Adaptive Collusion Attack to a Block Oriented Watermarking Scheme

In this paper, we propose an adaptive collusion attack to a block oriented watermarking scheme [1]. In this attack, traitors conspire to selectively manipulate watermarked blocks to remove the watermark information. To this end, the traitors compare the watermarked blocks generated from the same original block. If two watermarked block are not equal, they average these two blocks to generate a pirated block. Then, replace the watermarked blocks with the pirated blocks so as to build a pirated image. The pirated image has no watermark but has much higher quality than watermarked images. We also give a theoretical analysis on the probability of successful traitor tracing. Both theoretical and experimental results demonstrate that our attack is very effective when four or more traitors are involved in the collusion attack. In the cases of less than four traitors, we show how to integrate our collusion attack with an adaptive random attack to improve the qualities of pirated images as wel...

• Type: Conference
• Year: 2003
• Where: ICICS
• Authors: Yongdong Wu, Robert H. Deng
• Keywords: Collusion Attack, ICICS 2003, Pirated, Watermarked Blocks