Research On Selective Image Encryption Based On Coding

The development of modern technology always makes a huge variation into workand daily life of human beings. It only takes half of a century to enter into the Internetage from the computer-established age. Multimedia enriches the way of gettinginformation tremendously, among which image is the uppermost way to demonstratedifferent kind of information. Thousands of images are being transmitted on theInternet at any moment, the security and confidentiality of images must be protectedstrictly. However, since the structure of image data is special and owns its uniquecharacteristics, e.g. containing large volume of data, it is not suitable to encrypt imagesusing traditional cryptology directly because of its low efficiency. Therefore it meanssignificant to explore novel method to encrypt image data in the Internet environment.Image compression is the fundamental guarantee for storing and transmitting imagedata properly. The process of compression has its inborn superiority so that theencryption operation can be embedded into it seamlessly. The effect of compressionnot only boosts encryption efficiency but also spreads the encryption to other part ofcompressed data; furthermore, joint compression and encryption can beformat-compatible. Selective encryption is another hot research direction. Byselectively encrypting a tiny portion of image data that is critical for visualization, itcan reduce the encrypted data volume substantially; and at the same time it can gain agood security.In this disseration, by taking both into consideration the ideas of selectiveencryption and joint compression and encryption, we have made several contributions,which are listed as follows:(1) We focus on set partitioning in hierarchical trees (SPIHT) algorithm, which isan excellent compression encoding algorithm, and then propose a joint SPIHTcompression and selective encryption (JCSE-SPIHT) algorithm. The algorithmresolves some deficiencies and limitations of several proposed methods in existingliteratures, such as there may exist chances for attackers to perform cryptanalysis, oramong them some methods might be chosen-plaintext attacked facilely. InJCSE-SPIHT, only a tiny portion of crucial data is fast scrambled during the codingprocess. The time complexity of encryption can be reduced fromO (n2)to O (n)byusing the FRI procedure to accelerate the encryption operation, and also the algorithm is immune to chosen-plaintext attack by generating plaintext-dependent key streamwith the help of cryptographically secure pseudo random number generator (PRNG).The encryption effect is excellent, as well as it is format-compliant. The intensiveexperiments show that the encrypted data volume can be restricted in1%-4%, and it isproven that the algorithm possesses a good security and high encrypting speed.(2) We propose an image degradative encryption based on SPIHT (DeSPIHT). Itis different with the confidential image encryption because degradative encryption onlyencrypts the details of an image but reserves a recognizable outline for preview. Theoutput bits of SPIHT can be divided into several types and the algorithm encrypts theBL IS-O-sgnbits. The experimental results show that the average encrypted data volume isless than10%; furthermore the argument r can rebalance the security andcomputational overhead. The experimental results and security analysis demonstratethat the proposed algorithm has good security strength. It can be used to search imagesin unauthorized situation or applied in cloud computing. It’s a new efficient way toprotect SPIHT compressed images.(3) We make an induction and reorganization of the basic theory and somerelevant research achievements of image encryption and image compression. Inaddition, four image quality evaluation metrics, PSNR, MSSIM, LSS and ESS, areintroduced, which is later used to evaluate the security of proposed encryptionalgorithm. They take the luminance similarity, structure similarity and the edgesimilarity into consideration for evaluating the encryption effect jointly. Hence we usethese metrics to evluate encryption effect in our proposed algorithms.In the end of this dissertation, conclusions, as well as the discussion of theapplication and promotional work in relevant domain in this paper, are given. Futurework is also illustrated.