Graduate Defense: Kamryn Parker

Thesis Information

Title: Risk Assessment and Solutions for Two Domains: Election Procedures and Privacy Disclosure Prevention for Users

Program: Master of Science in Computer Science

Advisor: Dr. Hoda Mehrpouyan, Computer Science

Committee Members: Dr. Michael Ekstrand, Computer Science; Dr. Amit Jain, Computer Science; and Dr. Jaclyn Kettler, School of Public Service

Abstract

Risk is something that surrounds us each and everyday, and learning how to manage risk in different areas is necessary to limit its impact. Two different areas of risk have been identified for this thesis: election day incident infrastructure and user privacy disclosure prevention. We ask if it is possible to leverage information relating to risk to create procedures that handle it as an overall issue in order to apply procedures to similar areas of research. Understanding how to identify and prevent these potential areas of risk is important to secure information not just for a single person, but potentially for whole populations. Developing plans of action to prevent severe problems within these areas may not only improve election day incident communication but also personal trust in technology to share private information accurately.

This thesis proposes interdisciplinary techniques through qualitative analysis, quantitative analysis and machine learning tools to build solution frameworks for these unique challenges. In Idaho, communication between election officials is currently fractured when handling problems on election day. Many times, officials use multiple streams of communication (i.e. text, calls, emails, etc.) to relay information to one another. This fractured process risks inability to track progress made on incidents, potential inefficiency with resolution, and information loss between different communication streams. Thus, we survey election officials in multiple states to assess the need for a centralized incident management and communication tool to assist with incidents. Personal information security is an important part of a person or organization’s identity. If private information is leaked unintentionally to incorrect audiences, it could reflect poorly on the person or organization. Therefore, we analyze how to prevent the risk of undesirable privacy disclosure using unsupervised and transfer learning techniques to leverage public and private information for prediction.

Results pointed favorably to the implementation of a communication tool for election administrators to assist with efficient communication and response for election day problems. Learning outcomes and best practices towards preventing unwarranted private information disclosures for users will also be presented. Further analysis is advised to implement these contributions in real-world environments.