AI-Driven Adversarial Defense Framework with Generative Adversarial Network for Secure Healthcare IoT Ecosystems
Lisa Mmesoma Udechukwu
*
University of Southern California, 3551 Trousdale Pkwy, Los Angeles, CA 90089, United States of America.
Tunbosun Oyewale Oladoyinbo
University of Maryland Global Campus, 3501 University Blvd E, Adelphi, MD 20783, United States of America.
Nanyeneke Ravana Mayeke
University of the Cumberlands, 104 Maple Drive, Williamsburg, KY 40769, United States of America.
Temilade Oluwatoyin Adesokan-Imran
University of Ibadan, Oduduwa Road, 200132, Ibadan, Oyo, Nigeria.
Rukayat Oluwabukola Olasege
Ottawa University, 1001 South Cedar Street, Ottawa, KS 66067, United States of America.
*Author to whom correspondence should be addressed.
Abstract
This study developed and assessed an AI-driven adversarial defense framework using Generative Adversarial Networks (GANs) to secure healthcare IoT ecosystems against rising cybersecurity threats in medical settings. The research drew on datasets (CICIoMT2024, WUSTL-EHMS-2020, BoT-IoT, and Kaggle) and peer-reviewed studies to achieve three objectives: building a detailed threat model, designing a GAN-based defense optimized for healthcare IoT, and rigorously testing its effectiveness. The threat model revealed 127 vulnerability vectors, with adversarial attacks (32%) most prevalent, and a mean risk score of 7.82, highest for critical care devices (9.34). The GAN framework, featuring a multi-layer generator–discriminator pair and 128-dimensional encoder, achieved a mean accuracy of 95.8% against major adversarial attacks (FGSM 97.1%, PGD 94.8%, C&W 95.9%, UAP 96.4%), outperforming traditional defenses by 39.4%. With an MTTD of 82 ms, the system enables real-time deployment, allowing healthcare providers to integrate it directly into hospital IoT networks for proactive protection. Limitations include reliance on secondary data and high computational cost. Recommendations include hybrid datasets, explainable AI integration, real-world pilots, standardized metrics, and federated learning to enhance scalability and adaptability.
Keywords: Healthcare IoT, generative adversarial networks, adversarial attacks, cybersecurity defense, threat model