A novel content-selected image encryption algorithm based on the LS chaotic model

Many holistic image encryption schemes have been proposed in recent decades that achieve a certain security level. However, considering the actual application environment, sometimes it is not necessary to encrypt the whole image, but rather a selected specific content of the image. Thus, this paper proposes a chaos-based image encryption algorithm for specific contents with non-regular size. First, we construct a novel LS chaos model that combines the Sine map and Logistic map and then improve its chaotic complexity by perturbing control parameters and introducing delayed states. Second, we obtain the coordinates of each pixel of specific image content using the PSPNet semantic segmentation model trained on the Cityscapes dataset and the Faster-RCNN target detection model trained on the MS-COCO dataset. Finally, the selected pixels are encrypted using a proposed non-regular size encryption algorithm. Several simulation experimental results prove that this new LS chaos model exhibits superior chaotic properties and generates complex pseudo-random sequences. Furthermore, our chaotic encryption algorithm effectively implements image encryption for specific content and shows excellent encryption performance, indicating that our encryption algorithm is highly competitive in terms of security capability.