Design of a system for recommendation based upon personal preferences and the preferences of others considering privacy issues

Any kind of information is available nowadays on the Internet or on TV. It is neither possible to consume all of the information nor one is interested in every type of information. On the other hand it may occur that interesting information was just missed. Recommendation systems based on former purchases are already known from large online stores. Since the collection of information consumption allows for creation of detailed personal profiles. Especially regarding information consumed on the Internet or on TV there arise privacy issues that may prevent people from using such system.

The task of the thesis is to

• evaluate existing recommendation systems
• investigate Bayesian Networks and other learning approaches
• evaluate centralized client-server approaches vs. P2P approaches
• design a solution that definitely considers privacy (anonymity)
  • privacy between users
  • privacy between user and central server, if such approach is used
• evaluate the solution
  • number of hits, false positives, false negatives

Contact Persons