APPLICATION OF PROBABILISTIC CARDS OF BRIGHTNESS GRADIENTS IN DIGITAL IMAGES FOR ADJUSTING PARAMETERS OF STEGANOGRAPHIC DATA INSERTION

Abstract

In the context of the accelerating pace of emergence and the expanding spectrum of new cyber threats, steganography remains an important tool for protecting sensitive information. Unlike cryptographic protection, it conceals the very fact of information transmission in digital data arrays. The steady growth in the volume, speed, and range of information circulating in modern information systems necessitates the creation of appropriate data processing technologies. In this sense, the development of adaptive, low-resource steganographic algorithms that take into consideration the statistical properties of information carrier containers and content becomes particularly relevant. Their implementation aims to ensure a balance of «interests» between the controversial conditions, requirements, and properties inherent in the process of steganographic data embedding. Given the totality of the tasks at hand, resolving the specified contradictions is undoubtedly relevant and has a pronounced applied character. The research conducted simulation modeling of procedures for synthesizing probability cards of brightness differences for different types of images (under conditions of global and local - block statistics). During the modeling, the threshold values for «smoothing» and «similarity» (P_Z1 and P_Z2) and the dimensions of the N×N image Basic Blocks (BBs) were varied. The evaluation of the results was performed: - the PSNR metric; difference card analysis; the number and average length of the formed series of BBs. The modeling confirmed that the use of adaptive two-stage smoothing (P_Z1 and P_Z2) in the range of their values from 7 to 14 brightness gradations allows to significantly reduce the total amount of series of BBs and increase their average length. This improves the combinatorics of elements in the array series of BBs for the purpose of their subsequent multiplexing and reduces the computational complexity of the algorithm. According to the processing results, the key structural elements of the images are fully preserved, while all changes are localized in their low-information areas. The obtained PSNR values in the specified range of P_Z1 and P_Z2 parameter adjustments remain consistently high. Integrating probabilistic cards of brightness differences into the cycle preprocessing for source data ensures the possibility of express evaluation of the source data and allows for more efficient determining of the necessary statistical correlations in the interacting pair of «content↔container». Thanks to this, the initial conditions for forming arrays series of BBs and the combinatorics multiplexing of series parameter are improved. The considered approach ensures a balance between visual quality, speed, and computational efficiency of the investigated algorithm. This makes it a promising for implementation as part of mobile platforms and/or as a way to mitigate the effects of resource constraints in the hardware and software platforms used. The results obtained provide a basis for further automation of the selection of current processing parameters for concealed data, depending on the statistical properties of the «content↔container» interaction pair.