Entropy-Based Attack on Community Detection in Directed Networks

Community detection algorithms, which cluster nodes into community structures, can inadvertently compromise user privacy. To counter this, community deception algorithms have been developed, strategically altering edges to influence the outcome of community detection. However, existing approaches have notable limitations, including inadequate privacy protection across various dimensions and a lack of consideration for edge directionality. In this paper, we introduce a novel attack on community detection algorithms aimed at obfuscating entire community structures. Our method utilizes the variation in structural entropy of the directed network caused by the presence of the target community structure, employing it as the basis for optimizing our attack strategy. We compared our approach against state-of-the-art methods, yielding promising results.