Applying Epsilon-Differential Private Query Log Releasing Scheme to Document Retrieval

Web search logs are valuable and widely used for improving Information Retrieval (IR) research. However, these query logs contain sensitive data, which makes them difficult to be released directly even for research purposes. More companies can release their query logs if adequate privacy protection can be placed. This workshop paper introduces our research project on this privacy preserving query log releasing problem. In this paper, we propose a framework using differential privacy on query logs to guarantee high levels of privacy which achieves -differential privacy. We also provide a proof for why the user IDs for each of the individual search records can not be released in order to achieve differential privacy. Experiments show that our approach is a strong option for maintaining both high privacy and high utility. We believe this project is an important step forward to find practical solutions to this query log releasing problem.