Application of Neural Networks in Data Communication Analysis

The purpose of this paper is to analyze modern trends in the field of data communication analysis with a specific focus on identifying commonly used methodologies in the detection of unusual communication patterns. This paper provides a brief overview of the techniques employed in data communication analysis. We have examined particular techniques used for identifying abnormal communication patterns within network packets, associated with malicious attacks and their features. Given the information, we analyzed the utilization of artificial intelligence techniques for identifying unusual communication patterns, using feedforward neural network topologies. We proposed two types of topologies to address this problem – multilayer perceptron and convolutional neural network. The study includes the creation of unique datasets that contain both our generated data and openly available datasets, which are subsequently analyzed. The outcomes of this research include an assessment of the classification capabilities, solution complexity and processing time of multilayer perceptron and convolutional neural network topologies used in data communication analysis, based on the specific data flow and used computational system.