DNA Encoded Color Image Encryption Based on Chaotic Sequence from Neural Network

Communication in any form requires secrecy to ensure the message is received and interpreted only by the intended receiver. Fast moving communication technologies enables users to communicate multimedia messages like images, videos, animations and so on. Encryption of images aim to hide image content by the confusion and diffusion process. Secret keys are required to encrypt the images and other multimedia messages. Chaotic maps are good sources of secret keys, but when used for multimedia encryptions, they lack periodicity. Digitalization of chaotic maps result in poor secret keys unfit for cryptography. This paper proposes a novel Iterative Neural Network (INN) framework built from a hybrid Convolutional Logistic Hénon map. The hybrid map is studied for its chaotic behavior using bifurcation diagrams and Lyapunov Exponent of the map is found to be positive. The key stream generated are tested for their statistical randomness using NIST SP 800–22 test suite. The secret keys are used along with DNA encoding to perform the color image encryption. The secret keys and the cipher images are analyzed for their security characteristics like entropy, correlation coefficient, ability to withstand differential attacks. This novel encryption is able to withstand differential attacks and the neural network based key stream generation is found to have 5% more entropy than existing techniques and the correlation coefficient of generated keystream is increased by 8% compared to other existing techniques. The proposed INN based encryption of images was able to resist differential attacks and security of cipher images are improved making it a better candidate in the era of multimedia cryptography.