Detection of malicious transactions in databases using dynamic sensitivity and weighted rule mining

The development of an Intrusion Detection System for Database Security has grown into an undeniable necessity in the past few years. In order to maintain scalability and dynamicity of database systems, detecting privilege abuse is of foremost importance. Most researchers have presented either a static or database dependent solution to the existing problem of Insider Attacks. Our Approach, Dynamic Sensitivity Driven Rule Generation Algorithm (DSDRGA) detects intrusive transactions and therefore safeguards crucial data from modification. Sensitivity of data attributes and the data dependency rules are dynamically determined by mining frequent sequences from normal user data access patterns retrieved from the audit log. The extent of weighted similarity between the mined rules and the incoming transactions is used to classify the transaction as malicious or normal user behavior. DSDRGA gives promising results in the detection of malicious transactions when evaluated on a characteristic banking dataset.