Developing realistic attack models for privacy preserving record linkage and algorithms to prevent such attacks

点击翻译按钮获取中文摘要

Analyzing large scale data sets (big data) is gaining importance in the Social Sciences and Official Statistics. The analysis of such data often requires linking multiple data sets. In many countries, this linkage has to be done without a unique personal identification number. This process is called record linkage in statistics and computer science. Record linkage under the special restrictions given by European law and the federal and non-central organisation of data protection in Germany requires special techniques (privacy preserving record linkage). Previous approaches are not based on realistic attack models. The work program of this project is the development of such attack models and algorithms to prevent these attacks. The project will define quality requirements for privacy linkage and develop a formal security model. The analysis of existing proposals with regard to the formal security model will be followed by the development of new safeguards. The proposal describes three previously unpublished methods. Existing and newly developed procedures will be studied mathematically, with simulated and with real world data. The goal of the project is the development of cryptographically secure privacy preserving record linkage techniques for large data sets.