Abstract. Currently, the spread of information manipulation and cyber propaganda on social networks and messengers poses a serious threat to society. Cyber propaganda is widely used to form different points of view on social media and to manage society. This phenomenon affects political conditions, economic stability, as well as the formation of public opinion. This article discusses the methods used to monitor cyber propaganda. Hybrid models combine the results of several algorithms for a deeper understanding of the text context. This method is especially effective when working in multiple languages and with large amounts of data.
The article examines the process of defining cyber propaganda using several models. In addition, the results obtained on CNN, Random Forest and hybrid models are compared and show their effectiveness. The results showed high accuracy, but the CNN model also has high computational costs and difficulties with data processing. This study identifies the advantages and disadvantages of hybrid models and discusses their effectiveness in real-world conditions.

Keywords: cyber propaganda, hybrid model, CNN, Random Forest, social networks.