Countermeasure to non-linear collusion attacks on spread spectrum fingerprinting

In the spread spectrum (SS) fingerprinting technique, the averaging attack has been supposed to be the most damaging attack. It is because a number of nonlinear collusion strategies is supposed to be well approximated by averaging collusion plus additive noise. In this study, we intensively evaluate the conventional assumption, and reveal its drawbacks when a SS sequence is embedded in a digital image. Furthermore, we propose a simple and effective method to improve the drawbacks by introducing a spreading operation at the embedding. Our experimental results reveal that the damages caused by non-linear collusion strategies can be reduced in such a way that the host signal sampled from an image is kept secret.