A Fragile Watermarking Method for Content-Authentication of H.264-AVC Video

Abstract

This work proposes a blind fragile GOP-based watermarking technique that ensures the H.264-AVC video's content authenticity. In the H.264 video, the self-authentication code is created by securely hashing a number of intra prediction modes of I-frames in addition to motion vectors of P-frames and B-frames. The resulting authentication code is subsequently injected into the IPCM-Least block's Significant Bits (LSB) of some I-frames' luma and chroma pixel samples. To manage and regulate authentication data and watermark processes (embedding, detection, and verification), a key based on video footage and a secret key has been created. Intra prediction modes and motion vectors from the bitstream can be partially decoded to detect and confirm the hidden watermark data without the need for the original movie or complete video decoding. Several experiments were conducted to assess the sensitivity of the proposed method against signal processing, rate control and transcoding, conventional GOP-based and Frame-based attacks. Experimental simulations demonstrate that the implanted fragile watermark is sensitive to several low to harmful and content-preserving attacks. With only a modest bit-rate increase and hardly any perceptual quality deterioration, a large payload capacity is achieved. Experimental results show that the proposed method is suburb in term of imperceptibility, and lower effect on bit-rate. Applications like tamper-proofing, content-authentication, and covert communication are ideally suited for this technique. Based on achieved results, the devised method is capable of detecting any form of spatial and/or temporal manipulations which make it ideal for real-time applications.