This article presents a method for detecting steganographic changes in images using convolutional neural networks. The use of convolutional neural networks made it possible to automatically detect characteristic features and patterns in images, which ensured the high efficiency of the method. Dividing the image into small segments made it possible to search for changes with limited use of computing resources, in particular for the analysis of images on web resources, which can become part of a firewall to prevent steganographic changes. Experiments confirmed the high efficiency of the method in detecting steganographic changes, in particular when modifying images using various steganography methods. It is also found that the detection efficiency depends on the choice of database for training convolutional neural networks.