Publications
2 entriesArticles and papers on AI security, assurance, and governance in healthcare and critical infrastructure.
The Governance Gap in Healthcare AI: Why Policy Alone Can't Secure Medical Machine Learning
Why governance standards like ISO/IEC 42001 can't, by themselves, tell you whether a specific AI model has actually been tested against data poisoning or backdoor attacks — and why that gap matters most for AI making real clinical decisions.
EO 14409
HSCC guidance
ISO/IEC 42001
NIST AI RMF
Fusing Spectral Signatures and Activation Clustering for Backdoor Detection in Healthcare Imaging Models
A fusion rule combining spectral signature analysis and activation clustering into a single per-sample poisoning score, implemented as an open-source detection pipeline and evaluated on healthcare-imaging and natural-image benchmarks. Published on arXiv:2609.14290 and Zenodo, DOI 10.5281/zenodo.22732363.
MITRE ATLAS AML.T0020
NIST AI RMF · Measure
Spectral signatures
Activation clustering