AI-built test environments
Agents that build test versions of external services, so AI-generated security workflows can be tried without connecting to the real service.
Explore the projectCurrent research.
Two projects lead my work: agent-generated test environments and benchmarks for training small models to write security playbooks.
EXPLORE FURTHER
More research
Related projects in security automation, threat intelligence, and evaluation.
Agent-based playbook generation
An agent system that uses retrieved context and test feedback to turn security objectives into SOAR workflows.
Turning threat reports into detections
Connecting unstructured threat reports with log evidence and validated detection patterns.
Comparing human and AI evaluations
Studying how rubrics and evaluation procedures shape the agreement between human judgments and LLM assessments.
Learning defensive strategies against APTs
Modeling lateral movement and constructing simulation environments for reinforcement-learning-based defensive strategies.
Connecting attack evidence with graphs
Linking intelligence, kernel audit logs, and attack behavior to investigate multi-stage threats.