There is increasing interest in applying IoT healthcare systems in heart disease classification for remote monitoring and diagnosis. The traditional methods confront problems related to data protection, missing, and high dimensionality, and they require storage security, accuracy, and quality. Further, the existing approaches do not incorporate proper encryption with better machine deep learning algorithms to protect the identity as well as the accuracy of the system. To overcome these challenges, this study presents a novel solution covering secure data transfer through the Registered Attribute-Based Encryption Algorithm and the Steerable Graph Neural Network classifier for proper heart disease classification. In this study, heart disease is defined and classified from the cardiovascular disease dataset. To enhance the data’s quality, the first step of this ineffective means-mode normalization process was to denoise and scale the IoT data. RAbEA guarantees secure attribute-based encryption for the IoT healthcare data; SGNN improves classification performance by capturing intricate interactions of the data attributes. Furthermore, hyperparameters of the SGNN, utilizing Termite Alate Optimization (TAO) for optimization, present a new approach to escape high dimensional search space local optima and provide a superior classification rate. The suggested RAbEA-SGNN strategy offers exceptional security and achieves 99.79% disease classification accuracy compared to current methods.