CoHide: Overlapping community hiding algorithm based on multi-criteria learning optimization

Overlapping community detection algorithms have attracted a lot of attention for their ability to better reflect the diversity and complex relationships in real-world social networks and complex networks. However, overlapping community algorithms are vulnerable to expose private information of users to unscrupulous elements. Furthermore, existing overlapping community-based hiding models define the entire overlapping region as a danger zone, which is inefficient and unrealistic as it cannot accurately identify critical and vulnerable users. In addition, existing hiding algorithms change the topology of the original network through their edge addition and deletion strategies, which can disrupt the network trends. Therefore, in this paper, we propose an overlapping community hiding algorithm based on multi-criteria learning decision analysis and a network growth model, called CoHide. The algorithm is divided into two components: firstly, the High-risk Seed node set Extraction (HrSE) algorithm is used to efficiently and accurately extract the set of high-risk nodes vulnerable to attacks; then the Network Growth-based Community Hiding (NGCH) algorithm is used to achieve a quicker community hiding process. In addition, we define community tendency indicators and propose OL-Permanence, a community hiding indicator applicable to overlapping networks, based on the Permanence indicator to evaluate the hiding effect. The effectiveness of the proposed CoHide algorithm is verified by experiments based on three public datasets and one real Twitter dataset.