Files in this item



application/pdfESTRADA-DISSERTATION-2016.pdf (2MB)
(no description provided)PDF


Title:Dynamic reliability and security monitoring: a virtual machine approach
Author(s):Estrada, Zachary
Director of Research:Iyer, Ravishankar K.; Kalbarczyk, Zbigniew T.
Doctoral Committee Chair(s):Iyer, Ravishankar K.
Doctoral Committee Member(s):Bailey, Michael D; Campbell, Roy H.; Hwu, Wen-Mei W.
Department / Program:Electrical & Computer Eng
Discipline:Electrical & Computer Engr
Degree Granting Institution:University of Illinois at Urbana-Champaign
hardware-asssisted virtualization
dynamic analysis
Abstract:While one always works to prevent attacks and failures, they are inevitable and situational awareness is key to taking appropriate action. Monitoring plays an integral role in ensuring reliability and security of computing systems. Infrastructure as a Service (IaaS) clouds significantly lower the barrier for obtaining scalable computing resources and allow users to focus on what is important to them. Can a similar service be offered to provide on-demand reliability and security monitoring? Cloud computing systems are typically built using virtual machines (VMs). VM monitoring takes advantage of this and uses the hypervisor that runs VMs for robust reliability and security monitoring. The hypervisor provides an environment that is isolated from failures and attacks inside customers’ VMs. Furthermore, as a low-level manager of computing resources, the hypervisor has full access to the infrastructure running above it. Hypervisor-based VM monitoring leverages that information to observe the VMs for failures and attacks. However, existing VM monitoring techniques fall short of “as-a-service” expectations because they require a priori VM modifications and require human interaction to obtain necessary information about the underlying guest system. The research presented in this dissertation closes those gaps by providing a flexible VM monitoring framework and automated analysis to support that framework. We have developed and tested a dynamic VM monitoring framework called Hypervisor Probes (hprobes). The hprobe framework allows us to monitor the execution of both the guest OS and applications from the hypervisor. To supplement this monitoring framework, we use dynamic analysis techniques to investigate the relationship between hardware events visible to the hyper-visor and OS constructs common across OS versions. We use the results of this analysis to parametrize the hprobe-based monitors without requiring any user input. Combining the dynamic VM monitoring framework and analysis frameworks allows us to provide on-demand hypervisor based monitors for cloud VMs.
Issue Date:2016-10-17
Rights Information:Copyright 2016 Zachary J Estrada
Date Available in IDEALS:2017-03-01
Date Deposited:2016-12

This item appears in the following Collection(s)

Item Statistics