Publication

Abstract : A private framework for heart attack prediction using powerful AI techniques and stringent privacy protections. Differential Privacy (DP) is being used in the preprocessing stage to secure the privacy of sensitive health data, thus allowing the data to remain useful while still blocking access to individual records. To it, a hybrid ensemble learning approach for feature learning robustly through Random Forest, Gradient Boosting, and stacked Artificial Neural Network (ANN) boosting for improved predictive accuracy. Also, in coping with data decentralization problems, Federated Learning (FL) and Lottery Federated Learning (Lottery FL) allow collaborative training of models without revealing the local data. To this end, while there is FL, quantum key exchange (QKE) is safeguarding the model parameters exchange, thereby implementing a quantumresistant encryption layer. Adversarial testing also encompasses Membership Inference and Byzantine Attacks to evaluate the strength of resistance. This holistic approach has a high score on one of the main criteria of the provision of heart disease risk prediction in a real-world scenario for the healthcare sector in a secure, scalable, and privacy-preserving manner.