An approach to the identification of the state of elements in cyber-physical systems based on principal component analysis

Abstract

The close integration of modern cyber-physical systems with production and technological ones as well as with critical information infrastructure requires improving the monitoring process. The monitoring process is necessary with a constant increase in the possible points of entry into the system. The processing of a large amount of data coming from monitoring systems requires significant computing power. In this regard, it is relevant to reduce the dimension of the feature space while maintaining an acceptable monitoring accuracy. The proposed solution should be invariant to the dimension and orders of magnitude from which the time series supplied to the input of the monitoring system are composed. To obtain the most informative features in the formation of their set, it is proposed to use principal component analysis, and a method based on decision trees for their classification. A computational experiment was performed to confirm the applicability of the developed approach. The data of the network traffic for the research stand of the cyber-physical system water treatment were used in the experiment. The accuracy of the set of methods on the analyzed data was 98.74 %. The comparison with known studies showed an increase in the F-measure up to 0.925, which is 4.8 % higher than the most effective method used to date, namely the Isolation Forest method. The developed approach allows one to significantly increase the speed of identification and to detect anomalies of information security and functional safety of cyber-physical systems with high accuracy by reducing the dimension of the original feature space. The proposed approach can be used in event monitoring systems that deal with information security. The presented theoretical results can be useful for researchers of information security and functional safety of cyber-physical systems.

Keywords

• information security
• functional safety
• cyber-physical systems
• identification of anomalies
• time series analysis
• principal component analysis
• monitoring systems