Many real-world systems involve group interaction that cannot be captured by pairwise graphs. Hypergraphs address this by allowing edges to connect arbitrary numbers of nodes simultaneously. This paper investigates whether degree-preserving rewiring of size-three hyperedges can improve information propagation on hypergraphs under the SICP spreading model. In this spreading model, each infected node randomly selects one hyperedge and attempts to infect its nodes. We propose two rewiring strategies based on the Neighbourhood Coefficient (NC), a measure of the expected hyperedge size connected to a given node. The first strategy groups nodes with similar NC into the same hyperedges (assortative), while the second connects nodes with dissimilar NC (disassortative). Both strategies preserve the distribution of hyperdegrees within the hypergraph. Experiments on six real-world datasets show that disassortative rewiring consistently increases mean spreading, with a Pearson correlation of approximately −0.95 between assortativity and mean infected nodes across five datasets. Assortative rewiring instead increases variation between seed nodes without improving average spread. The overall extent of the effect is small, presumably because we only rewire size-three hyperedges, since it seems that larger hyperedges, which are left unchanged, likely dominate the spreading process. These results suggest that in hypergraphs, unlike in pairwise networks, disassortativity promotes broader information propagation, and that hypergraph structure plays a fundamentally different role in spreading dynamics than pairwise degree assortativity does. ... Read more

Controlling the transmission of information, misinformation, or infection has become a critical challenge across modern digital and physical systems. While classical network science models these processes using pairwise graphs, many real-world architectures feature complex group interactions that are more accurately captured via hypergraphs. This thesis investigates the problem of network immunization via node removal to inhibit spreading under the Susceptible-Infected Contact Process (SICP) diffusion model, a scenario where infected nodes spread infection through shared hyperedges and do not recover. We evaluate 14 distinct node removal strategies grouped into three categories: purely structural metrics, spreading-process-aware metrics leveraging a simple probability model and community-aware metrics that utilize hypergraph modularity to target boundary nodes bridging different communities. Through controlled simulations across six real-world datasets, we evaluate how effectively these strategies minimize final epidemic prevalence under both one-shot and adaptive removal modes. Our findings reveal that despite the intuitive appeal of leveraging community boundaries and process-aware probabilities, traditional structural metrics, specifically adaptive node degree, remain remarkably robust and often outperform more complex strategies. ... Read more

In this paper, we introduce a way to improve Susceptible-Infected Contact Process (SICP) spreading on hypergraphs. Hypergraphs are graphs in which a single edge can connect multiple nodes; such an edge is called a hyperlink. SICP models epidemic spreading/information diffusion using infected and susceptible nodes.

The main contribution of this paper is the proposal of a novel strategy for adding hyperlinks to the hypergraph to increase information spread. This novel method, NIPHD, takes into account the approximate probability of a node being infected and the number of hyperlinks to which it is connected. This strategy vastly outperforms others in 6 real-world datasets. With it, we observed increases of up to 10% in the number of infected nodes when doubling the number of edges connecting 3 nodes. This corresponds to a 5% increase compared to random addition, although this varies per hypergraph. ... Read more

Hypergraphs provide a natural representation of systems involving higher-order interactions, where a single hyperlink may connect any number of entities simultaneously. Unlike pairwise networks, the removal of a single hyperlink can disrupt interactions among multiple entities at once, making it important to identify hyperlinks that are critical for information transport. While targeted removal strategies have been widely studied in pairwise networks, their adaptation to hypergraphs and their effectiveness in disrupting shortest-path information transport remain largely unexplored.

This paper presents a comparative evaluation of five hyperlink-removal strategies based on concepts previously proposed in the network science literature. By comparing their effects on shortest-path information transport, we assess which structural properties best indicate hyperlink importance and hypergraph vulnerability. The strategies are evaluated on six real-world datasets. To enable fair comparisons across structurally different hypergraphs, hyperlinks of different cardinalities (or orders) are analyzed separately, and strategy performance is compared after removing equivalent proportions of hyperlinks. Performance is assessed through the reduction in global efficiency, whereas the size of the largest connected component provides a complementary measure of structural fragmentation.

The results show that strategies incorporating global structural information consistently outperform local heuristics. In particular, Hyperlink Betweenness Centrality achieves the largest reduction in both global efficiency and connectivity across all provided datasets. The findings further demonstrate that hypergraph structure strongly influences vulnerability, with larger and more overlapping hyperlinks increasing robustness to targeted attacks. These results provide new insights into hyperlink importance and the factors governing resilience in higher-order networks. ... Read more