一种动态更新失真代价的自适应k隐写算法

基于失真函数的自适应隐写技术在嵌入过程中，忽略了嵌入操作相互间的影响，隐写策略无法随载体统计特性的改变自适应地调节。考虑嵌入操作的交互影响，该文提出一种基于动态更新失真代价的k隐写算法。首先分析了中心像素与其邻域的相关性，理论证明了在4-邻域修改情况下中心像素的最优修改方式，进而提出了失真代价更新策略MDS(Modification Degree Strategy)；并结合该策略设计实现了一种自适应k隐写算法。实验表明，五元修改方式下算法UNIWARD-MDS(Pentary Version)在高嵌入率下(0.5~1.0 bpp)的抗SRM检测性优于S-UNIWARD(Pentary Version)，同时在抵抗maxSRMd2检测时不同嵌入率下均优于S-UNIWARD(Pentary Version)；三元修改方式下算法HILL-MDS和UNIWARD-MDS(Ternary Version)抗检测性能优于对应的自适应隐写算法HILL和S-UNIWARD(Ternary Version)。

Adaptive k Steganography Based on Dynamic Updating Distortion Cost

Adaptive steganography ignores the interactive impact introduced by the embedding operation during the embedding operation. Considering the cross impact of the embedding operation, an adaptive k steganography is put forward based on the dynamic distortion cost updating strategy Modification Degree Strategy (MDS). First, the analysis is conducted to prove the optimizing modification of the central pixel under the condition of the neighborhoods modifications. Then the MDS of updating the distortion cost is presented to adjust the distortion cost according to the modification of neighborhood. Finally, the steganography scheme is proposed using the MDS. The experimental result illustrates that the UNIWARD-MDS (Pentary Version) has a better performance than S-UNIWARD (Pentary Version) at the embedding rate 0.5~1.0 bpp when resisting the steganalysis SRM. Meanwhile the UNIWARD-MDS (Pentary Version) is better than the S-UNIWARD (Pentary Version) at the embedding rate 0.05~1.0 bpp when resisting the maxSRMd2 detection. The HILL-MDS and UNIWARD-MDS (Ternary Version) perform better than the corresponding schemes HILL and S-UNIWARD (Ternary Version).