Multiscale Evolutionary Perturbation Attack on Community Detection

Community detection, aiming to group nodes based on their connections, plays an important role in network analysis since communities, treated as meta-nodes, allow us create a large-scale map of simplify its analysis. However, for privacy reasons, we may want prevent communities from being discovered certain cases, leading the topics community deception. In this article, formalize detection attack problem three scales, including global (macroscale), target (mesoscale), and node (microscale). We treat optimization further propose novel evolutionary perturbation (EPA) method, where generate adversarial networks realize attack. Numerical experiments validate that our EPA can successfully algorithms all i.e., hide or disturb structure whole by only changing small fraction links. By comparison, behaves better than number baseline methods six synthetic real-world networks. More interestingly, although is Louvain algorithm, it also effective attacking other algorithms, validating good transferability.