Secure Medical Image Steganography using Lightweight Cryptography with Hybrid Machine Learning and Deep Learning
Year: 2026
Type: conference-paper
Abstract: Advances in digital health have led to widespread use of telemedicine and imaging in the field. This change brought a hidden challenge into practice, namely the security issue related to the confidentiality of information carried by X-rays. Therefore, the paper will explore a way to tackle the security problem with regard to the preservation of accuracy of classification. The suggested solution will be tested on a large dataset of musculoskeletal X-rays called MURA, containing about 40,000 pictures. Features will be extracted with the help of four models: MobileNetV2, ResNet50, EfficientNetB0, and Xception. After that, classifications will be performed using Random Forest, SVM, Logistic Regression, and XGBoost. As opposed to one-classification, this hybrid approach involves weighting the two classifications, which ensures stability. With regard to security, a lightweight encryption will be implemented through elliptic curve cryptography, whereas hiding the confidential content will be achieved by means of least significant bit steganography. Finally, as the last stage, the model will undergo testing, which will include proving that neither the classification nor the transmission process affect the image adversely.
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7204104979