An efficient image encryption using deep neural network and chaotic map

Abstract

Inspite of progressive growth of cryptography, encrypting sensitive information of an image is still a computationally complex task. After reviewing existing literature, it is now known that security problems are yet not solved and there is an open scope of further research. In most recent times, it has been noticed that neural network has proven cost effective optimization mechanism in offering security towards images. However, such implementation are computationally expensive process and do not solve various diversified attacks on image. Hence, the prime purpose of proposed system is to introduce an analytical research methodology for presenting a sophisticated framework where deep neural network has been used for optimizing the performance of simple encryption approaches. The robustness of optimization principle is further added with chaotic map concept for enhanced security performance. The study outcome shows that proposed implementation offers much better security performance without any negative effect on image quality.

Introduction

Data is one of the essential assets for all the enterprise which are required to be effectively protected. The concern associated with the data protection originates when the data are subjected for storage, transmission, and processing over any form of network system. At present, there are various forms of data security algorithm but it has its own scope of application, beneficial factors, as well as limiting factors too. The application of the data security approaches is also specific to the forms and structure of the data and it also depends if they are implemented over centralized or distributed system. Security has always been the major concern in many enterprise applications that are hosted over distributed networking system.

The Information Security refers to the procedures and approaches which are intended and applied to secure sensitive, private and confidential information or data from illegal access, usage, misappropriation, revelation, obliteration, alteration. Confidentiality, integrity, and availability are the main components of information security. The current state of world poses that most of the Information are exchanges across the internet and the storage of these data in open networks have created an environment in which illegal users can obtain the important information from that image data. Hence vulnerability of image data is significantly more in this regards.

In open environments, there are several security problems associated with the processing and transmission of digital images. Therefore it is necessary to affirm the integrity and confidentiality of the digital image being transmitted. A cryptography technique is one of the efficient mechanisms in order to resist potential threats. Both of symmetric-key and asymmetric-key encryptions are considered as primary types of encryption. However, performing encryption over images is quite a challenging task in contrast to normal data owing to massive level of information carried within an image. The most challenging aspect of applying encryption over an image is the degradation of an image while extracting the encrypted image.

At present, there are various security mechanisms towards assisting resisting attacks over an image, but they are not found completely fruitful towards balancing security demands and image quality demands. Most recently, deep neural network is found to offer better results even in presence of unstructured data with complete independence towards data labelling. Apart from this the chaos behaviour of chaotic map also offers extensive security. Hence, integration of deep learning and chaotic behaviour can offer a better form of encryption method of images. Therefore, the proposed paper presents a model where deep learning- chaotic map has been used for performing better optimization towards strengthening the encryption performance of image.

The rest of the paper is organized as follows: Section 2 discusses about the existing literatures where different techniques are discussed for detection schemes used in power transmission lines along discussion of research problems and proposed solution in 3. Section 4 discusses about algorithm implementation followed by discussion of result analysis in Section 5. The last Section 6 concludes the paper.