Adversarial Attacks on Influence Maximization in Homogeneous Hypergraphs

Adversarial strategies targeting influence maximization within homogeneous hypergraphs is an essential area with applications in viral marketing, rumor control, and epidemic prevention. We define the Adversarial Influence Maximization Problem (AdvIMH) and propose a Deep Reinforcement Learning-based algorithm ADV-DRL for selecting attack nodes. Additionally, we introduce a hypergraph embedding algorithm HyperVec to obtain vector representations of hypergraphs. By modeling real-world high-order relationships in datasets such as Music and Restaurant hypergraphs, we demonstrate the effectiveness of ADV-DRL through extensive experiments. Results show that ADV-DRL significantly outperforms baseline algorithms in reducing influence spread, highlighting its potential to identify vulnerable nodes and strategically diminish information propagation in hypergraphs. This work not only broadens the scope of influence maximization applications but also offers an innovative approach for tackling adversarial scenarios in networked systems.