Research On Anti-Collusion Digital Fingerprinting Technology

Coupled with the rapid development of information technology, piracy emergences everywhere. For the need in fighting pirate and protecting copyright of digital works, the idea of digital fingerprinting used to protect copyright has gotten more and more attention from internal and external investigators and come into the focus in their study.The main idea of fingerprinting is to distribute different copies of data with different fingerprints to different users. When the distributor found that one copy is illegally redistributed, he could identify the users who violated the protocol by the fingerprint trail. According to this idea, digital fingerprinting system performs fingerprint coding, fingerprint embedding, fingerprint extracting and fingerprint ''tracing. The extracting process always corresponds with the embedding process and the tracing process is based on the coding. In this sense we separate the fingerprinting problem into embedding and the coding problem, as the main contents in our paper.Firstly, we introduced the significance and development, and theory of fingerprinting technology in a whole and analyzed that the lower bound on the fingerprint length needed approximately grows logarithmically in the number of uses and linearly in the number of pirates in theory. In code design, we refer the length of fingerprint and the capability of anti-collusion as the criteria of evaluating fingerprint. After a thorough investigation, we discovered that traditional code such as c-secure code and I code have a good property of anti-collusion capability but the length is extremely large for most applications. BIBD code has a certain anti-collusion capability and reduces the length of fingerprint greatly, but it is hard to get and has a great expense on system. Then we presented a new fingerprint generating idea to combine BIBD code and I code. In our design, the fingerprint is easy to generate, the length of which is greatly reduced. It can resist many usual collusion attacks and trace the users in collusion efficiently. In fingerprint embedding, we adopted the DWT technique and spread spectrum algorithm that make the fingerprint embedded into data imperceptible and robust, and in the extracting process used correlating algorithm. Lastly, we simulated our fingerprinting system and the attack to our fingerprint as well on Visual C++ 6.0 platform. The experiment verified that our fingerprint-coding scheme has good capability to trace users who attacked the system by collusion in usual ways.