Methodology for Analyzing the Effectiveness of GAN Model Architecture for Classifying Encrypted Traffic

The object of the study is the architecture of generative adversarial networks (GANs) for traffic classification. The subject of the study is the methodology for analyzing the classification accuracy of GAN models. The article develops and tests a methodology for evaluating the effectiveness of GAN architecture, as well as methods for optimizing such models. Since the GAN discriminator is trained not only on real data, but also on synthetic data, this allows it to “predict” future changes in the analyzed data. Therefore, the results of the study can be applied primarily in the development of Deep Packet Inspection (DPI) and Intrusion Detection System and Intrusion Prevention System (IDS/IPS) modules for analyzing network protocols and services, as well as in other areas where input data can often change its parameters.