A Kind Of Anti-collusion Digital Fingerprint For Multimedia

These years, as an effective pirates tracing technology, the digital fingerprinting for multimedia attracts attention from more and more researchers. To solve some problems in existing multimedia fingerprinting schemes such as the longer length of fingerprint code, the influence of the noise brought by the processing after collusions, the resistance for minority collusion and so on, some researches on the fingerprint encoding and tracing are made. The main contents are as follows.Firstly, after analyzing the problems that the ECC-based multimedia fingerprinting scheme is not suitable for the multimedia carrying with smaller embeddable positions and the outer code encoding is more complicated, a novel concatenated fingerprint code for multimedia is proposed. The concatenated code is constructed by the convolutional code used as the outer code and the Hadamard code plus a dither sequence used as the inner code. The resistance against the interleaving collusion and the averaging collusion with the noise introduced by the processing after collusions is proved. Experimental results show that the code length of the proposed fingerprint code is less than the ECC-based fingerprint code under the settings of the smaller total user number and the smaller colluder number, and the proposed scheme has a better resistance to the averaging collusion.Secondly, to solve the problems that the binary random code can not resist the minority collusion performed by three users and the random noise could influence the collusion resistance of the fingerprint code, a random code based binary multimedia fingerprinting scheme is brought forward. The q-ary random code is used as the outer code and the extending binary random code constructed based the binary random code is used as the inner code which solves the problem that the binary random code can not resist the minority collusion performed by three users. The collusion resistance of the proposed scheme is analyzed under the settings that the number of colluders is far smaller than the total user number and the bit error rate is samller than 0.5. Experimental results show that the proposed scheme has larger code length but lower storage cost and computational complexity of detection, and when the bit error rate is less than 0.5, the scheme can trace at least a colluder with high probability under some kinds of general collusions.