Privacy loss and exploitation in e-commerce preference searching

• Main Author: Smith, Rhys
• Published: Cardiff University 2009
• Subjects: 020
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.584673

An area of e-commerce that is very much an active area of research is that of using an individual's preferences to enhance search. The development of this research area, and the model used to produce all existing methods, has an implicit assumption that the vendor to whom the consumer is releasing their preference information is trustworthy. This assumption results in two major issues: the certainty of privacy loss, and the potential for exploitation. Motivated by a wide ranging investigation into the concept and history of privacy and the methods used to protect it, along with the conclusion drawn from this investigation that the previously used methods of privacy protection via legal means can no longer keep pace with technological evolution, this thesis presents an alternative approach to searching with a consumer's preferences that enables the main goal of preference searching whilst also minimising privacy loss and the potential for exploitation. A proof of concept implementation of this approach, called "Gradual Partial Release", is presented. Essentially, its aim is to minimise privacy loss and exploitation by splitting a consumer's preferences up into multiple subsets of these preferences partial release - to be released one at a time to the vendor - gradual release - until sufficient results are returned. Three different Gradual Partial Release algorithms, that split up preferences into subsets in different ways, are presented, along with measures enabling quantitative measurement of privacy loss and exploitation to allow evaluation of their effectiveness. An evaluation was performed of the effectiveness and efficiency of the Gradual Partial Release algorithms, comparing the effectiveness (in terms of minimising of privacy loss and exploitation) of each algorithm and to the current approach to preference searching. Experiments show that the proposed Gradual Partial Release approach enables the basic idea of preferences searching whilst simultaneously offering the possibility of reduced privacy loss and reduced exploitation.