NDFT-based Audio Watermarking Scheme with High Security

Uniform discrete Fourier transform (DFT) has the drawback of public frequency points. On the other hand, nonuniform discrete Fourier transform (NDFT) could set up random sampling points in frequency domain as desired rather than fixed frequency points in DFT. That is DFT is the special condition of NDFT. This paper utilizes NDFT to instead uniform DFT in the algorithm of audio watermarking scheme with the purpose of providing the probability of hidden embedding positions. Moreover, to further improve the systematic security, in our method, we use coupled-chaotic sequence to randomly select the NDFT-domain frequency points used for embedding watermark, which overcomes the flaws brought by single chaotic map which is finite word length effect and vulnerable to repeated group attack. Good experimental results have shown the proposed scheme possesses higher systematic security than DFT-based watermarking scheme