An Encryption Method of Multiple Images and Multiple Region of Interest Based on DNA Sequence and Chaotic System
Jun Peng, Ke Xu, Qing Li, Shangzhu Jin, Yingxu Wang · 2022 IEEE 17th Conference on Industrial Electronics and Applications (ICIEA) · 2022
Most of the existing image encryption methods based on the region of interest (ROI) use traditional methods to manually screen regions, which are notoriously difficult to implement serialized and cannot handle multiple ROI well. In addition, the low dimensional Logistic map widely used has the shortcoming of small key space and poor security. Furthermore, DNA computing, which is popular in current research, also has problems such as fixed rules and low sensitivity. Therefore, this paper improves the advanced deep learning object detection algorithm YOLOV4 tiny to process multiple ROI more quickly and accurately and then designs an elegant encryption strategy for multiple images and multiple ROI combining Logistic map, Chen system, and DNA computing. Specifically, we calculate the initial value of the Logistic map from the multiple ROI, which significantly increases the keyspace and key sensitivity. At the same time, we utilize the Chen system to design a unique DNA coding and computing rule for each ROI dynamically to improve the pixel-level scrambling and diffusion. Finally, we combine multiple images into a larger picture, and use SHA256 to calculate its hash value as the initial value to generate a chaotic sequence to complete row-column scrambling, and realize serialized multiple images encryption. Simulation experiments and security analysis results indicate that our method has good encryption and decryption effects, as well as high security and robustness, and can resist statistical attacks and brute force attacks.