Detecting malicious transactions using Fuzzy Association rule mining

In recent years, databases have become a very crucial part in all organizations and hence database security has become very essential. In order to protect organizational databases, intrusion detection systems (IDS) are deployed. Non-signature based IDS are found to be reasonable better than signature based IDS. In this paper, a new data mining based approach Fuzzy Association Data Dependency Rule Miner (FADDRM) has been proposed for detecting malicious transactions. The proposed anomaly based approach focuses on mining data dependencies between data items in the database using fuzzy association rule mining. The data dependencies are mined using the transactions from the database log. The transactions which are not compliant to the data dependencies are treated as malicious transactions. The proposed approach is exemplified using a data set for typical banking organization and the result shows that FADDRM can detect malicious transactions more effectively as comparison to other approaches cited in literature.