Chronological Gazelle Optimization with Deep Learning-Based pixel prediction for video steganography in H.264 video for defence applications

Abstract

With the recent progress in Information technology and the internet, there has been an increase in violations of information security and privacy, particularly in the defence domain. In this work, a secure video steganography method using deep learning-based pixel prediction in H.264 video is presented. The pixels in the keyframe on which the secret image is embedded are predicted using a Deep Convolutional Neural Network (DCNN) optimized by the Chronological Gazelle optimization algorithm (CGOA). Later, embedding is carried out using Haar Wavelet Transform (HWT). The security of the proposed steganography technique has been analysed by performing steganalysis using a Convolutional Neural Network (CNN). The efficiency of the approach is examined by considering evaluation measures, like structural Similarity Index measure (SSIM), Normalized Correlation (NC), Peak Signal to Noise Ratio (PSNR), Bit Error Rate (BER), and Mean Squared Error (MSE), and has attained values of 0.979, 0.974, 49.624, 4.655, and 0.790, revealing imperceptibility and robustness.