Privacy-Preserving Healthcare Data Security Using Large Language Models and Adaptive Access Control

Protecting privacy-sensitive data in the digital healthcare sector is imperative due to the escalating threat of data breaches, unauthorized access, and cyberattacks. Traditional security systems, including rule-based and signature-based approaches, may struggle to adapt to evolving circumstances and manage high-risk situations. This study presents a security solution that uses adaptive access control systems and a Large Language Model (LLM) to protect sensitive patient information and electronic health records (EHRs). This method ensures HIPAA and GDPR compliance through context-aware risk assessment, anomaly detection, and query sanitization procedures, enhancing data security. Using a dynamic threat modeling technique, the proposed system identifies zero-day vulnerabilities and malicious behavior through real-time healthcare data transmissions. Comparative analyses show that the proposed LLM-based methodology outperforms traditional security techniques regarding accuracy, recall, precision, and F1 score. This improves threat detection rates and reduces false positives. The results demonstrate the effectiveness of LLM-based security solutions in securing medical records and ensuring patient privacy, thus providing robust and flexible protection.