Research On Chaotic Image Encryption Method Based On The Combination Of Generative Adversarial Networks And Neuron Mechanism

With the rapid development of computer information technology,information as an important resource has broken the traditional restrictions on time and space distribution,especially in the fast development of information today,information resources have become an important component in the development of human society,information security is an important technique to protect information resources.Compared with text information,image information has the features of rich data,high visualization and strong pixels correlation,which makes the traditional text information cryptosystem unable to meet the security requirements of image information.Because of the good cryptographic characteristics of chaotic system,it has been used more and more in image information security protection.However,the chaotic system exists cyclic degradation,the key space and the randomness of sequence need to be improved,so designing a chaotic system with better chaotic dynamic performance and combining with the image encryption method with high efficiency and security,so as to better ensure the security of image information.According to the characteristics of chaotic system and the security requirements of image encryption and decryption,taking improving the Lyapunov exponent and chaotic key space of chaotic system as the main line,this paper studies the chaotic image encryption method based on the combination of Generative Adversarial Networks(GANs)and neuron mechanism.The new one-dimensional chaotic system and new four-dimensional hyperchaotic system with large Lyapunov exponent and many chaotic control parameters are proposed,which are respectively combined with complex mathematical models such as GANs,Deep Convolution Generative Adversarial Networks(DCGANs),DNA coding and cloud model to construct a pseudo-random sequence generator,which generate pseudo-random sequences with good randomness.The new one-dimensional chaotic system and new four-dimensional hyperchaotic system combine with neuron mechanism,finite field,quaternion,improved Feistel block network and improved generalized Feistel block network to encrypt the image with block scrambling and diffusion.The main researching content is as follows:1.In order to obtain large key space and Lyapunov exponent and improve the performance of chaotic dynamics,the new one-dimensional chaotic system and new four-dimensional hyperchaotic system are proposed by adding chaotic control parameters based on Logistic system,quadratic chaotic system,Li hyperchaotic system,Lorenz chaotic system and hyperchaotic system proposed in relevant literature,and the performance of chaotic dynamics and chaotic sequence are analyzed.It shows that the key space,Lyapunov exponent,chaotic dynamics performance of the chaotic system and the randomness of chaotic sequence of the chaotic system proposed in this paper are improved.2.Aiming at the problems of chaotic sequence periodicity and different degrees of characteristic degradation caused by the limited calculation accuracy of chaotic system,the new one-dimensional chaotic system and new four-dimensional hyperchaotic system are combined with complex mathematical models such as GANs,DCGANs,DNA coding and cloud model to construct the new pseudo-random sequence generators.The randomness of the generated pseudo-random sequence is tested.The experiments show that the pseudo-random sequences have good randomness.At the same time,the periodicity and characteristic degradation of chaotic system are improved.3.In order to improve the scrambling diffusion effect,encryption efficiency and security of image encryption,combining the pseudo random sequence as key stream with the neuron mechanism,finite field,quaternion,improved Feistel block network and improved generalized Feistel block network to encrypt the block image with scrambling and diffusion are proposed,which ensures high security of the image encryption method.4.In order to verify the security of the proposed image encryption method,this paper discuss and analyze from both qualitative and quantitative aspects.The experiments show that the proposed image encryption method has good security.