Deep learning approach and cover image transportation: a multi-security adaptive image steganography scheme

In this era of information security and communication, a major priority is the achievement of a robust and secure steganography system when thinking about information concealment. The development of such an information-hiding scheme demands that the scheme be able to hide a secret message within the cover media. The most vexing issues in existing steganography protocols are imperceptibility, security, and capacity, and researchers have frequently emphasized a trade-off between these issues. Scholars have consistently ignored the balance between security and payload because resolving one problem has been shown to have an impact on the other, and vice versa. To overcome these problems, an effective method known as the Conventional Neural Network based Edge Detection Method (CNN-EDM) has been presented for image steganography in this study. The CNN-EDM is used to improve the contributions of the proposed scheme. Four main stages were used to achieve the objectives in this research, beginning with the cover image and secret image preparation, followed by embedding, and culminating in extraction. The last stage is the evaluation stage, which employs several evaluations to benchmark the obtained results. A standard database from the Signal and Image Processing Institute (SIPI) containing color and grayscale images with 512 x 512 pixels was utilized in this study. Different parameters were used to test the performance of the suggested scheme based on security and imperceptibility (image quality). The image quality was evaluated using three important metrics: histogram analysis, peak signal-to-noise ratio (PSNR), and structural similarity index (SSIM). Furthermore, two metrics were used to evaluate the security properties of the proposed system: the Human Visual System (HVS) and Chi-square (X2) attacks. The evaluations showed that the proposed scheme can enhance the capacity, invisibility, and security features and address the already existing problems in this domain.