http://www.cs.umanitoba.ca/~arindam/pubs/MiM07_aina.pdf

Peer-to-peer (P2P) networking is the primary method by which pirated music is disseminated.  To combat this epidemic, the Recording Industry Association of America (RIAA) has begun to seed corrupt files that track the file downloader’s IP address so that he may be prosecuted. 

The similarity between file sharing via P2P networks and the information cascades we discussed in class is that when one hosts a file, others will see that he has downloaded the file and will be more likely to download that file.  But this is problematic because there is a high potential for spreading corrupt files, especially if the RIAA wishes to flood the P2P network with “users” hosting the said files. To combat this, some popular P2P networks such as Kazaa have begun to implement file-rating systems, which are meant to speak of the file’s integrity.  But the problem with their rating system’s implementation is that each user’s input has the same weight on a file’s rating.  This is obviously silly because there is no way of knowing the reviewer’s credibility.  It could be the case that the RIAA hires people to assign high ratings to corrupted files.

To improve on this method of rating, the authors of the paper have created a system in which “As the stories (files) spread, each node makes an accept, reject, and forward decisions based on the credibility measure which is computed from the estimates of the originating node’s trust and the particular story’s trust.”  To estimate the node’s trust level, Baye’s Theorem, which we discussed in class, is used. 

It is clear that a rating system that uses conditional probability will be far superior to one that does not.  Because we have a way of knowing the reviewer’s credibility along with his review of the file, by Baye’s Rule, it is more likely that a file with a high rating is of good integrity than if we were to just use the rating system that Kazaa uses.