High-Security Video Steganography Through Pallier Encryption and Pixel Value Differencing In Dynamic Objects

In the era of digital communication, safeguarding personal and confidential information against unauthorized access has become a growing concern. Conventional video steganography methods face several limitations, including low data-hiding capacity, poor visual quality, weak robustness under attacks and inadequate integration of cryptographic mechanisms. To address these gaps, this study proposes a high-security video steganography framework that securely embeds a secret image within the moving objects of video frames. The primary objectives of the proposed work are to enhance confidentiality through Paillier cryptosystem-based encryption, ensure accurate detection and tracking of dynamic objects using Faster R-CNN and the Unscented Kalman Filter (UKF) and improve embedding efficiency and imperceptibility using the Discrete Wavelet Transform (DWT) combined with Pixel Value Differencing (PVD) in high-frequency regions. The encrypted secret image is embedded into dynamically detected objects, ensuring that the hidden data remains visually imperceptible while maintaining the structural integrity of the video content. During extraction, the reverse process retrieves and decrypts the concealed image using the same Paillier key, ensuring end-to-end data confidentiality. Experimental results demonstrate that the proposed method achieves superior performance compared to existing techniques such as LWT, LSB and LSBMR, with a Peak Signal-to-Noise Ratio (PSNR) of 38.51 dB, Normalized Correlation (NC) ≈ 0.999 and low Bit Error Rate (BER = 0.696), while maintaining strong resistance to attacks like cropping, filtering and noise addition. Implemented in Python, the proposed method ensures secure, high-quality and imperceptible data hiding, offering significant potential for applications in secure multimedia communication, surveillance and digital forensics.