Mechanisms for database intrusion detection and response

Data represent today a valuable asset for companies and organizations and must be protected. Most of an organization's sensitive and proprietary data resides in a Database Management System (DBMS). The focus of this thesis is to develop advanced security solutions for protecting the data residing in a DBMS. Our approach is to develop an Intrusion Detection and Response (IDR) system, integrated with the core DBMS functionality, that is capable of detecting and responding to anomalous SQL commands submitted to a DBMS. For the intrusion detection mechanism, the key idea is to learn profiles of database users from the SQL commands submitted by them to the DBMS. A SQL command that deviates from these profiles is then termed as anomalous. For responding to such anomalous and potentially malicious SQL commands, we introduce a policy-driven intrusion response mechanism that is capable of issuing an appropriate response based on the details of the anomalous request. Such response actions include fine-grained actions such as request suspension and request tainting; we introduce an access control system based on the notion of privilege states to support such fine-grained responses. For the management of the response policies, we introduce a joint threshold administration model that mitigates the risk of insider threats from malicious database administrators. A major component of the thesis involves prototype implementation of the IDR mechanism in the PostgreSQL DBMS. We discuss the implementation details on the same and report experimental results that show that our techniques are feasible and efficient.