Secured and efficient MLP algorithm based on Sand Cat Swarm Optimization and homomorphic encryption for healthcare data

Security is necessary to ensure that patient data is transferred securely across connected devices due to it is sensitive and confidential.Still Still, certain areas of the data transmission network, such as server systems and Internet of Things (IoT) devices are unsecure, allowing hackers to access and exploit sensitive data.To address these issues, the machine learning algorithm is made more secure and effective by using the Multilayer Perceptron (MLP) algorithm with Optimized Sand Cat Swarm Optimization (SCSO) and CKKS based Fully Homomorphic Encryption (FHE). In this proposed solution, healthcare private data is used as an input for data security. LDA is first used for reducing the dataset dimensionality, which enables the development of a better model without losing any data. The LDA outcome is then sent into CKKS-FHE to encrypts the data by applying the AES-256 bit for key generation. The encrypted data is fed into MLP machine learning for secure training, and SCSO optimization is used in MLP to tune the hyperparameter to avoid overfitting while also predicting heart disease using a classifier. Performance measures for encrypting data are used to assess and compare the proposed method with existing methods. For the proposed model, the encrypted data achieved the following performance metrics: 88.41%, 87.56%, 89.32%, 87.15%, 89.65% and 89.13 for accuracy, precision, NPV, specificity, recall, and F1_score, respectively. From the experimental outcomes, the evaluated outcome of the proposed approach for encrypting data is better to secure the health care data efficiently.