Slide #3 (Introduction to Rights Management)
  At the current stage of Information Age, information can be much more
  easily copied than it could be before. This worries
  writers and artists, because their intellectual properties can now
  be hard to protect. What rights management tries to do is to protect the
  intellectual property rights holder from fraud. Rights management
  defines the relationship between "Intellectual Property", its
  "Creator", and the "Consumer". It specifies what the consumer
  is allowed to do with the data (to view, to print), and how much each
  of these usages costs.

Slide #4 (Case Analysis - Document Distribution)
  Next, let us look at an example scenario. Alice is an
  artist, and sells Bob a digital document over the Internet. Bob's
  concerns are, as a buyer, whether or not the information received is
  what he wanted, and also if the information is correct. His 
  concerns can be addressed by non-repudiation. The bigger problem
  here, however, is what to do to protect the artist or the seller, in
  this case, Alice. 

  Alice is interested in controlling the distribution of her art,
  protecting herself from fraud or copyright infringements, and
  defining allowed usage patterns. To address these concerns, she can
  attach copyright information to the digital data. This information should
  be separate, but not separable from the data. As a safety net, Alice
  is also interested in legal protection, in case of fraud. 

  One of the problems is that, in the real world, we have third party
  attackers. They can, for 
  example, eavesdrop a line, and distribute the stolen
  information. They can also change the information (for instance,
  stock market data), or commit plagiarism (we have seen examples of
  plagiarism of papers authored by people here at CMU), and
  redistribute the material under their own names.

Slide #5 (Case Analysis - Protecting Images and Text)
  We have two cases here, depending on what we want to protect. Let us
  go over the list for image watermarking briefly. 

  As I said before, we would like the copyright information to be
  separate from the content. In fact, we would like it to be
  invisible. But we would like also to keep it inseparable from the
  image, so that nobody can remove the watermark. This leads us to
  the second point, which says that if you remove the
  watermark, then, with high probability, you will make changes to the
  image. These changes result in low quality images, which do not sell
  as well as the original ones. 

  We would like also to have robustness, which means that we would
  like watermarks to survive common image manipulations, such as lossy
  compression, cropping, filtering, etc. 

  In the case of text, things become more difficult. In the past,
  people have used different methods. One possibility is to treat text
  as an images. We can embed the data by changing the inter-line
  spacing, or spacing between characters or words. But this is not
  robust against attacks where the text is submitted to an OCR
  (Optical Character Recognition) system,
  and reprinted. This method is not useful for searching or handling
  texts either. We will see later how current schemes protect text.

Slide #6 (Unsolved Copyright/Legal Issues)
  In the real world, we have other problems. Even if we had perfect
  technical solutions, people can still 
  infringe copyright laws and get away with it. For instance, we can
  have a server on the Internet, and a country that does not follow
  the Berne convention will allow anyone to freely distribute the
  information on the server, without any legal action taken against
  the criminal. This happens, for instance, in Indonesia, where
  complete tracks of audio CDs are extracted and used illegally. 

Slide #7 (Different Models and Discussion of Problems)
  Next, we show different models to deal with the problem. 

  First, we have the electronic commerce framework, which does not
  really deal with the problem. 

  Second, we have super-distribution. The idea there is that anybody
  can acquire or distribute copyright protected information. But
  you are charged as soon as you use the information. This is called
  usage-based revenue collection scheme. In the super-distribution
  scheme, we also have the idea of "composite objects", where an
  object can have different parts (images, text, audio, etc), and the
  viewer pays for what he or she chooses to use. 

  And lastly, we have the Intertrust approach, which uses language
  features to express things that can be done with a document, and
  uses tamper-resistant boxes for secure executions. (The audience was
  skeptical about how well this scheme could work.)

  (The details of the second and the third approach were not given in
  the talk.)