Group-oriented and Collusion Secure Fingerprint for Digital Images

Digital fingerprinting is a technique for identifying unauthorized copy and tracing back to its user. The distributor marks each individual copy with a unique fingerprint. The present fingerprinting schemes generally have many difficulties and disadvantages in the case of large-size users. In this paper, we present a new fingerprint scheme which can be used in confidential departments where there are a limited number of users. This scheme is composed of an outer RS (Reed-Solomon) code and an inner code based on BIBD (balanced incomplete block design) code and orthogonal code. We first construct inner code matrix based on BIBD block and orthogonal code. Since the BIBD code just has good performance in resisting linear collusion attacks, we expand the inner code to improve inner code’s ability of resisting nonlinear collusion attacks. In reality, the probabilities of participation in collusion in different combinations of users vary. So we group inner codeword into subsets and distribute each subset to a group of users. The code scheme turns into a type of grouporiented fingerprint which is favorable to trace colluders and to avoid involving innocent users. Then we take the inner codeword as the RS code’s symbols and make the outer code satisfy the request of resisting collusion attacks. In the end, by concatenating inner and outer code, we get the final fingerprint. The performance of the algorithms is proved and analyzed by theory. Compared with traditional codes for digital fingerprint, under the same collusion size and error probability, our scheme shortens digital fingerprint length significantly, and test experiments confirm the code’s ability of resisting collusion attacks.