Secure Communication Using Steganography in Image Processing

Abstract

Research in information security and secrecy is becoming more and more important, as well as, demanding as the information is exponentially exploding. There has been a good bit of focus on cryptography, but with cryptanalysis and crypto attacks, researches have looked into the alternative means, like steganography. Steganography conceals the message into the cover file. In this dissertation we have focused on image steganography and we have tried to improve the key parameters of the system which are capacity, imperceptibility and robustness. In the first part of the thesis the presented work comprises of two contributions. In the first contribution the secret information is preprocessed by using latest devised right translated gray substitution box and BCH error correction codes. This process enhances the security of the information as data retrieval is impossible without the information of mapping rule applied and secret key. Differential embedding to the LSBs of optimized chaotically selected pixels through GA is another step to make detection difficult and avoid error to propagate. In second contribution we introduce an advance technique for embedding based on representing cover image pixels using a new 13-bit prime series representation resulting into 3-times increase in capacity. The payload has been reduced by applying 2D DCT to secret image and thresholding of the coefficients ensuring high imperceptibility. An innovative algorithm has also been proposed to ensure the uniform spread of message into the cover image by adjusting average separation between the chaotically selected pixels of cover image for embedding based on size of secret information and cover image In the second part of the thesis the presented work is an application of compressed sensing. The main advantage we have gained is the huge increase in security of the information along with the payload reduction. Utilizing compressed sensing made the system more secure as reconstruction of the data is impossible without knowing the measurement basis, as the generation process of measurement basis is random and requires huge computations to predict them. We have presented two problems. The first contribution focuses on combining compressive sensing and steganography to enhance security of image steganography. The secret image is encrypted and compressed using compressive sensing. The encrypted data is then embedded to randomly selected pixels of cover image using a secret chaotic key. Simulation results presented show the efficient recovery and reconstruction of secret image using reduced payload. In the second contribution we have focused on security and payload capacity enhancement of an image steganography system for an audio message by using compressed sensing theory. However, in order to utilize compressed sensing, the audio message is first converted to an equivalent grayscale image which is sparsified using 2D-DCT and thresholding. The sparsified image is further compressed using the proposed compressed sensing algorithm which enhances the security to a high level and also payload capacity improves significantly, without losing imperceptibility of the system. The compressed image is embedded in chaotically chosen pixels of the cover image. At receiver the compressed sensing reconstruction algorithm is used to reconstruct the grayscale image which is then converted back to the audio message. Presented results indicate that the proposed system is highly imperceptible, secure and robust against various image processing attacks. It is able to reconstruct secret audio message with high PSNR value.