A Triple Key 3-d Chaotic Image Encryption Method using Double Chaotic Neural Networks

Abstract: The thesis highly aims on the security of images for secure communication. To achieve this, the need is to confuse any unauthorized who is trying to get the image. Here the image planes are taken into consideration. Three chaotic functions are studied and implemented into the system to create a secure system. The Red, Green and Blue components as on X,Y and Z axis are extracted. The scrambling of these pixels using Arnold cat maps does not create any change in the intensity or color of the image. A Chebyshev map is also induced which again creates a set of manipulations on the generated keys. A Triple key encryption is carried out on each of the plane using a 3-D chaotic neural network using three different keys on individual networks. This creates a more complex system where, if either of the key is wrong, a scrambled image will be the decrypted output. As s result a highly secure system will be in effective. Nine different keys will decide the predictability of the framework. Neural network performs XOR operation and XNOR operations works on two stages that is dependent only on the flow of the operation. Chaotic systems provide unpredictability of execution. The direction of manipulations cannot be predicted as a slight change in the value of the keys will make many deviations in calculations. Analysis of the method results in infinite value of PSNR. Higher quality of encryption nearly 96%. Hence the use of multi- dimensional chaotic functions creates much security due to the unpredictability’s and randomness.

Keywords: PSNR ; CI; NPCR; UACI; QoI; CNN