PMCDM: Privacy-preserving multiresolution community detection in multiplex networks

Abstract

Multiplex networks can effectively represent network data with communications generated by sensors across multiple platforms in Internet of Things (IoTs). Multiresolutional community detection (MCD) in multiplex networks is essential to understanding the functional modules of complex systems in IoTs, consequently the studies on MCD have attracted much attention in recent years. However, most existing works require full access to the communications of systems in IoTs, which are sensitive and private. To overcome this requirement, in this paper, we propose a privacy-preserving MCD framework (called PMCDM) for multiplex networks, aiming to detect the community structures at multiple resolution levels while preserving the link and weight privacy of the systems in IoTs. PMCDM first encrypts the sensitive link and weight information of multiplex networks using the differential privacy and homomorphic encryption techniques, respectively, and subsequently devises a multi-greedy algorithm (called DH-Louvain) for MCD on the encrypted networks by optimizing a weighted modularity density. Extensive experiments on both the eleven LFR benchmark and the ten real-world multiple networks show that PMCDM can effectively detect the multiresolutional community structures of these networks while preserving the network privacy.