Structural stegdetectors performance in case of side-informed message embedding

Counteraction against sensitive information leakage with usage of hidden (steganographic) communication channels is topical issue today. Development of advanced side-informed message embedding methods lead to significantly increasing the robustness of formed stego files, in particular stego images, to known steganalysis methods. Providing the high accuracy of stego image detection in such case requires obtaining additional information about precover (higher-quality representation of the cover), which does not available in most cases. For increasing the accuracy of stego image detection there was proposed to use the multilevel image model. Performance analysis of multiclass stegdetector, based on mentioned model, as well as advanced statistical stegdetectors was provided. It was shown that applying the proposed multilevel image model allows significantly decrease the detection time in comparison with known statistical stegdetector by preserving the fixed accuracy of stego image detection.