Secured Digital Watermarking Using Neural Networks

Neural network has evolved as a technological breakthrough that has the potential to solve hard computational complexity. In the recent past, the neural network has played a significant role in solving problems in a multi-domain environment, i.e., security, AI, data science, image processing, etc. In security, digital watermarking is used to provide two major components of CIA, i.e., integrity and authentication. The conventional ways of utilizing watermarking have certain drawbacks which the neural network can fill. This chapter explored and discussed referred methods and approaches used for embedding the digital watermark based on neural networks. The machine learning models utilized for watermarking were demonstrated with the help of a taxonomy tree. It also includes qualitative analysis of different watermarking techniques based on neural networks. The study also includes several attack scenarios in the implementation of digital watermarking with their cons and pros. Finally, the performance analysis of image authentication techniques was compared using the parameter values of PSNR (peak signal-to-noise ratio), SSIM (structural similarity index measure), and NC (normalized correlation) to suggest a suitable method for neural network implementation in watermarking process.