Section 6.9. Summary of Database Security

6.9. Summary of Database Security

This chapter has addressed three aspects of security for database management systems: confidentiality and integrity problems specific to database applications, the inference problem for statistical databases, and problems of including users and data of different sensitivity levels in one database.

Both confidentiality and integrity are important to users of databases. Confidentiality can be broken by indirect disclosure of a negative result or of the bounds of a value. Integrity of the entire database is a responsibility of the DBMS software; this problem is handled by most major commercial systems through backups , redundancy, change logs, and two-step updates. Integrity of an individual element of the database is the responsibility of the database administrator who defines the access policy.

The inference problem in a statistical database arises from the mathematical relationships between data elements and query results. We studied controls for preventing statistical inference, including limited response suppression, perturbation of results, and query analysis. One very complex control involves monitoring all data provided to a user in order to prevent inference from independent queries.

Multilevel secure databases must provide both confidentiality and integrity. Separation can be implemented physically, logically, or cryptographically . We explored five approaches for ensuring confidentiality in multilevel secure databases: integrity lock, trusted front end, commutative filters, distributed databases, and restricted views. Other solutions are likely to evolve as the problem is studied further.

The emerging field of data mining shares some security problems with databases. Confidentiality, especially personal privacy, is a challenge to maintain, and inference across multiple databases is a further threat to confidentiality. Each database is usually owned and controlled by one party that is responsible for data correctness. Conversely, data mining applications often deal with multiple databases from different sources, which complicates ensuring the correctness or integrity of data mining results. Using comparable data items as keys becomes an issue with amalgamated databases being mined.

Many of the techniques discussed in this chapter are particular to database management systems. But the analysis of the problems and the derivation of techniques are typical of how we analyze security needs in any software application. In a sense, we must do a threat analysis, trying to imagine ways in which the security of the application can be breached. Once we conjecture ways to destroy integrity, confidentiality, or availability, we conjure up designs to help us build the security into the application's design, rather than after the fact. In the next chapter, we examine the security implications of another specialized form of application, networks.