A novel algorithm for colour image steganography using a new intelligent technique based on three phases.

A three-phase intelligent technique has been constructed to improve the data-hiding algorithm in colour images with imperceptibility. The first phase of the learning system (LS) has been applied in advance, whereas the other phases have been applied after the hiding process. The first phase has been constructed to estimate the number of bits to be hidden at each pixel (NBH); this phase is based on adaptive neural networks with an adaptive genetic algorithm using upwind adaptive relaxation (LS ANN _ AGAUpAR1 ). The LS of the second phase (LS ANN _ AGAUpAR2 ) has been introduced as a detector to check the performance of the proposed steganographic algorithm by creating a rich images model from available cover and stego images. The LS of the last phase (LS CANN _ AGAUpAR3 ) has been implemented through three steps, and it is based on a concurrent approach to improve the stego image and defend against attacks. The adaptive image filtering and adaptive image segmentation algorithms have been introduced to randomly hide a compressed and encrypted secret message into a cover image. The NBH for each pixel has been estimated cautiously using 32 principle situations (PS) with their 6 branch situations (BS). These situations have been worked through seven layers of security to augment protection from attacks. In this paper, hiding algorithms have been produced to fight three types of attacks: visual, structural, and statistical attacks. The simulation results have been discussed and compared with new literature using data hiding algorithms for colour images. The results of the proposed algorithm can efficiently embed a large quantity of data, up to 12 bpp (bits per pixel), with better image quality. [ABSTRACT FROM AUTHOR]