Query log perturbation method for privacy preserving query

Database as a Service (DaaS), a form of cloud computing, has recently attracted considerable attention. Users require their sensitive data to be protected from a database administrator that serves as a third party managing the data. We have proposed a secure query execution model for such an environment [10, 11]. Key features of our approach are to represent each tuple of each scheme as a plaintext table with one bloom filter index and to replace queries with keyword searches of the bloom filter index. In [12], we have defined an attack model in which attackers guess features of a plaintext table by observing bit patterns of the bloom filter index: further, we considered a defense against such as attack. We must also assume in this model that attackers can access query logs and may infer features of a plaintext table using such query logs. In this paper, we define an attack model by using query logs and propose a method to defend against the attack by executing fake queries.