On Mathematical Models for the Discordant Size Content Replacement Attack in Watermarked Audio Signals

An important aspect of any watermarking scheme is its robustness against attacks. For this, it is essential to know the behavior of the attack when handling a signal. The discordant size content replacement attack is a very severe attack against audio watermarking schemes. This attack replaces a set of samples of an audio signal with another collection of samples of different sizes. This study presents two mathematical models that describe the discordant size content replacement attack using different approaches: matrix representation and decimation, and interpolation. Each model includes content replacement attacks of equal, larger, and smaller sizes. Evaluations are carried out with different degrees of manipulation. The objective is to obtain a mathematical description that analyzes the behavior of the attack on an audio signal, and use the models to design self-restorable schemes of audio signals that are robust against this attack. The models showed proper functioning in their evaluations.