Using Search Results to Microaggregate Query Logs Semantically

Abstract

Query log anonymization has become an important challenge nowadays. A query log contains the search history of the users, as well as the selected results and their position in the ranking. These data are used to provide a personalized re-ranking of results and trend studies. However, query logs can disclose sensitive information of the users. Hence, query logs must be submitted to an anonymization process to guarantee that: (a) no sensitive information can be linked to an identity; (b) the analysis of the anonymized data produces similar results than the original data, i.e. minimize data distortion. Latest anonymization approaches utilize microaggregation, a statistical disclosure control technique that provides a privacy comparable with $$k$$ -anonymity, attempting to minimize the data distortion. We propose a new method that uses search results to optimize microaggregation, providing more data reliability than the existing methods.