Blind decoder for binary probabilistic traitor tracing codes

This paper presents a new decoder for probabilistic binary traitor tracing codes which is based on classical hypothesis testing and estimation theory. This new decoder is blind, in the sense of ignoring a priori the collusion attack it is facing. It basically performs a joint estimation of the collusion channel and the probability that each user takes part in the collusion. The experimental results shown in the paper have been obtained with the classical Tardos code, although the proposed decoder works with arbitrary probabilistic binary codes. Another remarkable advantage of this blind decoder is its ability to successfully cope with collusion channels far more general than the classical marking assumption, including channels that produce erasures and random decoding errors.