Joint image compression and encryption schemes

Abstract

Thanks to the rapid growth of computer networks and information technology, the number of multimedia applications related to images is rapidly increasing, thus the security issue of image storage and transmission becomes much more essential. Many encryption algorithms have been proposed to ensure images' safety. Among these encryption schemes, strategies of integrating image compression process with encryption algorithms have received much attention, since compression is a must-do step for most images we see on the Internet. Among various compression standards, JPEG is the most commonly applied method for lossy compression of digital images. This thesis presents our contributions in realizing image encryption during the JPEG compression process. In this thesis, four different joint image compression and encryption schemes are proposed by introducing encryption techniques into the intermediate stages of JPEG. All the encryption algorithms are format compliant to JPEG. The first encryption scheme is realized by applying new order-8 orthogonal transforms for blocks' transformation, followed by block permutation on quantized 8×8 blocks. These new transforms are produced by embedding rotation angle of π to 8×8 DCT's flow-graph structure. Through controlling the number of rotation angles embedded, we can achieve the quality control of the final encrypted images. In the second encryption scheme, we improve the first scheme by allowing more rotation angles, not just π , to be embedded into the flow-graph of DCT, and the coding efficiency of these newly generated transforms is better than that of the transforms used in the first scheme. Additionally, we enhance the whole cryptosystem's diffusion property by applying plain-image-dependent key for encryption, and propose a new data hiding technique for concealing the encryption key. The encryption scheme is JPEG compression friendly, can achieve a good balance between the encryption power and the compression efficiency. In the third encryption scheme, we realize joint image compression and encryption by using 16×16 DCT, in which the encryption techniques are mainly conducted at JPEG's transformation stage and entropy coding stage. To enhance cryptosystem's diffusion property, we also establish a relationship between the secret key and the plain-image, like the second scheme. For the transformation stage encryption, we segment the plain-image into non-overlapping 16×16 blocks, and use order-16 DCT for transformation. Then we distribute all the coefficients of 16×16 transformed blocks into 8×8 blocks, and do block permutation and DC coefficients confusion. For the entropy coding stage encryption, we shuffle all Run/Size and Value (RSV) pairs of AC coefficients, and embed end-of-block (EOB) identifiers to keep format compliance. As for the fourth encryption scheme, we use adaptive block-size (8×8 size block and 16×16 size block) for JPEG image encryption. Given a plain-image, we first segment it into 8×8 size blocks or 16×16 size blocks, according to two different segmentation schemes. Then we transform these different size blocks using the corresponding size of DCT, followed by the 8×8 blocks permutation and same run value RSV pairs' shuffling operation. The fourth scheme has comparable compression efficiency as the second scheme, and can achieve better statistical attack defense ability than the second scheme.