Independent security research focused on how modern systems actually fail- cloud infrastructure, container isolation, Kubernetes orchestration, and AI platforms. Published findings, disclosed vulnerabilities, and conference presentations going back to 2020.
A chain attack on Kubernetes orchestration, combining container isolation bypasses and misconfiguration paths to achieve host-level persistence. Presented at BSides USA (VA). 1M+ systems affected.
Adversarial attack research targeting LLM provider data pipelines via unvalidated API endpoints, demonstrating how model poisoning can be introduced through external-facing interfaces. Presented at REcon.
A new class of malware that uses AI-driven decision-making to change its execution logic while running. Instead of following a fixed attack chain, it adapts to defensive controls, permission boundaries, and execution outcomes—creating non-deterministic behavior that challenges signatures, sandboxes, replay analysis, and traditional behavioral detection.
Chaining IAM misconfiguration, metadata service access, and container privileges to achieve cluster-admin in a real AWS EKS environment. Full disclosure published independently.
Exploiting shared kernel resources in Google Colab’s multi-tenant environment to break isolation boundaries between user sessions.
Investigating how Google Cloud Shell and similar ephemeral environments handle isolation, and where those assumptions break down.
A misconfiguration chain in Azure Key Vault access policies allowing unauthorized secret extraction across tenant boundaries under specific conditions.
Regularly contributing to the global security community by sharing findings at major industry conferences and global conferences, events.
Identifying non-obvious entry points by challenging the fundamental assumptions of architectural design.
Mapping complex dependencies to find weaknesses where components interact with external systems.
Moving beyond single-step bugs to create high-impact exploit chains that demonstrate business risk.