Models of the Spread of Disinformation on the Internet

Introduction. In our time, the development of digital technologies never stops. People, with each passing day, increasingly gain access to rapid information, and at the same time, they are more and more often deceived due to the lack of quality verification of content for authenticity. Fakes regularly fill the information space. The main danger lies in the fact that users can not only consume news but also become distributors themselves. Information and psychological operations (IPSOs) are a tool for influencing public opinion. Deceptive content poses a threat also because it can be supported by an emotional component, supplemented with truthful information, and involve the distortion of facts to complicate the recognition of an information attack. The spread of fakes can even lead to the destabilization of the situation in the state. That is, disinformation is also a threat to national security because it is often aimed at spreading panic and undermining the morale of society; this poses a particular threat during a war in the country, when the reliability of information is critically important. Usually, the focus in existing solutions is directed at the analysis of individual features (nature of the text, dynamics of dissemination, etc.), and they are also limited in searching for dependencies between objects, since most of them are based on data analysis only in Euclidean space. Therefore, in the fight against disinformation, we need more advanced solutions where the mentioned problems will be addressed. Objective of the paper. Development of a hybrid model for the spread of disinformation on the Internet by combining a neural network-based solution and an information dissemination model. The proposed solution should ensure high accuracy in fake detection, as well as demonstrate flexibility and resilience to changes in the environment. Results. It is proposed to use SEIRA models to simulate the spread of disinformation in social networks, previously detected using GNN based on real data from social media. Keywords: disinformation spread, modeling, hybrid model, neural networks, social networks.