An Efficient Key Generation Method in Audio Zero-Watermarking

This paper proposes modification of previous audio zero-watermarking, proposed by Chen et. al [1], in terms of its watermark-key generation. We investigate that the size of previous watermark detection key is around two or three times larger than its watermark data. In order to yield an efficient key rather than generating three secret keys, we will exploit data indices, instead of field indices, to store the key. Compared to previous method, our watermark key's size is relatively the same as its watermark data. Despite of having smaller key size, the experiment demonstrates that our scheme has maximum imperceptibility level, indicated by SNR value, and can resist to several non-malicious attacks, such as echo addition, band pass filtering and re-sampling. In addition, we also address the invertibility and quasi-invertibility of our scheme for resolving rightful ownership, and present attack which can cause confusion to rightful claim. We list the objective of attacks, abilities of attacker, and a scenario to evaluate its security. According to the experimental result, our proposed scheme still remains several limitations, regardless of certification authority existence.