Development of a Subsystem for Steganalysis of Digital Images Based on a Convolutional Neural Network to Detect and Prevent Attacks Using Hidden Steganographic Channels

Abstract

This article presents a way to implement the subsystem for steganalysis of digital images circulating in the information system. This subsystem expands the functionality of existing intrusion detection / prevention systems in terms of detecting covert channels used in computer attacks. In the presented solution, a parametric model of a convolutional neural network is proposed and implemented to detect a payload in digital images, performed by a number of steg-nesting algorithms recognized in real attacks. A software implementation of a modular generator of a training sample (dataset) that supports these algorithms has been developed. An experimental assessment of the accuracy has been carried out