Public traceability in traitor tracing schemes

Traitor tracing schemes are of major importance for secure distribution of digital content. They indeed aim at protecting content providers from colluding users to build pirate decoders. If such a collusion happens, at least one member of the latter collusion will be detected. Several solutions have already been proposed in the literature, but the most important problem to solve remains having a very good ciphertext/plaintext rate. At Eurocrypt '02, Kiayias and Yung proposed the first scheme with such a constant rate, but still not optimal. In this paper, granted bilinear maps, we manage to improve it, and get an "almost" optimal scheme, since this rate is asymptotically 1. Furthermore, we introduce a new feature, the "public traceability", which means that the center can delegate the tracing capability to any "untrusted" person. This is not the first use of bilinear maps for traitor tracing applications, but among the previous proposals, only one has remained unbroken: we present an attack by producing an anonymous pirate decoder. We furthermore explain the flaw in their security analysis. For our scheme, we provide a complete proof, based on new computational assumptions, related to the bilinear Diffie-Hellman ones, in the standard model.