Intelligent Concealment of Healthcare Information: A Novel Machine Learning-Driven Multi-Layered Steganographic Approach for Enhanced Medical Data Privacy

Abstract

Rapidly, the adoption of cloud computing, the “Internet of Medical Things (IoMT)”, telemedicine, and electronic medical records has made an increasing number of people concerned about the security of medical information being transmitted. This study suggests a novel multi-layered steganographic approach designed using an intelligent machine learning system for privacy preserving healthcare communication. The structure is based on the NIH Chest X-ray dataset of 112,120 X-rays from 30,805 patients. The basic steps of medical image preprocessing include resizing, image normalization, CLAHE enhancement, median filtering, and AES-256 encryption, ResNet-50 based deep feature extraction, adaptive region selection using XGBoost and DWT-LSB embedding. The medical images are preprocessed by resizing, image normalization, CLAHE enhancement, median filtering and AES-256 encryption followed by ResNet-50 deep feature extraction, XGBoost based adaptive region selection and DWT-LSB embedding. Experimental results give an impressive PSNR of 54.32 dB and SSIM of 0.978, embedding capacity of 0.175 bpp, and extraction accuracy of 99.21%. The proposed framework is also able to decrease the time of operation from 19.7% to 37.2% in terms of steganalysis detection ratio and 0.99 in terms of AUC, and reduce the encryption/decryption time to 1.72 s and 1.63 s, respectively. Based on these outcomes, the framework introduced in this study is proven to be efficient, robust, and highly secure for concealing healthcare data while maintaining user privacy in the present context of digital healthcare.