Joint Encryption Model Based on a Randomized Autoencoder Neural Network and Coupled Chaos Mapping

Abstract

Following an in-depth analysis of one-dimensional chaos, a randomized selective autoencoder neural network (AENN), and coupled chaotic mapping are proposed to address the short period and low complexity of one-dimensional chaos. An improved method is proposed for synchronizing keys during the transmission of one-time pad encryption, which can greatly reduce the usage of channel resources. Then, a joint encryption model based on randomized AENN and a new chaotic coupling mapping is proposed. The performance analysis concludes that the encryption model possesses a huge key space and high sensitivity, and achieves the effect of one-time pad encryption. Experimental results show that this model is a high-security joint encryption model that saves secure channel resources and has the ability to resist common attacks, such as exhaustive attacks, selective plaintext attacks, and statistical attacks.