Efficient Anti-Scalable Anti-Collusion Fingerprinting

Digital fingerprinting schemes are techniques to protect the copyright of digital contents. One of the important problems is a collusion attack such that several users combine their copies of a same content to modify or delete the embedded fingerprint. Trappe et al. proposed the AND anti-collusion codes (AND-ACC) against the average attack. However, the scheme is vulnerable to linear combination attack (LCCA) and cannot support a large number of users. For this last issue, Seol and Kim highlighted a code (called SK) by extending the AND-ACC. Unfortunately, that code is weak against majority attack and LCCA. In this paper, we improve the SK scheme by adding to each group the subcode. The new code can resist average attack when the inter-group collusion is less likely than intragroup collusion due to geographic conditions. Non-blind detection statistics with the knowledge of the host and soft-threshold detection are used to identify colluders within each guilty group. Our model increases the probability of tracing )) ( log ( 1 N n O  colluders within each guilty group, where N is the fingerprinting length and n is the number of users in each group. Experiment results on the real images show that our code is robust to average attack.