Access and privilege in secure big data analysis

Abstract

The distributed data sources and strict security controls of the Enterprise Level Security (ELS) architecture present challenges for data mining. The ELS architecture is a secure enterprise system that enforces strict security controls in a uniform way across an enterprise. It includes end-To-end bilateral authentication for all human as well as machine interactions and verifiable claims-based access controls. Claims provisioning is automated and centrally managed based on authoritative attributes of active entities in the enterprise. While these security provisions are necessary for secure systems, they present some unique challenges to big data analyses. Key among these are non-standard schemas, non-standard access and privilege, restricted access to analysis outcomes, and overall privilege handling. Some of the distributed data sets may be fully or partially accessible, or even not accessible. Users with limited access may compute different results than those with broad access. We discuss the problems encountered for data mining in an ELS architecture and possible solutions.