Least significant bit steganography technique using modulus operation with pixel selection derived from K-map and Gaussian elimination

Abstract

As demand for information exchange across the network increases, so is the need for safe covert communication, which can be achieved using steganography. Many steganography techniques have emerged; however the performance of the techniques rely heavily on three major factors, which are the payload, imperceptibility and robustness. These elements are always in trade-off. In this research, a new steganography technique that emphasizes on high imperceptiblity with reasonable embedded capacity is presented. The proposed embedding approach leverages on the LSB substitution technique and neighbourhood operation to precisely determine how much changes is required for each target pixel of host image in terms of its gray scale. Thus, a 9x9 filter called Matrix Distribution Array (MDA) is introduced based on the Karnaugh Map grouping principle to generate nine possible 4-pixel groups for pixels selection. A modulus operation is then performed for each group to obtain a group residue (Res). An initial change of the target pixel value is calculated based on the difference between Res and secret information. Afterwards, the Gaussian Ellimination technique is then applied together with MDA on the change value to obtain a final figure of the change required. Finally, the target pixel is modified and rounded accordingly by subtracting its original gray scale with the change value. The embedding process is repeated until a stego-image is eventually produced, likewise, an extraction process is performed using a similar procedure but in a reverse manner. The experimental results show that the imperceptibility of the proposed method improved significantly by 8% to 16% when tested with fixed embedding capacity ranging from 6 kB to 116 kB as compared to the LSB substitution technique. The result also reveals that the embedding capacity improves up to 50% while maintaining reasonable Peak Signal-to-Noise-Ratio (PSNR) value between 35dB to 40dB.