چكيده به لاتين
Due to some inherent features of image such as high correlation among pixels, image encryption is somehow different from text encryption; so, traditional algorithms are not suitable for image encryption.During the last few years, several image encryption algorithms have been proposed. Unfortunately, because of weakness in the keystream generation, poor diffusion function, insensitivity to changes in plaintext and keystream, many of the schemes are insecure against various conventional attacks, especially on chosen-plaintext chosen-ciphertext attack. Therefore, in this thesis, we tried to study the strengths and weaknesses of algorithms with permutation and succession structures based on chaos, neural network, fuzzy integral and frequency transformations. By applying chosen-plaintext chosen-ciphertext attack which is described in the thesis, we show that the cryptosystems are not robust to resist attacks. Both mathematical analysis and experimental results confirm the feasibility of these attack.
Moreover, several pseudo-random number generators were introduced based on chaos, hash function, 2D logistic and coupling quantum chaotic map, choquet fuzzy integral, and a new chaos function. They were used to design algorithms for encrypting grayl-level and colour imagein in time and transform domains. The algorithms employ the image data in order to increase the resistance of the cryptosystems against differential attacks.
The experimental results reveal that the new image encryption algorithms have the advantages of large key space (2128), low Peak Signal-to-Noise Ratio, high mean square error, high entropy, and high sensitivity (NPCR > 99.6%, UACI > 33.4%). Also, the distribution of gray level values of the encrypted image has a semi-random behavior. Also, it is shown that this algorithm yields better security performance in comparison to the results obtained from other algorithms. Overall, it seems that the proposed algorithm can be a good candidate for image encryption.
