Collusion Attack Resistant SVD Based Secured Audio Watermarking Scheme with Chaotic Encryption

In this paper, a blind, robust, and inaudible audio watermarking algorithm based on Singular Value Decomposition (SVD) is proposed. The algorithm produce multiple copies with different fingerprints/watermarks on the same audio using different SVD values making it robust to collusion attack. The proposed algorithm uses modified mean value and quantization index modulation technique of the singular values for watermark embedding. Instead of using the SVD matrix or the SVD matrix in cascade with the other transform directly for watermark embedding, the proposed schemes uses Baker's Chaotic mapping for encryption of the matrix without the cost of imperceptibility. The proposed method thus produces security as well as collusion attack resistance over all the SVD based methods where the location for the watermark embedding is known as the diagonal elements of the SVD matrix produced from the audio directly. Experimental results presented in this paper demonstrate the effectiveness of the proposed algorithm against signal processing attacks with good imperceptibility.