Selective dual-key encryption for real-time protection of medical CT images using hybrid chaotic systems

The safe transmission and storage of medical images has become a major problem due to the growing usage of telemedicine platforms and cloud-based healthcare systems. Due to the massive data quantities, considerable spatial redundancy, and extremely sensitive diagnostic information contained in medical CT images saved in the DICOM format, traditional full-image encryption is computationally costly for real-time applications. A selective dual-key medical picture encryption scheme that preserves computational performance while safeguarding critical diagnostic areas is presented in this paper. The technique uses a ROI-based approach where two different secret keys are used to encrypt the foreground and background regions separately. A key-dependent dynamic substitution box is generated using a hybrid chaotic system that combines logistic and tent maps, providing high nonlinearity and resistance to chaotic degradation. To further enhance security, the encryption process integrates multi-stage diffusion mechanisms, including DCT-based diffusion, pixel-wise XOR chaining, global chaotic permutation, and bit-level whitening. Experimental results obtained from CT images in the SIIM dataset [45] demonstrate that the proposed algorithm effectively suppresses pixel correlations, produces uniform histograms, and achieves near-ideal Shannon entropy (≈ 7.99). High NPCR and UACI values confirm strong diffusion and sensitivity to minor changes in both keys and input pixels, while NIST-inspired statistical tests indicate that the encrypted outputs exhibit randomness characteristics close to ideal random sequences. These results confirm that the proposed framework provides robust security and is suitable for real-time medical image protection.