Total disclosure of the embedding and detection algorithms for a secure digital watermarking scheme for audio

This paper discusses the modification of a robust digital audio water-marking scheme to allow the disclosure of the embedding and detection algorithms. The chosen scheme uses MPEG 1 Layer 3 compression to determine the position of the mark bits in the frequency domain. The marking positions would be exposed if the original embedding algorithm was disclosed. In fact, it is shown that even if an attacker did not know the exact tuning parameters used for embedding, he or she could still produce an approximate superset of the marking frequencies from only a marked copy and successfully attack the file. To avoid this problem, a secret key is introduced in the embedding and detection processes. The secret key includes the seed of a pseudo-random number generator which is used to compute the exact marking positions. The modification is then analysed in terms of capacity, imperceptibility, robustness and security. The experiments show that the modified scheme preserves most of the properties of the original one, such as robustness against MP3 compression for the most frequently used bit rates, and does introduce additional security as the mark is more difficult to erase when the embedding and detection algorithms are disclosed.