Context-Aware Image Encryption via Adaptive Entropy-Oriented Segmentation

Recently, the demand for secure visual data transmission has increased significantly in applications such as cloud computing, telemedicine, and intelligent surveillance systems. This study presents an adaptive image encryption framework called Adaptive Entropy-Based Segmentation (AESeg), which, unlike conventional segmentation-based encryption approaches that rely on predefined geometric partitions, introduces an entropy-driven segmentation paradigm where the spatial distribution of information directly guides the allocation of cryptographic strength. In the proposed framework, high-entropy regions associated with rich textures and structural details are protected using strong cryptographic mechanisms, whereas low-entropy homogeneous regions are secured using lightweight encryption operations. By integrating entropy-based segmentation with the Dynamic Cipher Composition (DCC) architecture and adjacency-aware cryptographic allocation, AESeg enables context-aware encryption while reducing unnecessary computational overhead. Experimental results show that AESeg achieves entropy values close to 7.99, NPCR above 99.6%, and UACI above 33.4%, while reducing encryption time by 28–35% compared to conventional uniform encryption approaches. These results demonstrate that the proposed method achieves an effective balance between security robustness and computational efficiency.