Detection Types of Steganographic Attacks Based on Semi-Fragile Watermarks Using Machine Learning

This article discusses how to detected types of steganographic attacks on digital images using semi-fragile watermarks using machine learning techniques. This article discusses different methods of introducing watermarks: into the spatial domain and the transformation domain. The attacks in this case are filters that can be applied to images: blurring, adding noise, sharpness, contrast, color balance. Methods have been identified that allow you to determine with higher accuracy the specific type of attack on the embedded digital watermark in the image. The peculiarity of this method is that the original image does not need to be used to determine the attack (blind method [1]).