Malware Patching Strategies in Edge Intelligence IoT Systems: A Differential Games Approach With Spiking Neural Networks

Malware poses a significant threat to Internet of Things (IoT) systems, leading to unauthorized access and data breaches. Evolving malware continues to pose severe threats to IoT systems, necessitating defense mechanisms capable of adaptive learning and real-time strategic decision-making. To this end, we present a novel IoT malware propagation and patching model based on differential games and spiking neural networks in edge intelligence (EI)-enabled IoT systems. Specifically, built upon optimal control theory, we reveal the process of dynamic evolution between infected IoT devices and associated edge nodes based on differential games under malware propagation in EI-enabled IoT systems. The dynamic state transitions are described by differential equations, followed by introducing attack/defense intention for theoretically attaining the optimal IoT malware patching strategies. Moreover, we design a novel IoT malware propagation-patching approach named differential games-based deep spiking Q-network (DGDSQ) for practical patch optimization decisions in EI-enabled IoT systems. Additionally, we conduct experimental simulations to comprehend that DGDSQ-assisted edge nodes are more effective than double deep Q-network and dueling double deep Q-network against IoT malware propagation.