Security and Privacy Challenges in Healthcare Internet of Things Applications in the Era of Agentic AI

Abstract

In recent years, Healthcare Internet of Things (HCIoT) applications have enabled remote patient monitoring, realtime diagnostics, and personalized treatment. As these systems start to use agentic AI, AI that can perceive, reason, and act on its own, by offering smarter care at the doorstep, they also bring new security, privacy, and ethical risks. This survey reviews security challenges for HC-IoT in the context of agentic AI, such as adversarial machine learning, data poisoning, model inversion, privacy leakage, and manipulation of autonomous decisions. We analyze existing defenses and show why many traditional security frameworks are not enough for dynamic, adaptive agentic AI systems. Then, we outline open research challenges and future directions, such as explainable AI security, zerotrust architectures for autonomous agents, federated and privacypreserving learning, post-quantum cryptography for wearables, and neuro-symbolic reasoning for secure decision-making. This work provides a clear roadmap for researchers, practitioners, and policymakers working to secure the next generation of intelligent healthcare systems.