Features of Applying Pretrained Convolutional Neural Networks to Graphic Image Steganalysis
S. N. Tereshchenko, Artem A. Perov, Alexander Leonidovich Osipov · Optoelectronics Instrumentation and Data Processing · 2021
Abstract This paper investigates the usage of convolutional neural networks in order to analyze graphic images for presence of data introduced by steganographic methods. It is shown that a deep convolutional neural network can be trained to classify the presence of hidden data in graphic images, achieving an accuracy of 0.928 according to the weighted AUC metric. The hypothesis of the effectiveness of applying the concept of transfer learning in the sphere of steganography is tested. The efficiency of the proposed technology is demonstrated on a large number of experiments.