Hybrid-HAVA: Enhancing AI Security in Healthcare with GAN-Based Adversarial Filtering

Abstract

The rapid adoption of Artificial Intelligence (AI) in healthcare has significantly improved 
diagnostic accuracy and patient care. However, AI-driven medical models remain highly vulnerable to 
adversarial attacks, which can manipulate predictions and compromise patient safety. The Healthcare 
AI Vulnerability Assessment Algorithm (HAVA) provides a post-attack vulnerability index (PAVI) 
but lacks real-time defense mechanisms. This article proposes Hybrid-HAVA, an enhanced security 
framework that integrates HAVA with Generative Adversarial Network (GAN)-based filtering to 
proactively defend against adversarial perturbations. Experimental results demonstrate that Hybrid
HAVA outperforms conventional defense strategies by mitigating adversarial noise while preserving 
diagnostic accuracy.