Virtual Machine, Real Threats: Demystifying Virtualization Obfuscation for Resilient Software Security
Date of Award
Spring 2025
Project Type
Dissertation
Program or Major
Computer Science
Degree Name
Doctor of Philosophy
First Advisor
Dongpeng Xu
Second Advisor
Radim Bartos
Third Advisor
Elizabeth Varki
Abstract
Software obfuscation serves a dual purpose: It protects intellectual property and thwarts malware analysis, yet it can inadvertently enable advanced exploits and obstruct vulnerability discovery. This dissertation explores the multifaceted impact of obfuscation on software security, focusing primarily on virtualization-based obfuscators—widely recognized as among the most effective yet least understood forms of code obfuscation. This dissertation begins by demonstrating how conventional obfuscation can unintentionally facilitate sophisticated code-reuse attacks, enabling attackers to assemble more complex exploits. This dissertation then provides a systematic study of virtualization obfuscators, introducing a comprehensive taxonomy of VM diversification techniques, an automated tool to identify these techniques in real-world obfuscators, and an evaluation of how enhanced knowledge of VM internals can bolster deobfuscation. Finally, this dissertation investigates the challenge of vulnerability discovery in heavily virtualization obfuscated programs, proposing a hybrid fuzzing framework that combines runtime memory mutation and bottom-up fuzzing to effectively detect deep software flaws. By consolidating these efforts, this dissertation offers a roadmap for understanding, detecting, and mitigating modern obfuscation threats, ultimately empowering both the security research community and practitioners to build more resilient software.
Recommended Citation
Zhang, Naiqian, "Virtual Machine, Real Threats: Demystifying Virtualization Obfuscation for Resilient Software Security" (2025). Doctoral Dissertations. 2946.
https://scholars.unh.edu/dissertation/2946