Title: Enhancing System Security: A Multi-Faceted Approach to Preventing Data Leaks

Program: Computing PhD

Committee Chair: Nasir Eisty

Committee: Nasir Eisty, Edoardo Serra, Amit Jain

Abstract: Data breaches pose significant challenges in sectors such as finance, healthcare, and government, where the safeguarding of sensitive information is critical. This dissertation focuses on investigating the mechanisms behind data leaks, particularly through covert channels and physical attacks like cold boot attacks that exploit system vulnerabilities and circumvent traditional security measures. My dissertation is structured around three key objectives: First, I explore the potential for data ex-filtration through clock modulation in x86 cpu's, demonstrating that manipulating cpu clock speeds can enable high-speed data transmission. Second, I propose a robust framework designed to protect sensitive data, such as private keys, from physical attacks, ensuring security even when an adversary gains physical access to the system. Experimental results validate the effectiveness of this framework in thwarting such attacks. Third, I develop an AI-driven classification model to map software vulnerabilities responsible for data leaks from the Common Vulnerabilities and Exposures (CVE) database to common coding errors identified in the Common Weakness Enumeration (CWE). Overall, this research seeks to enhance data security by addressing vulnerabilities at both the hardware and software levels.