Analysis of attack-defense game for advanced malware propagation control in cloud

In modern society, cloud computing has emerged as an indispensable infrastructure. Nevertheless, as the cloud ecosystem grows increasingly vast and complex, a series of novel security challenges have surfaced, among which artificial intelligence (AI)-empowered advanced malware has provided network attackers with even more stealthy and potent weapons. While existing malware detection technologies can still maintain a certain level of defense against traditional security threats, the instant detection and response to these sophisticated AI-crafted threats become exceedingly difficult, consuming substantial remediation time and security resources. To address the balance between control costs and effectiveness, recognizing the intricately intertwined and dynamically interactive nature of the offensive and defensive parties, this paper introduces the framework of differential game theory, delving into the strategies for controlling the propagation of advanced malware in cloud environments. Firstly, we construct an advanced malware propagation control model targeting each virtual machine. On this basis, we define the specific categories of strategy selection for both the offensive and defensive sides, as well as their respective cost-benefit relationships, and formulate an attack-defense game problem. Subsequently, we rigorously demonstrate, from a mathematical theoretical perspective, that the optimal solution (i.e., Nash equilibrium) to the attack-defense game problem is indeed attainable, and we devise a dedicated accelerated algorithm for its solution. Finally, we conduct comparative experiments on three real-world datasets using three distinct strategies, and the analysis results show the effectiveness of our proposed method.