<p>Software-defined networking (SDN) is an innovative method that extracts the fundamental infrastructure from applications and network services while allowing direct programmability of network control. The Healthcare Internet of Things (H-IoT) has grown exponentially as a result of advancements in wireless communication methods. Smart medical gadgets use sensors and actuators to gather information about the human body, which is subsequently transmitted to the fog layer for analysis. Nevertheless, the existing models have some drawbacks in terms of detection accuracy, scalability, and computational efficiency, especially when they are used for complex healthcare IoT networks. Hence, this research presents a novel hybrid deep learning (DL) based method that utilizes the Global Mask Transformer for SDN-IoT intrusion detection for healthcare. Initially, raw input data are gathered from the MCAD-SDN dataset and pre-processed; then the features are selected using a hybrid tent chaos assisted Harris Hippo Optimization (HTC-HHO) method, which is used to select optimal features and reduce dimensionality. Then, selected features are used in the Hybrid Global Mask transformer approach with one dimensional conventional layer and residual connection (1D-HGMT) model, are effectively used to categorize and detect various types of benign and malicious attacks. Experimental results establish that the proposed system, compared with the existing approach, attained an accuracy and specificity value of 99.95% and 99.91%, respectively. Experimental analysis used performance metrics such as accuracy, precision, recall, f1-score and throughput. This study demonstrates the efficacy of the proposed approach in identifying various forms of assaults in healthcare applications, and improving privacy, detection speed, and scalability.</p>