Benchmarking Centrality Heuristics for Misinformation Containment: A Structure-Dependent Cross-Network Study — Code and Data

Simulation and analysis code, together with the complete raw result files, for a systematic benchmark of eight shield-selection (node-immunisation) algorithms for misinformation containment. Misinformation spread is modelled with a standard Independent Cascade model augmented with node immunisation, in which shielded nodes act as permanent, non-propagating firewalls. Performance is measured by the Containment Rate, CR = 1 − |M_final|/n. The eight benchmarked algorithms are Monte-Carlo Greedy, Degree Shield, PageRank Shield, Betweenness Shield, Dangling Shield, GCCDC Shield, Community Shield and Random Shield, each evaluated against a no-shield baseline across nine graph sizes (n = 200 to 2,000), four intervention timings (t = 0 to 3), a propagation-probability sweep (p = 0.05 to 0.20), a shield-budget sweep (k = 5 to 50), and two misinformation-source models. Generalization is assessed on four structurally diverse networks: the Facebook Ego Network, the Twitter Ego Network, the Twitch Gamers social network, and the arXiv CA-GrQc collaboration network. The repository contains seven self-contained experiment scripts, 56 raw result files covering every reported configuration, and a script that regenerates all quantitative tables and the data-driven figure directly from those files, so that every value reported in the paper can be independently reproduced without re-running the simulations.