A hyperchaotic image encryption algorithm based on LSTM neural network and lifting wavelet transform
Ning Mao, Xiaojun Tong, Miao Zhang, Zhu Wang
Abstract
Abstract In order to solve the problems of simple permutation-diffusion structure, low encryption efficiency and small chaos range of existing chaotic systems, this paper proposes a hyperchaotic image encryption algorithm based on LSTM neural network and lifting wavelet transform. By building upon the Lorenz chaotic system, we construct a new hyperchaotic system in this paper, which has more complex dynamic characteristics and higher Lyapunov exponent. Then, the image is encrypted by lifting wavelet transform and zigzag conversion algorithm. Finally, the trained LSTM neural network is used to process the Henon chaotic sequence to obtain the second key for encryption, which can effectively resist chosen-plaintext attack. The experimental findings indicate that the proposed encryption algorithm has good performance in key security, differential attack, statistical analysis and operation efficiency, which indicates that the algorithm has high security.