Design and Analysis of Wavelet based Steganography Algorithms for JPEG2000 Images

Abstract

Abstract Widespread use of multimedia data in commercial, scientific and defense applications has triggered new challenges for managing digital assets. Easy availability of digital editing tools are generating threats to the digital contents. All these necessitate security of digital data. There are many approaches which can be used for security of data. Information hiding is one of the approaches which is used in security of digital data. Further steganography is one of the techniques of information hiding that can be used for covert communication. The objective of the steganography is to hide the presence of secret data in any digital content from unauthorized users. It provides a better security for secret data in such a way that it is difficult for an eavesdropper to detect something is there. There are many different carriers which can be used in steganography but the most popular is digital images, which are widely utilized over the public networks. The important aspects of a steganography system are: embedding capacity, imperceptibility and un-detectability. Embedding capacity is the amount of secret data that can be embedded in a cover object, imperceptibility implies that the cover and its stego images are perceptually indistinguishable and un-detectability means attacker won’t able to detect that something is hidden inside the image. However the embedding capacity is mostly limited by the size and redundancy present in cover images. Also there is a tradeoff between embedding capacity and visual quality of stego image. However, most of the existing steganography techniques exhibit high distortion at low embedding capacity . Therefore, increasing embedding capacity and enhancing stego image quality are still challenging task, and this is our research objective. In this thesis, steganography algorithms are developed for spatial, frequency and compressed domain, following the footpaths of the prior researchers. The presented work is an effort to provide efficient algorithms having high embedding capacity while maintaining the characteristics of carrier iv image after embedding, so that the attacker won’t able to detect the presence of hidden secret data. Stego image produced by proposed algorithms preserve the statistical properties of an image after embedding the secret data. Steganography algorithms for uncompressed images have been proposed in both wavelet domain and spatial domain where we utilize the largest and smallest pixels of blocks of cover image as well as its wavelet subbands. Multilevel embedding approach is merged with spatial domain algorithm to enhance the embedding capacity. Spatial domain provides better results than wavelet domain algorithm in both embedding capacity and imperceptibility. For lossy JPEG2000 compressed images SVD and DWT based steganography algorithm is proposed at different bit rates. GA is applied to optimize the SF to maintain the visual quality of stego images produced by the proposed algorithm, so that trade-off between embedding capacity and visual quality is maintained. In the available literature, histogram is not utilized in steganography algorithms for compressed images. Histogram shifting based steganography algorithm applicable to lossless for JPEG2000 compressed images is proposed. The embedding of secret data is performed in the peak wavelet coeffi- cients of the cover image. To reduce the distortion between stego and cover images, OPAP is applied on stego image. In literature, none of the previous algorithms have used the Bit Plane Coding information prior to embedding secret data. In proposed work, we have utilized this information to propose steganography algorithms for JPEG2000 lossy compressed images. Further OPAP, EMD with MBNS and modified EMD are utilized in the proposed algorithms. These algorithms are extended to JPEG2000 videos, as these videos do not have motion compensation feature. In OPAP based algorithm, secret data bits are directly embedded into bit planes of significant wavelet coefficients and then OPAP concept is applied to enhance quality of stego images. Whereas in EMD with MBNS based algorithm, the secret data is converted into a series of symbols by using MBNS. These secret data bits are than embedded into bit planes of significant quantized wavelet coefficients using EMD concept. On the other hand, in modified EMD based algorithm, the secret data bits are embedded into significant quantized wavelet coefficients using modified EMD approach. Experimental results show that modified EMD v based algorithm is better than OPAP and EMD with MBNS based steganography algorithms. Experimental results reveal that preserving the image statistics using the proposed algorithm improves the un-detectability against the attack. The main contribution of proposed algorithms is that they can be used for information security and confidential communication.