Design of a 2D Rulkov neuronal hyperchaotic map: Dynamical analysis and application in medical target encryption

A lightweight encryption framework for medical image data is presented in this paper. A two-dimensional Hénon–Rulkov neuronal hyperchaotic map(2D-HRNHM), compressed sensing, and object detection technology are integrated to realize lightweight image encryption. The 2D-HRNHM chaotic map is constructed by coupling the Hénon map with a neuronal model to provide a pseudo-random key stream for the encryption algorithm. The phase diagram, bifurcation diagram and Lyapunov exponent spectrum of the constructed map are analyzed to verify its dynamic characteristics. A compressed sensing convolutional reconstruction network (CSCRN) is constructed to enhance the quality of image reconstruction. In the encryption workflow, regions of interest (ROI) are first extracted from medical image data by the target detection model. Subsequently, initial encryption and compression are performed on the ROI via chaotic sequences and compressed sensing, to reduce data volume while guaranteeing data security. Finally, the security level is further enhanced through an additional encryption step driven by chaotic sequences. Experimental results show that the ROI-based strategy reduces the encrypted data volume by approximately 40%. Across test images, the proposed method achieves average NPCR and UACI values of 99.6036% and 33.2931%, respectively, while the average information entropy of the encrypted images reaches 7.9996. At a compression sampling rate of 0.5, the reconstructed image achieves a PSNR of 35.2562 dB and an MSE of 19.3848, with an overall encryption time of 0.091 s. These results demonstrate that the proposed framework provides high security, efficient encryption, and robust reconstruction for medical image protection.