TRUeIDS: A Unified IoMT Intrusion Detection Framework with Zadeh’s Trust Fuzziness and R3GAN Reputation Robustness

Healthcare networks built on the Internet of Medical Things (IoMT) face heightened risks not only from conventional DoS, DDoS, spoofing, and reconnaissance attacks, but also from advanced adversarial attacks—including evasion, mimicry, history poisoning, membership inference, and model inversion—that most state-of-the-art IDS fail to address. To overcome these limitations, this paper presents TRUeIDS, a three-stage novel intrusion detection framework that strengthens security through integrated transient trust index and cumulative reputation learning. Stage I estimates short-term device trust using Zadeh’s Interval Type-2 Fuzzy Logic (IT2FL) to effectively manage uncertainty and adversarial noise. Stage II employs Regularized Residual Robust Generative Adversarial Network (R3GAN)-based reputation modeling to capture long-term stability, variability, and resistance to on–off behaviors. Stage III introduces a Zero-Trust decision mechanism that normalizes and fuses trust and reputation scores into adaptive thresholds, classifying devices into Allow, Challenge, or Deny states. Experiments conducted on the CICIoMT2024 dataset across binary, 6-class, and 19-class scenarios demonstrate robust performance, achieving accuracies above 98.9\%, MCC up to 0.9852, and ROC-AUC consistently beyond 99.2\%. These results confirm that TRUeIDS provides a mathematically rigorous and practically validated solution for secure, scalable, and resilient intrusion detection in IoMT environments.