Attacking Traitor Tracing Schemes Using History Recording and Abrupt Decoders

In ACM-DRM 2001, Kiayias and Yung [19] introduced a classification of pirate decoders in the context of traitor tracing that put forth traceability against history recording and abrupt pirate decoders. History recording pirate decoders are able to maintain state during the traitor tracing process while abrupt decoders can terminate the tracing operation at will based on the value of a "React" predicate. Beyond this original work, subsequently a number of other works tackled the problem of designing traitor tracing schemes against such decoders but with very limited success. In this work, we present a new attack that can be mounted by abrupt and resettable decoders. Our attack defeats the tracing algorithm that was presented in [19] (which would continue to hold only for deterministic pirate decoders). Thus we show that contrary to what is currently believed there do not exist any known tracing procedures against abrupt decoders for general plaintext distributions. We also describe an attack that can be mounted by history recording (and available) decoders.