How to Share Your Favourite Search Results while Preserving Privacy and Quality

摘要

Personalised social search is a promising avenue to increase the relevance of search engine results by making use of recommendations made by friends in a social network. More generally a whole class of systems take user preferences, aggregate and process them, before providing a view of the result to others in a social network. Yet, those systems present privacy risks, and could be used by spammers to propagate their malicious preferences. We present a general framework to preserve privacy while maximizing the benefit of sharing information in a social network, as well as a concrete proposal making use of cohesive social group concepts from social network analysis. We show that privacy can be guaranteed in a k-anonymity manner, and disruption through spam is kept to a minimum in a real world social network.