Models of multilevel databases security

Abstract

Multilevel security - is a security policy that allows to classify objects and users based on a system of hierarchical security levels and use this classification to organize an access control system. In relational data bases with multi-level security, any user reading or updating data in a table should be allowed to process only those lines that allow its access level represented by the security label. For each row in the table (or row attrib-ute), the privacy level is set by assigning a security label. The user can read or modify the row only if its label dominates over the label of the row. This requirement leads to the fact thatthe same domain object can be represented in the table by several records, each accessible only to users with the corre-sponding security label (the property of multi-valued rela-tions). Whereas the fundamental principles of relational data-bases building require the uniqueness of each tuple relation-ship. The way to resolve this contradiction is determined by the security model used. In addition, the multi-valued rela-tionship leads to the emergence of vulnerabilities in the form of hidden channels (covert channels), obtaining information through inference channels, semantic ambiguity and others. As an investigation direction in the field of database security, the technology of a multilevel secure database is developing rapidly. Many models of multilevel security in RDBMSs have been developed based on the Bell-Lapadul model, such as the SeaView model, the Jajodia-Sandhu model, the Smith-Wins-lett model and others that would completely or partially solve arising problems like hidden channels, semantic ambiguity, and others. However, no flawless solution or model has been proposed to date. Objective: SeaView, Jajodia-Sandhu, Smith-Winslett models analysis, identification of their advantages and disadvantages.