Bio-Inspired Optimization Algorithms for Image Steganographic and Cryptographic Applications

Abstract

The multimedia data communication in the form of images is done in different applications, namely, healthcare, defense, and satellite. These images are prone to numerous attacks on the communication channel, such as statistical attacks and differential attacks. These attacks are overcome by adding the security layer to the multimedia data using steganography, encryption, and hybrid methods based on it. The steganography methods hide the secret data in the cover media, whereas encryption methods scramble the secret data using a random key. Further, in the hybrid methods, the secret data is encrypted and then hidden in the cover image. Hence, in this research, these security methods are investigated to find out the challenges of the existing algorithms and enhance them using the bio-inspired algorithms. In the literature, numerous authors used the bioinspired algorithm to enhance the security methods. However, in the literature, several bio-inspired algorithms are available, and selection of the most optimum algorithm is a challenging task. In this research, the most optimal bio-inspired algorithms, namely, Egyptian Vulture Optimization (EVO), Green Heron Optimization (GHO), and Black Widow Optimization (BWO), are chosen due to their better exploration rate, minimum parameter tuning, and better convergence rate. Based on these algorithms, security methods in the fields of steganography, encryption, and hybrid approaches based on them are designed in this research. In this research work, two security methods are designed to enhance the security parameter known as imperceptibility for image steganography applications. In the first method, three parameters are determined before data hiding, such as optimal cover image index, block index, and secret data index, using the bio-inspired algorithms (EVO, GHO, and BWO), which were not claimed by other authors in the previous studies. On the other hand, in the second method, the secret data bits are matched with cover image LSB bits, and the matched index is determined. Thereafter, the matched index is hidden in the same cover image in the optimal way using the bio-inspired algorithm by finding the best starting index in the cover image and optimal secret data index. The bio-inspired algorithm searches the best indexes based on the objective function. In our work, the parameter mean square error (MSE) is taken as the objective function. Further, the benefit of the proposed image steganography method is that it is suitable for single- and multi-bit data embedding. The simulation evaluation of the image steganography method is done on the standard USC SIPI Image Database images. Further, for the evaluation purposes, several grayscale and color images are taken into consideration. Next, the evaluation is done based on subjective and objective analysis. In the subjective analysis, based on the visual quality, original cover images and their histograms are compared with the output images known as stego images. On the other hand, in the objective analysis, several performance parameters, namely, mean square error (MSE), root mean square error (RMSE), peak signal to noise ratio (PSNR), structure similarity index measure (SSIM), correlation coefficient (CC), entropy, university image quality (UIQ) index, image fidelity (IF), and normalized absolute error (NAE), are used to analyze the characteristics of the stego image with respect to the cover image. The result shows that the proposed image steganography method achieves high SSIM, CC, UIQI, IF near to one value, low MSE, RMSE, NAE near to zero value, and approximates similar entropy between cover and stego image as required in the steganography. Besides that, the proposed method achieves better PSNR without degrading the payload capacity as compared to existing methods. Two security methods are proposed next to overcome the statistical and differential attacks on the secret data for image encryption applications. In the first method, a random key of 512- bits is generated using the bio-inspired BWO algorithm for data encryption. The benefit of the BWO algorithm is that it searches for the best key among the n number of keys based on the objective function. Subsequently, this key is utilized in the image encryption method to perform substitution, permutation, and key scheduling steps. Besides that, the BWO mutation operation is performed in the permutation step. On the other hand, in the second method, the BWO algorithm searches the best parameter values of the chaotic logistic map for key generation based on the objective function. After that, an exclusive-OR operation is performed between the secret image pixel and the random key. Next, the encrypted matrix is randomly circularly shifted horizontally and vertically to achieve permutation. In both methods, entropy is taken as the objective function. The simulation evaluation is done in the standard USC SIPI image database. In the evaluation, several images are taken into consideration. The result shows that the encrypted image is found to be completely noisy from visual analysis, and its histograms are equally distributed. On the other hand, in the objective analysis, the proposed methods achieve high entropy (~7.999) close to the ideal value and also high number of pixels change rates (NPCR) (~99%) with a low correlation coefficient (near to zero value) and PSNR (near to 8-12 dB). Further, comparative analysis shows that the proposed method outperforms in terms of entropy and NPCR over the existing methods. Next, a privacy-preserving method is designed by hybridizing the image encryption and steganography methods. The novelty of the proposed privacy-preserving method is that the same evolutionary BWO algorithm is used for key generation, for secret data encryption, and for optimized data hiding. The benefit of the proposed method comes from the fact that the cover image plane chosen to hide the encrypted data is not fixed, as it is determined based on the pixel intensity value. Moreover, in the proposed method, only sensitive information about the user is encrypted. The visual analysis shows that the input and output images are similar, and the objective analysis shows that the cover plane and optimal starting pixel index are not static, achieving better PSNR (in the range of 54.1579-54.3132 dB), high CC, SSIM, IF, UIQI (near to 0.999 value), and similar entropy is obtained between input and output image. The proposed methods are useful for anyone who wants to communicate secret data in a more secure way.