This thesis analyses the performance of the open-source tool AWT2 at working with watermarks, powered by intercalation. Part of the focus of this thesis will be to examine the difficulties in perceiving a digital watermark in selected music files with emphasis on the analysis of detection and of the stability of the watermark depending on the genre of music when choosing different types of attacks.
Besides having a variety of positive effects on societal development, internet and digital technologies raised an issue of illegal distribution of digital audio formats. Broadband Internet connections enable downloading large amounts of data with almost no loss, which can enable unlimited distribution and digital copying of multimedia content without adequate supervision. Efficient protection of intellectual property (a music file) against intentional and unintentional alteration and reproduction is a problem that is partially addressed in this thesis and needs further discussion.
Digital watermark technology can offer one of the solutions for addressing uncontrolled distribution of digital content. By using existing patented software tool AWT2 to insert a watermark and its digital payload in music files, we want to test the operating effectiveness of the program. We have produced a Matlab script, which was used to automate the entire process, from the initial embedding of watermark to the performance of different types of attacks and detection of the watermark. At the end, we analyzed the performance of perception and the acquisition of the watermark from the attacked music files, taking into account five musical genres. Firstly, we evaluated the analysis of the performance of the open-source program AWT2 according to the number of obtained watermarks from music files, and secondly, we analyzed the performance of perception and the acquisition according to the choice of musical genre in connection with selected types of attacks.
In the empirical part of this thesis we performed Matlab script and analyzed the results. We have shown detailed performance of the script after the main blocks, explained the selected types of attacks and established a framework of operation and interpretation of the results. The results were then compared and statistically presented in tables and graphs contained herein for better understanding and overview of the data.
On average, the performance of embedded watermarks in selected music files was good, which means that the program AWT2 successfully completed its task. In the first phase of the empirical part of this thesis, when we did not use the selected attacks, the success of the perception and the acquisition of the embedded watermark is 94%. When we selected the normal mode, the musical genre Latin stood out the most. In the second phase of the empirical part this thesis, with the use of different types of attacks, the results differed according to the musical genre. At the N1b attack we did not detect the watermark in more than half of musical files. On the other hand, at the N2a attack we did not detect even one watermark. We have found that certain types of music (Latin) have an impact on the results of the program AWT2 (increased undetectable watermark), which is reflected in the increased representation of certain musical genres at deteriorating performance of perception of an attacked digital watermark. Therefore we can conclude that success of detection of the watermark depends also on the selection of the musical genre.