The rapid integration of cloud computing and Internet of Things (IoT) technologies has transformed modern healthcare into an intelligent, data-driven ecosystem capable of continuous monitoring and real-time analysis of patient information. Cardiovascular disease remains the leading global cause of mortality, demanding early, accurate, and secure diagnostic solutions. This study introduces a secure and efficient disease classification framework, the Multi-Head ProbSparse Cascaded Graph Network optimized with Draco Lizard Optimizer (MH-PSCGN-DLO), designed to enhance prediction accuracy while ensuring strong data protection. The system incorporates Multi-Level Encryption (MLE) to safeguard sensitive medical data, followed by an advanced preprocessing pipeline—Adjusted Min-Max, Decimal Scaling, and Statistical Column Normalization (AMM-DS-SCN)—to reduce noise and standardize heterogeneous IoT data. Parrot Optimization (PO) selects the most informative features, while MH-PSCGN combines Disentangled Cascaded Graph Convolution and Multi-Head ProbSparse Self-Attention for robust classification. The DLO optimizer further enhances model stability and convergence. Experiments on the cardiovascular disease dataset demonstrate 98.4% accuracy, 0.01 error rate, and significantly lower encryption/decryption times compared to existing methods. These results indicate that MH-PSCGN-DLO provides a reliable, secure, and computationally efficient foundation for real-time clinical decision support, with strong potential for future extension to other chronic disease domains.
