Combining Tardos fingerprinting codes and fingercasting

Forensic tracking faces new challenges when employed in mass-scale electronic content distribution. In order to avoid a high load at the server, the watermark embedding process should be shifted from the secure server to the client side, where (1) the security of the watermark secrets must be ensured, and (2) collusion-resistance against a reasonably sized coalition of malicious users needs to be guaranteed. The combination of secure content broadcasting, secure embedding and collusion tolerance aspects has been recently addressed and termed as Fingercasting. However, the proposed solution does not apply a special collusion-resistant code, but derives a limited resistance against collusion attacks from the underlying spread spectrum watermark. In this paper, we make the first step towards tackling this problem: we propose a construction that provides collusion-resistance against a large coalition in a secure watermark embedding setting. In particular, we propose to incorporate a variant of the collusion resistant random code of Tardos, currently the code with best asymptotic behavior, into a Fingercasting framework. Through statistical analysis we show that the combination is feasible for a small subset of possible Fingercasting system parameters.