A secure steganography technique for digital forensics in H 264 AVC videos

Abstract

In the presented work, one of the transform domain based techniques has been utilized to provide better robustness and security in addition to high imperceptibility and capacity. The presented work scrutinizes the current state-of-the-art techniques in the field of video steganography including video specific techniques in addition to spatial and transform- domain based techniques. The objective of this thesis is to present video steganography technique with suitable trade-off between the three basic requirements with high security for covert communication. To fulfil this objective, in this thesis, two techniques have been proposed by utilizing H.264/AVC video format for steganography which is the most common video standard at present. The proposed technique II- SRAT employed Discrete Wavelet Transform (DWT) on the Region of Interest (ROI) based on multiple moving objects tracking. After tracking multiple objects, each object is embedded with different secret images to improve the capacity. The proposed scheme is tested on numerous video sequences by evaluating both subjective and objective metrics. Different metrics employed for objective evaluation involve Mean Square Error (MSE), PSNR, SSIM, BER and Normalized Cross Correlation (NCC). The experimental results demonstrated the effectiveness of the proposed techniques against steganalysis with only maximum 56.6% average detection accuracy in case of IRT and 53.25% for SRAT. Though both the techniques performed well against steganalysis, however, the results conclude that SRAT performed better in terms of security with less detection accuracy as compared to IRT which means SRAT is more reliable for covert communication. The presented thesis work can be efficiently and effectively used for a secure communication using H.264/AVC videos for digital forensics. Finally, some future recommendations are also given based on the analysis and experiments done to continue research in this area. newline