Comparative Analysis of Traditional vs. Deep Learning-Based Steganography Techniques

Abstract

Over the past few years, the deep learning technique as it pertains to steganography, the art of hiding a message in public information, has progressed significantly. This chapter compares the different techniques of deep learning-based steganography and conventional steganography. The emphasis in traditional methods of embedding information like the LSB substitution rests on the number of messages that can possibly be embedded and the level of resistance towards statistical detection. In contrast, the deep learning-based steganography approaches incorporate artificial neural networks which enhance the embedding process to optimize the data capacity, security, and functionality of many different types of media. Deep learning methods are, however, innovative and provide more security and robustness, although their disadvantage is the high processing power. For that deep learning steganography is an advancement. This is an emerging trend in the field where deep learning is likely to be incorporated in the coming days as an important enhancement of other sketched vision system.