Contribution à l’étude expérimentale d’un système de cryptographie de signaux : application aux images médicales

Abstract

Over the last few decades, telemedicine has grown considerably thanks to the development of information and communication technologies. A lot of information is exchanged between health care practitioners. Much of this information consists of medical images, many of which contain sensitive patient information and require confidentiality. In order to ensure the security of images during their transmission from one point to another, several encryption algorithms have been developed, among which those based on the fusion or mixing of images are increasingly used. In fact, the majority of image fusion systems aim at providing richer information for better decision making. However, in recent years, image fusion techniques have been developed more for image encryption. In this paper, we propose two secret key encryption algorithms based on image fusion. Subsequently, a beginning of implementation of the first algorithm is led on the STM32F407ZET6 microprocessor embedded board in order to appreciate a possible implementation of the cryptosystem on this card. The first one uses the Henon, logistic-May and logistic-sine chaotic maps in the encryption process, where two images are scrambled separately and then mixed to produce two hybrid images. The second performs two-level target image fusion : the discrete cosine transform (DCT in english) is used first to merge the target images into two multiplexed images, each of which is permuted at the block level. The second level of fusion is performed by a non-linear relation inspired by the Cramer model. may-gaussian and gaussian-gompertz maps are used as pseudo-random number generators in this second algorithm. The performance evaluation tests of the proposed algorithms have been performed on standard images of the scientific community, and then on medical images. The results obtained after simulations, analyses and comparison with algorithms of the literature, show that they are satisfactory in terms of robustness, speed of execution time, quantity of images to encrypt and quality of reconstructed images. Moreover, the proposed algorithms present a simple structure, and are adapted to an implementation in a hardware architecture such as the STM32F407ZET6.