Kevin Landry

I am a U.S. Coast Guard veteran and an M.S. candidate in Digital Forensics at George Mason University, concentrating in Penetration Testing and Reverse Engineering. My undergraduate work is in computer science and cybersecurity from the University of Mary Washington.

My work sits where offensive validation, defensive telemetry, cloud infrastructure, and digital evidence meet. I build controlled environments that make security properties measurable: honeypots that expose real attacker behavior, detection-validation systems with independent ground truth, benchmark harnesses that separate model failure from evaluator failure, and forensic workflows that preserve provenance instead of relying on narrative confidence.

Current research extends that model into agentic AI security. SwarmKillChain examines how compromised or malicious agents appear across process, tool, protocol, and network layers, how trust and delegation fail in multi-agent systems, and where defenders can detect, delay, contain, or disrupt coordinated autonomous activity.

I also contribute to AI evaluation and benchmarking programs spanning software engineering, cybersecurity, agentic systems, multimodal prompt evaluation, rubric-based scoring, and deterministic test environments. The common thread is the same: evidence has to outrank the system's own claim that it worked.