Research On Image Encrption Algotithm Based On Generative Neural Network

In recent years,with the rapid development of computer technology,the issue of information security has attracted more and more attention.With the wide application of digital images,a large amount of image information is transmitted on the network,which may be stolen by illegal users during storage or transmission,resulting in information leakage.Image encryption technology is one of the important means to protect the security of digital image information.Although the traditional image encryption algorithm has achieved the function of image encryption to a certain extent,there are still some defects in terms of reliability,security,and encryption speed.The image encryption algorithm based on deep learning has become one of the research hotspots in the field of image encryption due to its advantages in nonlinear mapping and image generation.Based on deep learning technology,this dissertation proposes two image encryption algorithms.The main research work is as follows:(1)This dissertation proposes a generative adversarial network-based image encryption algorithm related to plaintext.Firstly,the Generative Adversarial Network is used to generate the key associated with the plaintext.In the encryption process,three encryption modules are designed:pixel replacement,image scrambling,and image diffusion.The decryption process is the reverse process of the encryption process.Experimental results show that the generated key has passed 15 randomness tests of NIST and has better randomness.The proposed encryption algorithm has good performance,has a large key space,and the ciphertext image information entropy is close to the optimal value.It can resist common attack methods such as noise attack,shear attack,and statistical attack,and improves the reliability and safety.(2)This dissertation proposes a grayscale image encryption algorithm based on Piecewise chaotic map and autoencoder.Image encryption is completed by combining Piecewise chaotic map scrambling and autoencoder network encoding.Firstly,the original image is scrambled by the Piecewise chaotic map,and then the scrambled image is diffusely encoded by the encoding network.A uniform distribution loss function is introduced in the encoding network to increase the randomness and security of the ciphertext image,and a structural similarity loss function is introduced in the decoding network to improve the quality of the decrypted image.The experimental results show that this algorithm has better encryption and decryption performance,the correlation of adjacent pixels of the ciphertext image is close to 0,and has a great advantage in encryption speed,which can provide new ideas and methods for image encryption.