Two-phase entropy based approach to big data anonymization

Data outbreak following the flourishing of technologies like cloud computing, social networks, and many more stimulate emergence of new paradigm called “Big Data”, which not only improved the decision-making, but also provided a means to easy privacy violation of an individual. Existing data anonymization techniques adhering to certain privacy model such as k-anonymity or l-diversity proved their worth, but demands identifying potential quasi-identifier set manually by the domain expert which is very much prone to human error. The paper focus on the problems of choosing appropriate quasi-identifier set and minimizing information loss due to anonymization process. The proposed approach is two phase entropy driven approach to find the quasi-identifier set in a first phase, which is extended to second phase to anonymize the dataset based on selected quasi-identifier set. Our approach ensures that dataset maintains its statistical properties along with minimal information loss and better privacy.