Akademik Veri Yönetim Sistemi
3rd International Conference on Applied Engineering and Natural Sciences, Konya, Türkiye, 20 - 23 Temmuz 2022, ss.1976-1977
In data hiding, the new carrier is sent to the
receiver after certain messages are hidden in a carrier object such as a image,
video, sound or text. The data hiding, which is divided into two as
steganography and watermarking, has various methods for processes such as
secret communication, content protection and authentication. In data hiding
methods used for secret communication, the carrier object which is contains the
secret message is very similar to the original object and does not arouse any
suspicion to third parties. In classical data hiding, there are many up-to-date
methods developed for images and videos that improve parameters such as data hiding
capacity, invisibility, and resistance to attacks. fields One of them is the
development of new data hiding methods using deep learning approaches. In this
study, first of all, the use of deep learning models for data hiding and the
new approaches and terms that emerged with the use of deep learning for data
hiding were evaluated. Then, different architectural structures such as
convolutional neural networks (CNN) and generative adversarial networks (GAN)
and their usage in data hiding are discussed. Finally, within the scope of this
study, the use of classical data hiding quality criteria for deep
learning-based data hiding methods and the possibilities of using some quality
metrics used for deep learning architectures in data hiding were evaluated. In
addition, the advantages of data hiding studies with deep learning compared to
classical methods, how they provide solutions to today's information security
problems, and which problems they can be applied to in the future are
emphasized.