The complex dependencies and attack surface expansion of the microservice architecture in the cloud-edge collaborative network make it possible for attackers to horizontally penetrate through a single vulnerability. We propose CoQMIX, a microservice intrusion tolerant scheduling method based on Cooperative game and multi-agent reinforcement learning (MARL).The Cooperative game model based on Nash bargaining is used to coordinate the interest relationship between cloud and edge nodes, and the QMIX algorithm is used to optimize the intrusion tolerance strategy. Experiments show that CoQMIX reduces the security threat value by 50.44% compared with DSEOM, and reduces the cloud edge intrusion tolerance cost difference by 68.66% compared with SmartSCR, which can achieve the global optimal balance between resource consumption and security protection.
PEI et al. (Thu,) studied this question.