Image Generation Method Based on the Recoverable Byte Sequence using the Neural Networks

Abstract

Due to the rapid development of information technologies, the tasks of ensuring information integrity, safety and confidentiality, as well as the possibility of guaranteed confirmation of its source, are becoming more relevant than ever. One of the possible solutions to this problem could be the steganographic methods, which allow both hiding the fact of information transfer and imperceptibly adding the useful data. Scientific literature describes a large number of steganographic algorithms. However, only insignificant number of works is devoted to the data hiding methods using the neural networks, and even less are devoted to generating containers for them. A method for generating images based on the hidden information is proposed, which guarantees possibility of both hiding and subsequent extraction of information eliminating the need to select an appropriate container. As part of the method, an algorithm was developed that included description of the stages of input data preprocessing, transformation into the container image and extraction of the hidden information. Examples of the proposed method operation are provided. The method could serve both as a steganographic algorithm for hiding information and as the algorithm for adding information in the form of watermarks