Local Centrality Minimization with Quality Guarantees

Centrality measures, quantifying the importance of vertices or edges, play a fundamental role in network analysis. To date, triggered by some positive approximability results, a large body of work has been devoted to studying centrality maximization, where the goal is to maximize the centrality score of a target vertex by manipulating the structure of a given network. On the other hand, due to the lack of such results, only very little attention has been paid to centrality minimization, despite its practical usefulness. In this study, we introduce a novel optimization model for local centrality minimization, where the manipulation is allowed only around the target vertex. We prove the NP-hardness of our model and that the most intuitive greedy algorithm has a quite limited performance in terms of approximation ratio. Then we design two effective approximation algorithms: The first algorithm is a highly-scalable algorithm that has an approximation ratio unachievable by the greedy algorithm, while the second algorithm is a bicriteria approximation algorithm that solves a continuous relaxation based on the Lov\'asz extension, using a projected subgradient method. To the best of our knowledge, ours are the first polynomial-time algorithms with provable approximation guarantees for centrality minimization. Experiments using a variety of real-world networks demonstrate the effectiveness of our proposed algorithms: Our first algorithm is applicable to million-scale graphs and obtains much better solutions than those of scalable baselines, while our second algorithm is rather strong against adversarial instances.