An image encryption scheme using a chaotic neural network and a network with multistable hyperchaos

Abstract

In this paper a new image encryption scheme is proposed, by which the pixel values are bit-level shuffled by the encryption sequences generated by a complex network, while the pixel positions of diffused images are scrambled by the chaotic sequences generated by a time-delayed neural network. In this way the key space is greatly enlarged through the simultaneous use of the time-delayed neural network and the complex network with a high dimensional multistable hyperchaotic system as nodal model. In particular, different nodes in the complex network are selected to generate the encryption and the decryption sequences. To achieve the synchronization accuracy necessarily for correct decryption, the number of nodes needed in the fully connected complex network is investigated and determined. Through applications on different kinds of plain images, the proposed algorithm exhibits excellent performance on security, i.e., with better statistical and differential characteristics, better plain-image key sensitivity and a large enough key space. In addition the encryption scheme can also resist various knowing attacks.