PERMDEC: community deception in weighted networks using permanence

Community detection is used to determine the network structure and node relationships. However, it raises privacy concerns when locating and disclosing the members’ personal or community information. Community deception is a method of hiding a target community from community detection algorithms. It is accomplished by minimally rewiring the edges of the community in the network. In this paper, we propose PERMDEC, a novel community deception algorithm that operates on a weighted undirected network. PERMDEC determines which edges of a given community should be modified based on the parameter permanence loss and updates the network to hide a specific community. We tested PERMDEC on five community detection algorithms on eight real datasets with varying budget values. The performance is compared to the baseline method SECRETORUM using the deception score and NMI. In general, PERMDEC outperforms the existing method of deception for weighted networks.