A NEWWATERMARKING ATTACK USING BLIND SOURCES SEPARATION
Manal Taoufiki · International Conference on Digital Information Processing and Communications · 2013
In this paper, we show that the watermarking method based on sources separation can be easily detected and destroyed. Indeed, the previous method is based on a simple linear mixture between the watermarks (signatures) and the original signal (the signal to be protected) with grossly different proportions. In this paper, we propose an attack based on blind sources separation that has the ability to extract the signatures independently from the original signal using just one observed mixture.