Enhancing Utility and Privacy of Data for Software Testing

A fundamental problem in test outsourcing is how to allow a database-centric application (DCA) owner to release a smaller subset of its private data along with the application. In addition, the DCA owner needs tangible guarantees that the entities in this data are protected at a certain level of privacy, while retaining testing efficacy. We are trying to solve this problem by balancing four important dimensions: testing coverage, privacy, semantic correctness, and data minimization. We built a novel approach that enhances both utility and privacy of data for software testing using a novel combination of program analysis, clustering, and association rule mining approaches. To the best of our knowledge, there exists no prior approaches that synergistically address all of the aforementioned dimensions. We also proposed several avenues for future work based on our existing work.