Security analysis of digital image watermarking using deep learning inspired LSB and chaotic S-Box in cyber security

Abstract

Information Technology (IT) has experienced remarkable advancements, this revolutionized the way of living, working and talking. One most momentous innovation is the social media communication and data sharing. The increasing prevalence of vast data sharing in the modern digital landscape has raised significant concerns regarding data privacy. Cryptography and image watermarking are essential components in the domain of security, specifically in providing services for multimedia protection. These technologies play crucial roles in opposing the integrity, confidentiality, and ownership of digital multimedia assets such as images, videos, and audio files. This research combines Least Significant Bit (LSB) and Holistically Nested Edge (HED) with chaos-based Substitution Box (S-Box) to enhance the strength of digital data security. The proposed technique summarizes the benefits of suitable positions by HED and high capacity and robustness against image processing attacks by LSB and encryption services by chaotic S-Box. The performance metric of proposed technique is calculated on Lena, Baboon and Pepper images. whereas to analyze the robustness Bit Error Ratio (BER) is calculated against different attacks i.e., speckle, gaussian, salt and pepper noise and cropping attacks on different images. The outcome shows that proposed technique opposing most of the attacks and legitimize it is effectiveness. The average value of Peak Signal to Noise Ratio (PSNR) and Structure Similarity Index (SSIM) are 53.15 dB and 0.9 which indicates invisibility of watermark signal in carrier image. The average value of Normalized Correlation (NC) is 0.9 which indicates proposed technique has strong robustness.