IoT-enabled human-cyber-physical interaction and decision system for dynamic social network intervention: an application to drug transmission

Drug addiction spreads through concealed social interactions that form complex and adaptive transmission networks. Effectively intervening in such networks remains a pressing challenge, as practical strategies must coordinate real-time sensing with limited resources and contextual human judgment. This study presents a novel IoT-enabled human-cyber-physical interaction and decision system (IoT-HCPIDS) that bridges distributed sensing capabilities with strategic network intervention for modeling and blocking drug addiction transmission networks. The proposed architecture integrates four hierarchical layers: (1) IoT perception layer deploying detection device nodes for real-time behavioral data collection, (2) cloud-based data fusion layer constructing dynamic digital twins of evolving social networks, (3) intelligent model layer implementing mixed-integer programming (MIP) optimization for intervention strategy generation, and (4) human-in-the-loop application layer enabling role-specific decision support through adaptive interfaces. At the algorithmic core, a novel matheuristic VNS-MIP hybridizes variable neighborhood search with MIP-based local optimization, incorporating hazard-greedy construction heuristics and adaptive operators to navigate the intervention targeting problem. The framework is validated through real-world deployment in Guiyang City, China, with IoT-enabled monitoring of 2416 registered individuals. Experimental results show that adaptive dual-population segmentation improves total harmfulness reduction by 3.9% over conventional targeting and achieves full allocation of the pre-defined intervention budget. In addition, VNS-MIP consistently matches or outperforms exact solver performance within 48 seconds across all tested configurations. While demonstrated for drug control, the methodology generalizes to network intervention domains such as epidemic containment and misinformation control, where disrupting adaptive social structures under resource constraints similarly demands integrated computational and human expertise.