VIS: A Virtualization Introspection System for Securing KVM-based Cloud Platforms

 

Fang Yu and Sheng-Wei Lee

 

Abstract: Linux Kernel Virtual Machine (KVM) is one of the most commonly deployed hypervisor drivers in the IaaS layer of cloud computing ecosystems. The hypervisor provides a full-virtualization environment that intends to virtualize as much hardware and systems as possible, including CPUs, network interfaces and chipsets. With KVM, heterogeneous operating systems can be installed in Virtual Machines (VMs) in an homogeneous environment. However, it has been shown that various breaches due to software defects may cause damages on such a cloud ecosystem. We propose a new Virtualization Introspection System (VIS) to protect the host as well as VMs running on a KVM-based cloud structure from malicious attacks. VIS detects and intercepts attacks from VMs by collecting their static and dynamic status. By replaying the attacks on VMs under VIS, we collect VM behaviors and leverage artificial intelligence techniques to derive effective decision rules to cluster and detect malicious behaviors with unsupervised learning nature. The preliminary result shows the promise of the presented approach against several modern attacks on CVE-based vulnerabilities.