Machine learning based weight optimized genetic algorithm for digital video watermarking technique
Hussein Z. Almngoshi, M. Vedaraj, V. P. Sriram, M. Davidson Kamala Dhas, V. Arunraj, Sudhakar Sengan, Pankaj Dadheech · Journal of Autonomous Intelligence · 2024
Video piracy is growing amid the popularity of online streaming services and storage solutions, primarily because it is easy to share media through the internet. Content creators now have a simpler time sharing audio and video materials across different platforms because of a boost in data traffic. This has caused issues about the security of multimedia content and the security of Intellectual Property Rights (IPR), both of which have become fundamental in the contemporary digital age. Digital Watermarking (DW) is a revolutionary technology that enables multimedia IPR by successfully hiding and securing intellectual accuracy from cyberattacks. As evidence of the relative simplicity with which it can be combined, DW is now recognised as the main point of studies addressing data verification and IPR security measures. Watermarks are hidden tags that DW uses to secretly integrate into video recordings in order to detect IPR crimes and authenticate the reliability of data. In order to improve data source verification, the present research provides the Least Significant Bit (LSB) approach to DVW, which increases the possibility of Mean Square Error (MSE). The study references a Genetic Algorithm (GA) that decreases the negative impacts of LSB by minimising MSE while boosting the Peak Signal-to-Noise Ratio (PSNR), a key measure of the quality of watermarking. The research paper also uses statistical techniques and experiments to demonstrate how it analyses the complexity of computation, precision, resources, speed, and endurance as metrics for performance. With PSNRs exceeding 45.19 dB without attacks, the technique shows robustness over the background noise, filtering, and video encoding. With findings from experiments demonstrating a 75% Normalised Cross-Correlation (NCC), 97.89% accuracy for training, and 96.78% validation accuracy, the proposed approach indicates higher accuracy than hiding and security.