Clustering Approaches
Advances in information security, privacy, and ethics book series · 2018
Clustering-based approaches, also known as generalization approaches, are popular in anonymizing relational data. In social networks, these generalization approaches are applied to vertices and edges of the graph. The vertices and edges are grouped into partitions called super-vertices and super-edges, respectively. In most cases, these vertices and edges are divided according to some predefined loss function. One major drawback of this approach is that the graph shrinks considerably after anonymization, which makes it undesirable for analyzing local structures. However, the details about individuals are properly hidden and the generalized graph can still be used to study macro-properties of the original graph.