Home | Posts | Papers | Talks | About
Angles of Attack: Engineering Mission-Critical AI Systems
Zenodo | Google Scholar | ORCID
US Patent 12,093,400, 2024: Systems and methods for model security in distributed model training applications
A novel distributed security architecture for Federated AI/ML data poisoning defense. »
AI-DAL: Towards Security Design Assurance for Artificial Intelligence Systems in Production
The rise of artificial intelligence applications in society, and their accompanying security concerns, has created a need for regulatory oversight that is auditable, actionable, and adaptable to a rapidly changing technological landscape. Methods from safety-critical software engineering, particularly aerospace, are adapted to use in production AIML to aid both practitioners and regulators in establishing design thresholds for AIML system security. Assignment of AI Design Assurance Levels (AI-DAL) to projects/components, along with production of related compliance artifacts, is proposed as a means of consistently applying appropriate design requirements based on a system’s potential adverse impact. »
Securing AIML Systems in the Age of Information Warfare
A novel AI/machine learning security operations (AI/MLSecOps) architecture, including systems for operationalized security, auditing, data threat, and information warfare detection, along with OODA Loop-based game theoretic modeling of information warfare in AI/ML systems, and boolean path threat modeling & cyber resiliency metrics adapted to the canonical MLOps development cycle. »