TRUST AND RELIABILITY-BASED FRAMEWORKS FOR SECURE VOLUNTEER CLOUD COMPUTING

Author

yousef alsenani, Southern Illinois University CarbondaleFollow

Date of Award

6-1-2021

Degree Name

Doctor of Philosophy

Department

Computer Science

First Advisor

AHMED, khaled

Second Advisor

Crosby, Garth

Abstract

With increasing popularity and the need for low-cost green computing systems, new paradigms and models such as volunteer cloud computing (VCC) have recently emerged. Volunteer cloud computing shares the same philosophy as a desktop grid, which uses underutilized or spare resources, called volunteer hosts, on personal computers owned by individuals and organizations. The VCC generally targets globally distributed, highly heterogeneous, and non-dedicated machines. The underlying volunteer cloud infrastructure makes resource management and task scheduling a challenging process since it comprised of randomly joined/leave volunteer hosts having varied levels of availability, volatility, and trustiness. The volunteer cloud includes several unique characteristics that are differ from traditional cloud and desktop grid models, including: (a) high resource heterogeneity, (b) resource availability and volatility, (c) security and trust relationship, and (d) diversity of workloads. In addition, the critical need to guarantee the quality of service (QoS) for applications deployed in the volunteer cloud computing requires tracking and monitoring of the reliability and trust of highly non-dedcated volunteer resources.In this dissertation, we proposed two approaches to address reliability and trust challenges in volunteer cloud computing. Current reliability approaches are inadequate in overloaded systems as these approaches do not consider resource utilization and job behaviors or characteristics. As a solution, we propose reputation and resource-based model, called ReMot, that estimates the reliability of a volunteer host. ReMot consists of three stages which include the following: (1) extracting resource usage patterns from historical data of tasks and host machines; (2) predicting failure rates of virtual machines and resource utilization; and (3) applying the reliability estimator (RBE) algorithm to estimate the reliability of a volunteer host. To validate ReMotapproach, the we have utilized a large usage trace of real world applications made available by Google Inc. The results indicate that ReMot obtained more accurate reliability estimation than existing models and dynamically adapts to workload variations.The second research problem proposed ProTrust, a probabilistic framework that defines the trust of a host in VCC. In volunteer cloud computing, volunteers do not disclose resource information before joining the system. This leads to uncertainties about the level of trust and security in the system. The majority of available trust models are suitable for peer-to-peer (P2P) systems, which rely on direct and indirect interaction, and might cause memory consumption overhead concerns in large systems. We expand the concept of trust in volunteer cloud computing and develop two new metrics: (1) trustworthiness based on the priority of a task, named \textit{loyalty}, and (2) trustworthiness affected by behavioral change. We first utilized a modified $Beta$ distribution function, and the behavior of resources are classified into different loyalty levels. Then, we present a behavior detection method to reflect recent changes in behavior. We evaluated ProTrust experimentally with a real workload trace and observed that the framework's estimation of the trust score improved by approximately 15\% and its memory consumption decreased by more than 65\% compared to existing methods.