Sensitive knowledge hiding in tree and graph databases

Database sharing is a beneficial process which organizations sometimes need to do. Although it is a good practice, it may threaten the database security through disclosing sensitive knowledge. This is because sophisticated data mining tools nowadays are so developed that running any of the tools on published database may disclose the sensitive knowledge implied by the database. As a result, database publishing is not a secure way of database sharing. Hence, we reason that the sensitive knowledge in database must be firstly identified then it must be removed. The process is called the sanitization. Sensitive knowledge hiding is extensively studied mostly in the context of transactions. However, it needs to be studied for tree and graph structured databases as well. In this thesis, the sensitive knowledge hiding is extended for tree and graph databases. This work defines respective problems and develops solutions for both of them. Moreover, a framework, called FISHER, is developed for sensitive knowledge hiding which is able to hide sensitive knowledge from various kinds of other databases as well, including transactions, sequences, and spatiotemporal databases.