Research On Dynamic Datasets Privacy Preservation Technology

Privacy preservation becomes more and more critical in many data applications, because a large number of privacy is threatened in the information organization and management of databases. For an individual, because of the job and life, it is necessary to provide its personal information to hospitals, banks, enterprises and other institutions generally. This information may be released to meet the operation of certain institutions and the requirements of scientific research. However, this information often contains sensitive information that individuals do not want to be known by others. If the data is leaked after the releasing of information, it will infringe upon personal privacy.Therefore, privacy protection has become a major topic of information security field. Anonymization is an effective approach to prevent privacy leakage, the anonymization technique can preserve privacy while guarantee the truth of the published data, efficient anonymization has attracted much research work, most of which, however, has been done on static dataset, which has no update and need only"one-time"releases. In other words, most of the privacy preserving algorithms do not support the re-publication of datasets after insertions, deletions and updates. However, most of the real world data sources are dynamic. Applying the existing static dataset privacy preserving techniques directly causes unexpected private information disclosure frequently.Recently, a few researchers begin to pay attention to anonymizing dynamic datasets. Meanwhile, there are some deficiencies in these researches. In many fields, some attribute values change probably but some not. For example, in the medical field, one person's disease may be converted into another or be cured over time. However, once suffering from a permanent disease, such as"Lung cancer", it can not be cured or converted to other diseases. Therefore, anonymizing dynamic dataset in such situation is more challengeable.This paper investigates the anonymization work of permanent diseases dynamic datasets using medical databases. Taking the typical current anonymization techniques as example, we discuss exhaustively various inference channels of serial releasing dynamic datasets on medical records using existing methods, and then propose an efficient algorithm on the idea of"invariance". The experimental results show that our method protects privacy adequately and has low information loss metric.