Analysis and monitoring of user behaviors in social media and online social networks are of key concern for information security, public safety, and situation awareness. The authors’ goal is to model user activities, which are typically very dynamic. They propose a generalized hypergraph (GH) model to model user activity graphs.

The authors identify key abstractions for modeling user activities in a network: the level of terminal events; the level of actors (identifiers and characteristics); the level of real-life users; the level of services for evaluating activities; the collection of branded services; and the level of social events and processes. The authors then propose the GH model. In this model, the actor corresponds to a special edge in a hypergraph, rather than to a node. The generalized hypergraph allows for effective representation of the relationship between nodes on different hypergraph edges. The proposed GH model can accommodate relations of several functional types within an edge of a generalized hypergraph. It is also in essence a hierarchical hypergraph, in which nodes and edges can belong to different hierarchical groups with different functional modes.

After proposing the GH model, different types of relationships (edges) between actors and their activities are discussed: undirected or directed relationships (or both) between nodes; a multitype relationship between nodes; non-oriented or oriented edges of a hypergraph; complex relationships between nodes within edges of a hypergraph; node-to-node graph relations within different hypergraph edges; relations between hypergraph edges; and relations between a node and a hypergraph edge.

Overall, the paper provides interesting information about the use of a hypergraph-based approach for modeling dynamic user activities in a complex network for cyber-security and monitoring.