The Identifiable Parent Property guarantees, with probability 1, the identification of at least one of the traitors by the corresponding traitor tracing schemes, or, by IPP-codes. Unfortunately, for the case of binary codes the IPP property does not hold even in the case of only two traitors. A recent work has considered a natural generalization of IPP-codes for the binary case, where the identifiable parent property should hold with probability almost 1. It has been shown that almost t-IPP code...
The Identifiable Parent Property guarantees, with probability 1, the identification of at least one of the traitors by the corresponding traitor tracing schemes, or, by IPP-codes. Unfortunately, for the case of binary codes the IPP property does not hold even in the case of only two traitors. A recent work has considered a natural generalization of IPP-codes for the binary case, where the identifiable parent property should hold with probability almost 1. It has been shown that almost t-IPP codes of nonvanishing rate exist for the case t = 2. Surprisingly enough, collusion secure digital fingerprinting codes do not automatically possess this almost IPP property. In practice, this means that for a given forged fingerprint, say z, a user identified as guilty by the tracing algorithm can deny this claim since he will be able to present a coalition of users that can create the same z, but he does not belong to that coalition. In this paper, we study the case of t-almost IPP codes for t > 2.