Hiding data and detecting hidden data in raw video components using sift points

In this study, steganography and steganalysis, which are important sub-disciplines of data-hiding methods, have been emphasized and inferences have been made according to the results obtained by using different methods. Raw video components that were taken in real time have been used. These components are both spatial and temporal. Steganalysis studies were performed on real-time raw video components in which data were hidden using SIFT, a feature extraction algorithm. These studies are based both on pixel adjacency matrix in statistical/mathematical modeling and on the use of CNN detectors in deep learning. While data storage is performed in the carrier medium, non-repeating points are used from the highest quality 5 SIFT keypoints obtained separately for each frame of real-time raw video components. It has been confirmed that the SIFT keypoints obtained separately from the cover medias before and after the data-hiding transaction are exactly the same in the spatial domain and that there are no serious structural differences between them. Data-hiding transaction within these SIFT keypoints is achieved using the LSB method in the spatial domain. As a result of the data-hiding transaction made to the keypoints obtained by SIFT method, it has been observed that the secret message stored on the cover media is the same as the secret message obtained from the covered media. This study was not intended to find the hidden message as content or to disclose and/or destroy the hidden message as content. This study aimed to determine whether there is any hidden message content based on statistical/mathematical and deep learning. And the results have been evaluated.