抵抗帧间共谋的自适应量化嵌入隐写方法

Adaptive quantization embedding steganography method against inter-frame collusion

由于视频无法抵抗帧间共谋攻击,使得水印隐蔽性和鲁棒性降幅明显,因此提出一种基于自适应量化嵌入机制的抵抗方法,用以克服嵌入水印后导致的失真累加效应。通过构建共谋模型、模拟攻击原理及分析噪声来源获得抵抗策略,即在自适应量化嵌入机制的基础上,利用人眼对光学高频区域不敏感的视觉冗余特性实现水印隐写。仿真对比实验结果表明:经施加噪声干扰,从视觉质量角度观察水印的隐蔽性未受影响,说明该方法对噪声不敏感;帧平均、帧切除的首末帧峰值信噪比平均降幅比例为20.24%,帧重构、帧裁剪不同比例的平均波动幅度为2.77%,明显优于最低有效位(LSB)、差值扩展(DE)算法等对比方法的首末帧平均降幅比例41.17%和27.89%,平均波动幅度的13.74%和7.70%,表明所提方法具备较强收敛及调节帧间冗余的作用,使共谋者无法利用任何的帧间信息冗余来消除水印,进而无法实施有效的共谋攻击。

Concealment and robustness of the watermark in video is sensitive to resist inter-frame collusion attacks. So,for overcoming the distortion accumulation effect brought by embedded watermarks,a watermark method based on adaptive quantization embedding mechanism to resist inter-video collusion attacks was proposed. A strategy to resist collusion attacks was obtained by constructing a collusion model,simulating attack principles,and analyzing source of noise,i. e.,taking advantages of adaptive quantization mechanism and the insensitivity of human eyes to optical high-frequency regions to achieve steganography of watermark information. Simulation comparison experimental results show that concealment of the watermark is not affected by the visual perspective after noise interference is applied,which indicates that the encrypted video frame is not sensitive to noise. Frame average and frame cut of the first and last frames decrease 20. 24% evenly,and frame reconstruction/frame cropping fluctuation ranges are 2. 77% evenly,which is significantly better than the comparison methods such as Least Significant Bit(LSB)and Difference Expansion(DE)algorithm,whose reduction rates are 41. 17%and 27. 89% evenly and fluctuation range is 13. 74% and 7. 70% evenly. It indicates that the proposed method has effect of converging and adjusting redundancy between frames,so that the conspirators cannot use any information redundancy to eliminate the watermark,and thus cannot implement an effective collusion attack.