameter in the process of attack identification in cyber-physical systems

Korzhuk Viktoriia Mikhailovna, Groznykh Anton Vladimirovich, Zakoldaev Danil Anatolevich

This article describes the problem of attack identification in cyber-physical systems. A method for identifying of network-level attacks on wireless sensor networks based on behavioral analysis is described. A variation of the method with the confidence level parameter is proposed. This parameter allows to set the acceptable level of accuracy of attack identification and the number of features used. The results of experiments for 14 different attacks and different network topologies are presented. The values of a priori probability and the confidence level parameter are calculated algorithmically. It was found that the excess of the value of the confidence level parameter over the value of the a posteriori probability of the normal state by ~30% is sufficient; there is no need to use the a posteriori probability of the normal state exceeding 20%. Recommendations on the application of the proposed solution are formulated. The obtained results can be used as a part of intrusion detection system.

Keywords:cyber-physical systems, information security, cyber security, information security, wireless sensor networks, ZigBee, attack identification, confidentiality level