A Reversible Data Hiding Framework for Secure Brain MRI Protection Utilizing Deep Learning
Abstract
This study presents ConOs-Net, a deep learning–based reversible data hiding framework for secure brain MRI image protection. The model integrates CNN-based residual learning with the Osprey Optimization Algorithm (OOA) to achieve high-fidelity embedding and extraction. Experimental evaluation on the BRATS dataset demonstrates superior performance, achieving a PSNR of 46.38 dB, SSIM of 0.993, low BER, and high embedding capacity. The proposed approach effectively preserves diagnostic quality while ensuring full reversibility and data confidentiality, making it suitable for secure medical image transmission.
Fields of Study & Subjects
Advanced Steganography and Watermarking Techniques
Chaos-based Image/Signal Encryption
Digital Media Forensic Detection
Publication details
- Journal / source
- Engineering Technology & Applied Science Research
- Volume
- 16
- Issue
- 1
- Pages
- 31930–31938
- Publication date
- 9 February 2026
- Language
- EN