Using the qualitative characteristics of an image for comprehensive steganalysis
Yaroslav L. Grachev, V. G. Sidorenko · Dependability · 2025
Aim. The problem of image steganalysis is especially relevant given the use of steganographical concealment in graphic files for delivering malicious code and information as part of cyber attacks. That requires improvements to the existing methods of detecting steganographically embedded information. One method is to use a comprehensive steganalysis technique that involves concluding on the detection of embedded information based on the findings of a group of steganalysis methods, as well as auxiliary calculations. Methods. It is proposed improving the accuracy of hidden information detection by using qualitative image estimation. The paper demonstrates the relationship between the estimates and the increased rate of steganalysis errors. The method of comprehensive steganalysis that involves accounting for the qualitative characteristics of images allows improving the accuracy of estimation by reducing the rate of false positives. The paper uses statistical methods for calculating the qualitative characteristics of images, Spearman correlation, and machine learning. Results. A software package has been developed that integrates elements of the comprehensive steganalysis method described in the paper that includes both a group of steganalysis methods, and a set of evaluated qualitative characteristics of an image. The author evaluates the relationship between the qualitative characteristics of an image and the steganalysis errors in the case of empty containers. Test samples have been defined and machine learning models have been built that generate a conclusion as regards the detection of hidden information in an image. Conclusion. The proposed method enables improved accuracy of hidden information detection, while taking into account the estimates of the qualitative characteristics of an image as part of steganalysis, which is confirmed experimentally.