A Model for Availability and Security Risk Evaluation for Systems With VMM Rejuvenation Enabled by VM Migration Scheduling

Matheus Torquato,Marco Vieira,Paulo Maciel

doi:10.1109/access.2019.2943273

Abstract

Most companies and organizations rely nowadays on virtualized environments to host and run their applications. Some of these applications have stringent availability and security requirements. An important challenge for high availability in virtualized systems is software aging, which can lead the system to hangs or other types of failures. Software rejuvenation is applied to cope with software aging problems, whereas previous research suggests the use of Virtual Machine (VM) migration to reduce the downtime related to Virtual Machine Monitor (VMM) software rejuvenation. However, there is still a gap regarding the security implications of applying VM migration scheduling as support for VMM software rejuvenation. In this paper, we propose a security evaluation approach based on an availability model for virtualized systems with VM migration for VMM rejuvenation. The goal is to find the proper rejuvenation scheduling to reach the desired levels (or at least to avoid the undesired levels) of security risk and availability. We present three case studies comprising major security threats, namely Man-in-the-middle and Denial of Service attacks. Results provide insightful information regarding the tradeoff between availability and security risk when applying VM migration scheduling for rejuvenation purposes.

Highlights

Cloud computing security and dependability appear as essential challenges for industry and academia
The exact intercept is when the rejuvenation interval is equal to 57 hours, which means that larger rejuvenation intervals produce worse security results than the system without Virtual Machine (VM) migration scheduling
We highlight the following from this study case: 1) the policy that globally reduces the three metrics has the VM migration interval of 13.5 hours; 2) results from scenarios 5 and 7 reveal that migration policies for RiskScore of MITM (RS-MITM) and RS-Denial of Service (DoS) are incompatible, trying to reduce one metric will increase the other; and 3) an interesting approach for decision making in such complex scenarios is the definition of unacceptable service levels instead of desired service levels, which allows us to remove the worse solutions from the decision-making process

Summary

Introduction

Cloud computing security and dependability appear as essential challenges for industry and academia. Existing surveys show that Cloud Computing security is at the top of users’ concerns [1]. Likewise, assuring high levels of dependability in Cloud computing remains a significant research challenge. There is a need for developing holistic models for Cloud dependability evaluation [2]. Previous works highlighted indicators of software aging accumulation in Cloud components [3]. Software aging is a cumulative process that can lead software to hangs or other failures [4], whereas software rejuvenation is used to counteract software aging [5].

Objectives

Results

Conclusion