Logo image
Back
A Local, Portable, and Secure Knowledge-Enhanced Agentic AI System for HIPAA Compliance
Dissertation   Open access

A Local, Portable, and Secure Knowledge-Enhanced Agentic AI System for HIPAA Compliance

Md Abdur Rahman
University of West Florida Libraries
Doctor of Philosophy (PHD), University of West Florida
2026

Metrics

1 Record Views

Abstract

Protecting health data is the most essential thing when any AI systems need to use hugeamounts of health information. There are strong guidelines known as the Health Insurance Portability and Accountability Act (HIPAA) for collecting, processing, storing, and distributing these data. Conventional security methodologies, such as rule-driven and signature-dependent mechanisms, may often struggle to manage high-risk situations. Moreover, Large Language Models (LLMs) can keep secure sensitive health information for some cases. However, LLMs could become outdated because they are trained on data from a specific point in time. Also, retraining or fine-tuning them with updated information is also expensive and time-consuming, as billions of parameters need to be recalculated in terms of modified weights. Most importantly, they are often vulnerable to disclosing sensitive information. To address these issues, an agentic system for HIPAA compliance is proposed to check various sensitive healthcare documents using several agentic tools, whether they comply with HIPAA rules or not. In fact, it uses various data like HIPAA rules, regulations, sensitive healthcare data, company ocuments, etc., to convert it into vectorized high-dimensional data stored in a vector search library called FAISS (Facebook AI Similarity Search) as an external knowledges to help system through providing relevant context so that it can be used as addtional specific information to make a response with reasoning. Also, one of those tools checks for sensitive data and prevents it from being provided to the system as contexts. Moreover, new or updated HIPAA policy can be easily added without requiring repetition of the whole process. This work could reduce a significant compliance disparity between healthcare operations and HIPAA regulations. We also describe performance metrics confirming the effectiveness of this work.
pdf
A Local, Portable, and Secure Knowledge-Enhanced Agentic AI System for HIPAA Compliance3.44 MBDownloadView
Preprint Dissertation pdf Open Access

Details

Logo image