Image Encryption Based on Neural Network Architecture and Chaotic Systems

Abstract

In this digital era security is an issue that is to be addressed. For this purpose, a variety of encryption techniques are being developed. This paper looks into a technique of image encryption/decryption that utilizes 3 layers of mapping of pixels on different levels of the RGB planes. Two of them are intraplanar processes and one is an interplanar process to ensure security on all levels. The first layer is a 2 Dimensional logistic mapping that happens on all the 3 RGB planes but only in the plane. The second layer, inspired by Neural Network architecture, adds weights to the individual pixels and shuffling takes place between the 3 RGB planes. In this layer, 3 chaotic functions are used to generate respective weights for the pixels. These functions are Newton-Leipnik chaotic system, Rossler attractor equations, Volta chaotic system. The third layer is again a 2 Dimensional mapping using Duffing map that shuffles the pixels in the planes. This technique has three layers which further have sub-layers make it a more secure technique.