Statistical database modeling for privacy preserving database generation

摘要: Testing of database applications is great importance. Although various studies have been conducted to investigate testing techniques for design, relatively few efforts made explicitly address the which requires a large amount representative data available. As over live production databases often infeasible in many situations due high risks disclosure confidential information or incorrect updating real data, this paper we problem generating synthetic based on a-priori knowledge about database. Our approach fit general location model using characteristics (e.g., constraints, statistics, rules) extracted from and then generate learnt. may contain be used by attacker derive some information, present analysis method cell suppression technique. effective efficient remove aggregate private during generation.