Aiming at the problems such as small key space and incomplete color channel
encryption in traditional image encryption, this paper proposes a novel image
encryption algorithm based on generative adversarial network (GAN) and DNA
dynamic encoding. This paper introduces GAN into random key generation, and
uses GAN to learn and train the random key generated by hyperchaotic system.
A parallel chaotic system is used to generate two sets of pseudo-random
sequences, and DNA dynamic encoding is introduced to further transform them
to generate a new sequence. The pixel-level diffusion and scrambling of
images within and between channels are carried out by using random sequences.
The experimental results show that the randomness of GAN can significantly
expand the key space, and the proposed algorithm has significant advantages
in the security and anti-attack ability of ciphertext images.