Due to the large financial losses caused by illegal distribution of copyrighted materials, there is a strong necessity for the technology that would prevent such actions. In my diploma thesis I focused on the robust audio watermark and its resilience to attacks. I have described the tools for watermarking, watermark detection and watermark attacks. I tested the tools on three different, previously watermarked audio tracks. Thus I found out, that by embedding watermark, the distortion of the carrier signal is more significant in speech, while in music, especially classical, the distortion is almost negligible. Then I attacked those watermarks by noise, lossy transcoding, playback speed and pitch changing, as well as by analog loop channel. On the basis of estimates of the detectable audio distortion, which I carried out before and after the attacks, I was able to determine the most effective attacks. I also realised that the watermark is not resilient enough against attacks with the method of changing speed and pitch separately or proportionally, but with higher intensity. It turned out to be a big security hole, since after any such attack it is possible to preserve a relatively high-quality of audio track and to destroy watermark at the same time. I tagged the reasonableness of the watermark as questionable, because of the bad resilience to certain attacks, relatively detectable audio distortion and difficult infringement proving.
|
