A Novel Black-Winged Kite Algorithm with Deep Learning for Autism Detection of Privacy Preserved Data

Abstract

Autism Spectrum Disorder (ASD) is a complex neurodevelopmental condition that causes multiple challenges in behavioral and communication activities. In the medical field, the data related to ASD, the security measures are integrated in this research responsibly and effectively to develop the Mobile Neuron Attention Stage-by-Stage Network (MNASNet) model, which is the integration of both Mobile Network (MobileNet) and Neuron Attention Stage-by-Stage. The steps followed to detect ASD with privacy-preserved data are data normalization, data augmentation, and K-Anonymization. The clinical data of individuals are taken initially and preprocessed using the Z-score Normalization. Then, data augmentation is performed using the oversampling technique. Subsequently, K-Anonymization is effectuated by utilizing the Black-winged Kite Algorithm to ensure the privacy of medical data, where the best fitness solution is based on data utility and privacy. Finally, after improving the data privacy, the developed approach MNASNet is implemented for ASD detection, which achieves highly accurate results compared to traditional methods to detect autism behavior. Hence, the final results illustrate that the proposed MNASNet achieves an accuracy of 92.9%, TPR of 95.9%, and TNR of 90.9% at the k-samples of 8.