Study of Information Flows in a Network Based on Exact Solutions
Ruslan Politanskyi, Serhii D. Haliuk, Anna Ploshchyk, Serhii Toliupa · 2023
The paper presents a mathematical model that serves as the basis for determining sets of solutions for fully connected networks without loop flows with three or four nodes. Based on these solutions, software was developed to calculate the statistical distribution of entropy values within the network. Thestatistical propertiesof informationflow entropy in networks, with or without restrictions allowing uniform flow distribution, are analyzed. This characteristic is of practical importance for analyzing the dynamic state of the network and predicting flow redistribution processes from an initial unbalanced position, which inherently leads to increasing entropy. Another practical application of these results is their use as a foundation for algorithms to detect network imbalances resulting from significant external influenceswithout causing asubstantialchangein the total network load. This may include concealed DDoS attacks that replace useful network traffic in order to reduce bandwidth. The analysis method developed in this work can be adapted for networks with different topological structures, as the maj ority of networks do not have full connectivity or are generally loosely connected networks. The importance of this method is also on the rise due to the increased computing power available through cloud computing technologies and the accumulation of statistics on information flow distribution in networks for various purposes. This way, the network analysis technique discussed here can be implemented with the assistance of machine learning and artificial intelligence technologies.