A Review of Ten Year Research on Query Log Privacy

The research on search log anonymization is ten years old. Over this time, a number of methods to reduce the risks of user identification and disclosure of sensitive information from search log analysis have been proposed. However, the impact of these findings on the behaviour of data owners and publishers has been very limited so far. In this paper, we present a brief overview and a classification of the main approaches in the literature, arguing that there has been a shift towards a more pragmatic balance between the value of the data published and the risk of an adversary breaching the user’s privacy. Then we go on to discuss what are the critical issues that must be tackled before search log anonymization moves out of research laboratories and into operational settings. We also clarify some misconceptions and mistakes in the use of the AOL search query log dataset, which is the main (and virtually only) experimental data source in the field.