Design and Analysis of High-Capacity Anti-Collusion Hiding Codes

Abstract

A new methodology for the design of high-capacity anti-collusion hiding codes in a large-scale fingerprint-based traitor tracing system is proposed in this work. We consider a hiding code of MN bits, where M bits are used as the user ID and N bits are the length of the spreading codes. Since each user is assigned one out of 2 M ID numbers and one out of N spreading codewords, the total number of users is equal to 2 M N. To accommodate an even larger number of users, we propose a shifted spreading scheme that shifts the spreading codeword circularly by a certain amount. By allowing P shifts (with P mod N), the total number of users increases from 2 M N to 2 M NP. When multiple users perform a weighted collusion attack, we show that the task of identifying colluders and their attack weights can be formulated as a user detection and channel estimation problem in a multiuser wireless communication system with a multipath fading channel. For the latter, there exist code design techniques that choose the spreading codes carefully so as to reduce multiaccess interference of users with different spreading codewords effectively. By exploiting this analogy, we develop an anti-collusion code called OSIFT (Orthogonal Spreading followed by the Inverse Fourier Transform). We compare several hiding codes in a fingerprinting system consisting of hundreds of colluders. It is demonstrated by computer simulation that, when the OSIFT code is adopted, colluders and their attack weights can be found by the proposed fingerprinting system more accurately.