Software Testing, Bug Finding, and Exploitation

My research in this area focuses on creating advanced testing tools to automate software verification and vulnerability discovery. I specialize in developing directed fuzzing techniques that can efficiently navigate complex codebases to identify exploitable bugs.

Research Outcome: 1st Place Winner – DARPA AI Cyber Challenge (AIxCC)
I was a member of Team Atlanta, the 1st place winners of the two-year, DARPA-funded AIxCC global competition, securing a $4 million prize at DEFCON 33. My primary contribution involved developing the directed fuzzing engine for C/C++ programs within the team's autonomous Cyber Reasoning System (CRS), ATLANTIS. This engine was powered by Bullseye, a part of my PhD dissertation project, which utilizes advanced program analysis to guide vulnerability discovery. Our system's performance and architecture are documented in our technical report: ATLANTIS: AI-driven Threat Localization, Analysis, and Triage Intelligence System.

Systems Security and Confidential computing

I am interested in optimizing the performance and security of Trusted Execution Environments (TEEs), particularly within Intel SGX architectures. My work focuses on ensuring secure communication and high-performance I/O across hardware trust boundaries.

Research Outcome: RAKIS, a system designed to bridge the performance gap with secure I/O primitives.
As a Research Intern at Intel, I was the lead author of the research paper "RAKIS: Secure Fast I/O Primitives Across Trust Boundaries on Intel SGX," published at EuroSys 2025 (a top-tier "A*" systems conference). I led the design and development of RAKIS, a novel framework that enables secure, high-speed I/O for Intel SGX enclaves. My work addressed the critical performance bottlenecks of Trusted Execution Environments (TEEs) by creating optimized communication primitives that bypass expensive enclave exits while maintaining rigorous security boundaries. Full Paper via ACM Digital Library.

Human-AI Privacy

My current research explores the intersection of human psychology and data privacy, specifically examining the dynamics of Human-AI Trust. I am investigating how varying levels of social trust in AI agents influence a user's willingness to disclose different categories of sensitive data.