Game-Theoretic Adaptive Routing and Integrated Security Framework for Multi-Hop LoRaWAN Networks

Abstract

Multi-hop Long Range Wide Area Network (LoRaWAN) deployments supporting mission-critical applications face serious limitations when static routing protocols are unable to adapt to changing network conditions or routing-layer security threats. Our analysis indicates that static routing protocols can experience up to 43% degradation in packet delivery ratio under selective forwarding attacks compared to normal operating conditions. Current approaches offer limited capability to autonomously respond to topology changes, device failures, or varying traffic loads. This paper introduces a game-theoretic adaptive routing framework combined with reinforcement learning to dynamically optimize path selection in multi-hop LoRaWAN networks. The proposed framework operates entirely on the network server, while end devices perform only lightweight state reporting, ensuring compatibility with Class A LoRaWAN devices. Trust information derived from security monitoring is incorporated to balance performance objectives with security-aware routing decisions. The approach is evaluated through large-scale simulations with network sizes ranging from 100 to 2000 nodes. The results show consistent improvements in packet delivery ratio and latency compared to static routing across the evaluated scenarios, while maintaining stable performance under attack conditions. Attack detection achieves high precision (92.0%) with a low false positive rate (0.65%), with an average detection time of 26.6±6.8 seconds. Integration overhead remains limited, resulting in minimal additional energy consumption.