Mining Inter-transaction Data Dependencies for Database Intrusion Detection

Existing database security mechanisms are not sufficient for detecting malicious activities targeted at corrupting data. With the increase of attacks toward database-centered applications, an effective intrusion detection system is essential for application security. Although someresearches havebeen done on the database intrusion detection, methods for detecting anomalous activitiesin databases haveonly recently been explored in detail. In this paper, we present an approach employing inter-transaction data dependency mining fordetecting well-crafted attacks thatconsists a group of seemingly harmless database transactions. Our experiments illustrated the advantage of this new approach and validated the effectiveness of the model proposed.