A Method for Detecting Disinformation Spreaders Based on a Graph Representation of the Social Network Structure

The paper considers the problem of disinformation dissemination and detection in social networks. Based on the analysis, it was found that the problem of disinformation detection is closely related, first of all, to the problem of detecting disinformation distributors. A method for detecting disinformation spreaders is proposed based on the representation of a social network in the form of a directed graph, with the division of users into communities. In the study, the social network is presented as a connected directed graph, where each node represents a user. Users are connected to each other by edges and are united into different communities. The level of trust within a community is much higher than between users of different communities. Thus, if misinformation spreads within such a community, the probability of infection of all members of the community will be high. We introduce a classification of nodes (neighboring, border, and main) and show the role they play in the dissemination of information. An analysis of a Ukrainian-language news dataset, including over 2000 records, is conducted to identify users who can potentially become disseminators of information (both true and false). The number of items with the label “true” is 1283, and with the label “fake” is 891. We need to predict the boundary nodes, which are most likely to become information spreaders in the user community, since they are most tangential to all other nodes in the community. In the same way, calculations need to be made to predict the main nodes, which are most likely to become information distributors within the community itself. Thus, if disinformation reaches the boundary nodes (there are 11 of them in our dataset), it will spread throughout the community. In addition, it was found that users who have fake posts on the page are most likely to spread fake information. This may indicate that this account is fake. This study introduces a method for identifying misinformation spreaders in online social networks. The findings highlight that boundary nodes serve as conduits for inter-community dissemination, while core nodes play a significant role within their respective communities. The result of research using a real-world Ukrainian news dataset demonstrates that users who predominantly share fake posts are more likely to spread disinformation, potentially indicating the use of fake accounts. These insights can inform the design of advanced detection algorithms and support efforts to safeguard the information integrity of digital communication platforms. Further studies are planned to calculate a pair of “trust-reliability” scores for each network node using the Trust in Social Media (TSM) algorithm.