Privacy preserving release of blogosphere data in the presence of search engines

摘要

Users registered in a blogging platform and the subscriptions among them compose a social network with non-symmetric relations, whose data can be modeled as a directed graph. Release of such data for scientific analysis requires a pre-processing for ensuring no private information about people will be disclosed. The measures to be taken depend on the previous structural information a dishonest analyst is assumed to have. In this paper, the considered previous information is the sorting of blogs according to their PageRank relevance, which can be obtained by querying the blogging platform search engine. After analyzing the scenario, the n-rank confusion model is proposed. Experimental results show this model achieves a high privacy protection level while preserving the structural parameters of directed graph data to a high extent.