Improving security by using a database management system for integrated statistical data analysis

Abstract

International research collaborations access and integrate data collected in different countries. For different reasons, e.g., legislation, data owners need to control who has access to and how their data are analyzed. The analysis of data is performed in statistical software, which is usually called on top of a data management system, e.g., a database management system (DBMS). Therefore access to data is controlled by the DBMS, while statistical analyses are usually controlled by another system. To improve security we propose a novel architecture for executing statistical analysis on data stored in a DBMS. In the proposed architecture the statistical software is called from a DBMS. The architecture allows control of both data retrieval and statistical data analysis from one system, i.e., DBMS. We implemented a prototype for executing analysis programs by calling statistical software SAS from a relational DBMS IBM DB2 over data stored in DB2 database. This paper describes the proposed architecture and the implemented prototype.