Graph Neural Networks in Tasks of Extended Social Network Group Profiles Authentication

Currently, the business processes of many organizations are associated with the need to interact with group profiles of social networks and process the data they provide. Based on information from social networks, marketing, sociological research and various types of information analysis related to products, services, and events can be carried out. In this regard, methods of extended authentication, used during a user session and allowing to determine the authenticity of the user based on behavioral data, become relevant. In this work, an experimental analysis of data collected from various group profiles of a social network was carried out, during which a homogeneous graph was constructed using text data, each node of which was provided with a feature description, and the effectiveness of the DOMINANT (Deep Anomaly Detection on Attributed Networks) method using graph neural network for the problem of extended authentication.