Abstract
In order to fight the phenomenon of intellectual piracy in multimedia distribution platforms, the traitor tracing technique, called also active fingerprinting was proposed by researchers as a suitable solution. In fact, the media producer, before distributing any release of his content, should proceed by embedding in each copy a unique fingerprint to identify its owner or its purchaser and even to allow the tracing process in case of piracy attack. In the same context, some tracing approaches proposed in the literature have deserved a considerable attention. The main assumption in these approaches is that users having common characteristics have more probability to collude together. Hence, using a group-based fingerprint in the tracing process should enhance the traitors' detection rates. In this paper, the traitor tracing scheme we propose consists of two major contributions: (1) the generation step of the group-based fingerprint by clustering the users' identifiers, and (2) the two-level tracing process which is chosen adequately to be applied to the group-based structure of the fingerprint. In fact, in this strategy, the tracing levels is made in two successive steps: in the first one, guilty groups are accused using the Boneh Shaw code and in the second one, the retrieving back of traitorous users into these groups is made using the Tardos code. We evaluate the proposed tracing scheme against various types of collusion attacks: collusion into the same group and between different groups. Significant improvements using the proposed scheme are shown especially for two criteria: the detection rates, and the time taken by the accusation process to retrieve at least one colluder, confirmed by a comparison to the same scheme applied to non-clustered users' identifiers.