Security and robustness constraints for spread-spectrum Tardos fingerprinting

This paper presents a practical analysis of the impact of robustness and security on Tardos' collusion-secure fingerprinting codes using spread-spectrum watermarking modulations. In this framework, we assume that the coalition has to face an embedding scheme of given security level and consequently has to suffer a probability of wrongly estimating their embedded symbols. We recall the Worst Case Attack associated to this probability, e.g. the optimal attack which minimises the mutual information between the sequence of a colluder and the pirated one. For a given achievable rate of the Tardos' fingerprinting model, we compare the Improved Spread-Spectrum embedding versus a new secure embedding (called p-Circular Watermarking) considering the AWGN channel. We show that secure embeddings are more immune to decoding errors than non-secure ones while keeping the same fingerprinting capacity.