Community Deception Based on Genetic Algorithm Using Graph Local Structure

Many researchers have studied community detection attacks in different ways but there is no best method to fool all community detection algorithms. In this study, to concern privacy protection, we aim to generate the perturbation network using genetic algorithm based on heuristic approach from community disclosure by community detection algorithms. Clustering coefficient and randomization of initialization population are applied to ensure the diversity of the solution spaces and to remove the repetition of same chromosomes in population. This system proposes a local features-based mutation approach for generating the adversarial network effective in community deception. The experimental exhibits the optimal strategy with normalizing mutual information (NMI) and Modularity (Q) by running on difference datasets with variety community detection algorithms.