LLM-Based Binary Similarity Research
Investigating whether large language models can evaluate similarity between binaries using HLIL generated from Binary Ninja.
AI Security · Cybersecurity Engineering · Resilient Infrastructure
Roman Beames
Computer scientist focused on securing intelligent systems, understanding complex software behavior, and building resilient infrastructure for emerging technologies.
About
Building at the intersection of AI and cybersecurity.
I recently graduated summa cum laude with a B.S. in Computer Science from the University of Oklahoma. This fall, I will begin an M.S. in Cybersecurity Engineering at Auburn University. War Eagle!
My work explores how intelligent systems can be evaluated, secured, and safely deployed within complex technical environments, with interests spanning machine learning security, autonomous systems, and resilient cyber infrastructure, especially as these systems become increasingly embedded in everyday life.
- Graduate focus
- AI security, reverse engineering, and the reliability of increasingly autonomous systems.
- Questions I chase
- How do AI systems fail, how can we evaluate them, and what makes them trustworthy in practice?
- What I build toward
- Secure, understandable tools for inspecting and integrating AI systems in everyday life.
Selected Work
Research and engineering projects
CVAT Autonomous Driving Annotation Pipeline
Built a Docker-based CVAT annotation pipeline with model-assisted labeling workflows for autonomous driving datasets.
Secure Peer-to-Peer Messaging Tool
Developed a secure messaging prototype that paired encrypted socket transport with peer authentication, key exchange, and a GUI for comparing plaintext and ciphertext behavior.
Beyond the Lab
Leadership, service, and creative work
The same habits that matter in technical work show up in the places where I learned to lead: preparation, trust, communication, and care for the people beside me.