INFERENCE THROUGH SECONDARY PATH-ANALYSIS

Inference, the ability to deduce classified information from unclassified information1, is a leading security issue in the field of Multilevel Secure Database Management Systems (MLS DBMSs). Trusted systems are designed to prevent an unauthorized flow of data. However, inference techniques gather unauthorized information in a seemingly proper manner. Standard models for controlling information flow (i.e.: Bell-LAPadula) cannot detect unauthorized access gained by utilizing inference techniques. This paper addresses a specific class of inference attack: inference through secondary path analysis. This method of inference attack is characterized as one which seeks out alternative paths between two attributes whose primary path is classified. Our goal is to formally define this type of attack as a preamble to its solution.