改进的大容量多分辨率LSB音频隐写算法

算法将载体音频信号用可完全重构的滤波器组进行小波包分解,根据各子带的边际信号失真敏感度将其分组,然后用改进的LSB算法以不同的分辨率将秘密信息嵌入到选定子带的小波系数中。实验结果表明,在信噪比相同的情况下,该算法的隐写容量比传统时域LSB算法高160kbps～200kbps,同时可很好地保持不可感知性,并具有一定的不可检测性,能抵抗一类针对时域LSB算法的隐写分析算法。

A Modified Multi-resolutional LSB Insertion Algorithm for High Capacity Audio Steganography

This work presents a novel modification to the standard LSB (Least Significant Bit) algorithm that can obtain very high hidden data rate.The proposed algorithm uses the perfect reconstruction filterbanks to decompose the cover audio signal into 32 subbands,sorts and groups the subbands according to their marginal signal distortion,and then embeds the secret information inside the wavelet pack domain by modifying the LSB values of the corresponding wavelet coefficients with multi-resolution.For the same SNR(Signal to Noise Ratio) values,the proposed algorithm can outperform the classical LSB algorithm by 160kbps~200kbps of hidden data.Subjective experiments show that the proposed wavelet pack domain information hiding scheme is acoustically more transparent as well. Furthermore,it can resist a kind of steganalysis method proposed to detect LSB steganography.