Detecting malicious transactions in database using hybrid metaheuristic clustering and frequent sequential pattern mining.

Database systems have become imperative for organisations around the world to store and analyse information. However, as one of the ramifications of a massive surge in cloud-based activities and interactions brought forth by the advent of the internet era, the data is exposed to an audience broader than ever and there are a variety of new challenges putting database security in jeopardy. To be able to address the problem of data security, we propose a unique method for Database Intrusion Detection System build on frequent sequential pattern mining and a modified metaheuristic hybrid clustering of Grey Wolf and Whale optimization algorithm to determine malicious transactions in Role Based Access Control and non-RBAC supervised databases. Our proposed approach extracts data dependency rules from the database logs using CM-SPADE mining algorithm to detect outsider threats. It then assigns role profiles to the users based on the previous user activities using the modified metaheuristic clustering to detect insider threats. Thereby, identifying incoming transactions as malicious by matching the role profile of the user and comparing the adherence of the transaction pattern to the generated dependency rules. To evaluate the efficiency of the model we generated a synthetic dataset including malicious and non-malicious transactions adhering to the TPC-C benchmark, and the findings were encouraging, with levels of accuracy of around 97.8 percent. [ABSTRACT FROM AUTHOR]