An Optimal Active Defensive Security Framework for the Container-Based Cloud with Deep Reinforcement Learning

Due to the complexity of attack scenarios in container-based cloud environment and continuous changes state microservices, effectiveness active defense strategies decreases with microservice change. To tackle it, main focus is how establish a comprehensive threat model adaptive deployment strategy. In this study, we present an optimal defensive security framework (OADSF) for deep reinforcement learning. Firstly, based on characteristics container clouds threats paths attackers are analyzed from application layer layer. Then, propose Holistic System Attack Graph quantitatively analyze gain, quality service (QOS) efficiency scenarios. Finally, optimization moving target (MTD) strategy modeled as Markov decision process. Deep learning proposed handle space explosion under large-scale applications, thus solving configuration orchestration platform. We use Kubernetes build clusters. The algorithm implemented Python 3.7 Tensorflow 1.14. Simulation results show that method can quickly converge applications increase efficiency. Compared DSEOM SmartSCR, increased by 35.19% 12.09%, respectively.