MODELING AND EVALUATION OF THE COMPUTATIONAL COMPLEXITY OF THE MAIN PROCEDURES, THE INITIAL STAGES OF THE HYBRID STEGANOGRAPHIC ALGORITHM

Abstract

 In today's environment of increasing cyber threats, steganography plays a key role in ensuring data confidentiality by hiding it in digital data arrays (images, sound, file systems, etc.). Unlike cryptography, steganography hides the very fact of information transfer, which makes it indispensable for data protection in information and communication systems with limited resources. The need to create energy-efficient and adaptive algorithms that combine content robustness, high container quality, and low computational complexity emphasizes the relevance of the research conducted. The paper evaluates the computational complexity of input data preprocessing procedures and determines their impact on the efficiency of subsequent data encapsulation. As a basis for modeling, one of the possible variants of smoothing low-informative areas of the source images was selected. The following assessments were carried out: - performance by execution time for different types of images; PSNR metric; probability of color brightness difference and number of formed basic blocks (BBs) of images. During the modeling, the consequences of smoothing the textures of the source images were investigated in the conditions of varying smoothing matrix sizes and the value of the pixel brightness difference. The number of basic operations (arithmetic, logical, comparisons) for each block was calculated and the overall complexity of the corresponding stages of the algorithm was determined. The experimental evaluation was performed by measuring the execution time of the procedures, evaluating the image quality using the PSNR metric, calculating the probability of brightness difference, and counting the number of formed series of BBs. To evaluate the effectiveness of the investigated procedures, test images with different textural characteristics were used. The obtained results confirmed that the pre-smoothing stage improves the initial conditions for the formation of BBs series (compared to their absence). The introduction of the smoothing stage ensures noise reduction in low-information areas, which contributes to increased block homogeneity and expands the range of combinatorics of available encoding parameters (as a way to counteract attempts at unauthorized content extraction). Flexible settings for block size and brightness coarsening threshold values for image elements allow the smoothing process to be adapted to different image types and the current resource limitations of the hardware platforms used. The proposed approach to preprocessing data creates the necessary conditions for forming BBs series, improves content robustness, and maintains a low level of visual distortion. Adaptive settings for smoothing parameters allow the algorithm to be used effectively on platforms with limited resources.