The new release is available in source or binary form from the Flourish Player project web site or directly from SourceForge.

The Fraunhofer SIT MediaSearch Framework is a loosely coupled, event based software system that searches internet platforms such as BitTorrent, Rapidshare or YouTube for media that match specified search criteria, downloads them and examines them for the presence of digital watermarks. Since the amount of media available on the internet is huge, the system is required to be scalable so that large amounts of data can be processed simultaneously.

The scalability of the MediaSearch Framework faces several possible impediments that we will examine in this thesis. Especially noteworthy are the requirements of high network bandwidth for downloads and large computing resources for the retrieval of watermarks. The required scalability might be achieved by distributing the execution of the system to multiple machines to utilize their aggregate bandwidth and computing power.

In this thesis, we examine possible distributed architectures and their implications on the Framework and its scalability. We also discuss impacts on the reliability of the system as a whole and countermeasures against failures of parts of the system. Finally, we propose an implementation of a distributed architecture for the MediaSearch Framework, evaluate its benefits and shortcomings, and show that it is a feasible operation platform for complementing passive copyright protection using digital watermarks with an active search component.

If you wish to receive a full electronic copy of my master’s thesis, please feel free to contact me via e-mail.

First and foremost: What I tried first was using the @PostConstruct annotation on a method that should be called after deployment – but that doesn’t work, because the EJB is not guaranteed to be instantiated directly after deployment. As a matter of fact, at least the JBoss I was using lazily instantiates the EJB the moment it is first invoked.

After a while, I came across this thread in Sun’s forum, in which a Sun developer states that better support for this use case is planned for a future EJB version, and currently the only way to achieve the desired effect is to rely on a ServletContextListener, which provides a contextInitialized() method that is called as soon as a servlet context is initialized. However, like the “servlet” implies, there needs to be a web application to use this listener – so I ended up adding a web application to my enterprise application with no content besides that context listener that would poke the EJB in order to get started. Now although this method is far from elegant, at least it works reliably and most possibly independent of the container used; at least I can verify its functionality on the JBoss 5 used in my test system.

For my interest and somewhat deeper knowledge in watermarking especially of audio content, I found the following talks very enlightening (in order of the conference):

• Supraliminal Audio Steganography by Heather Crawford and John Aycock: Although their approach to audibly hide information in audio files using sounds from the domain resp. genre of the audio content itself is far from being perfect at the moment, it seems to be a promising approach for steganography at the semantic level that could prove quite difficult for conventional steganography to detect.
• An Epistemological Approach to Steganography by Rainer Böhme: A very illustrative discussion of the empirical nature of steganography (and also other disciplines of information hiding) and the need for a separation between models of the steganographic channels and the steganographic algorithms themselves.
• A Phase Modulation Audio Watermarking Technique by Michael Arnold, Peter G. Baum, and Walter Voeßling: Using a watermarking method modulating phase information in the Fourier Domain, the authors attacked the challenge to actively measure audience rates for broadcast media. The MPEG psycho-acoustic model is used to determine embedding thresholds in order to ensure maximum audio quality.
• Perception-based Audio Authentication Watermarking in the Time-Frequency Domain by my co-workers at Fraunhofer SIT, Sascha Zmudzinski and Martin Steinebach: The integrity of audio content is protected by extracting features forming a robust perceptual hash that is in turn embedded into the content itself using a watermarking technique. This way, the authenticity of the content can be verified, even after common processing such as lossy encoding or analogue transfer.

• Aspect Oriented Programming (AOP)
• Object Manager featuring dependency injection
• MVC framework
• Persistence framework

…and many more.

The framework seems to be very interesting and sophisticated, and although I couldn’t yet find time to try it, I will definetly do so in the next time. The framework is developed in conncetion with the TYPO3 content management system and can thus be expected to be of high quality.