Adaptive protocols for hypervisor security in cloud infrastructure using federated learning-based anomaly detection

A static security layer protecting the hypervisor from ever-evolving cyber attacks raises concerns about cloud computing security in the dynamic cybersecurity landscape. As cybercriminals modify their approaches, the security protocols should adapt accordingly. This paper introduces an adaptive communication protocol enhanced by federated learning (FL) to improve hypervisor security in cloud infrastructures. Federated learning is a decentralized machine learning (ML) approach that prevents data sharing while still allowing models to learn collaboratively across multiple hypervisors. Artificial Intelligence (AI)-based anomaly detection is incorporated into this framework to enhance hypervisor security in cloud infrastructures. The proposed system utilizes local and global anomaly detection models to dynamically adjust security protocols and protect hypervisors against threats such as hyperjacking, side-channel attacks, and virtual machine (VM) escape. Experimental results demonstrate the protocol's effectiveness, achieving a detection accuracy of 92.6%, significantly higher than the 85.2% from centralized learning and 78.4% from static protocols. Furthermore, the adaptive approach reduced communication overhead by 55% and training time by 32%, emphasizing its efficiency and operational performance. This research highlights the potential of integrating adaptive protocols with federated learning to enhance cloud security, offering a robust defense against evolving cyber threats.