Enhanced image encryption scheme with new mapreduce approach for big size images

Abstract

Achieving a secured image encryption (IES) scheme for sensitive and confidential data communications, especially in a Hadoop environment is challenging. An accurate and secure cryptosystem for colour images requires the generation of intricate secret keys that protect the images from diverse attacks. To attain such a goal, this work proposed an improved shuffled confusion-diffusion based colour IES using a hyper-chaotic plain image. First, five different sequences of random numbers were generated. Then, two of the sequences were used to shuffle the image pixels and bits, while the remaining three were used to XOR the values of the image pixels. Performance of the developed IES was evaluated in terms of various measures such as key space size, correlation coefficient, entropy, mean squared error (MSE), peak signal to noise ratio (PSNR) and differential analysis. Values of correlation coefficient (0.000732), entropy (7.9997), PSNR (7.61), and MSE (11258) were determined to be better (against various attacks) compared to current existing techniques. The IES developed in this study was found to have outperformed other comparable cryptosystems. It is thus asserted that the developed IES can be advantageous for encrypting big data sets on parallel machines. Additionally, the developed IES was also implemented on a Hadoop environment using MapReduce to evaluate its performance against known attacks. In this process, the given image was first divided and characterized in a key-value format. Next, the Map function was invoked for every key-value pair by implementing a mapper. The Map function was used to process data splits, represented in the form of key-value pairs in parallel modes without any communication between other map processes. The Map function processed a series of key/value pairs and subsequently generated zero or more key/value pairs. Furthermore, the Map function also divided the input image into partitions before generating the secret key and XOR matrix. The secret key and XOR matrix were exploited to encrypt the image. The Reduce function merged the resultant images from the Map tasks in producing the final image. Furthermore, the value of PSNR did not exceed 7.61 when the developed IES was evaluated against known attacks for both the standard dataset and big data size images. As can be seen, the correlation coefficient value of the developed IES did not exceed 0.000732. As the handling of big data size images is different from that of standard data size images, findings of this study suggest that the developed IES could be most beneficial for big data and big size images.