Access Type

Open Access Dissertation

Date of Award

January 2018

Degree Type

Dissertation

Degree Name

Ph.D.

Department

Computer Science

First Advisor

Shiyong Lu

Abstract

In the era of big data, scientific workflows have become essential to automate scientific experiments and guarantee repeatability. As both data and workflow increase in their scale, requirements for having a data lineage management system commensurate with the complexity of the workflow also become necessary, calling for new scalable storage, query, and analytics infrastructure. This system that manages and preserves the derivation history and morphosis of data, known as provenance system, is essential for maintaining quality and trustworthiness of data products and ensuring reproducibility of scientific discoveries. With a flurry of research and increased adoption of scientific workflows in processing sensitive data, i.e., health and medication domain, securing information flow and instrumenting access privileges in the system have become a fundamental precursor to deploying large-scale scientific workflows. That has become more important now since today team of scientists around the world can collaborate on experiments using globally distributed sensitive data sources. Hence, it has become imperative to augment scientific workflow systems as well as the underlying provenance management systems with data security protocols. Provenance systems, void of data security protocol, are susceptible to vulnerability. In this dissertation research, we delineate how scientific workflows can improve therapeutic practices in autism spectrum disorders. The data-intensive computation inherent in these workflows and sensitive nature of the data, necessitate support for scalable, parallel and robust provenance queries and secured view of data. With that in perspective, we propose $OPQL^{Pig}$, a parallel, robust, reliable and scalable provenance query language and introduce the concept of access privilege inheritance in the provenance systems. We characterize desirable properties of role-based access control protocol in scientific workflows and demonstrate how the qualities are integrated into the workflow provenance systems as well. Finally, we describe how these concepts fit within the DATAVIEW workflow management system.

Download

Share

COinS